MST-004 MST-004

STATISTICAL
Indira Gandhi
National Open University INFERENCE
School of Sciences

Block

1
SAMPLING DISTRIBUTIONS
UNIT 1
Introduction to Sampling Distribution 7
UNIT 2
Sampling Distribution(s) of Statistic(s) 29
UNIT 3
Standard Sampling Distributions-I 53
UNIT 4
Standard Sampling Distributions-II 69
Appendix 91
Curriculum and Course Design Committee
Prof. K. R. Srivathasan Prof. Rahul Roy
Pro-Vice Chancellor Math. and Stat. Unit
IGNOU, New Delhi Indian Statistical Institute, New Delhi

Prof. Parvin Sinclair Dr. Diwakar Shukla

Pro-Vice Chancellor Department of Mathematics and Statistics
IGNOU, New Delhi Dr. Hari Singh Gaur University, Sagar

Prof. Geeta Kaicker Prof. Rakesh Srivastava

Director, School of Sciences Department of Statistics
IGNOU, New Delhi M.S. University of Baroda, Vadodara

Prof. Jagdish Prasad Prof. G. N. Singh

Department of Statistics Department of Applied Mathematics
University of Rajasthan, Jaipur I.S.M., Dhanbad

Prof. R. M. Pandey Dr. Gulshan Lal Taneja

Department of Bio-Statistics Department of Mathematics
All India Institute of Medical Sciences M.D. University, Rohtak
New Delhi
Faculty members of School of Sciences, IGNOU
Statistics Mathematics
Dr. Neha Garg Dr. Deepika
Dr. Nitin Gupta Prof. Poornima Mital
Mr. Rajesh Kaliraman Prof. Sujatha Varma
Dr. Manish Trivedi Dr. S. Venkataraman

Block Preparation Team

Prof. G. K. Shukla (Editor) Mr. Prabhat Kumar Sangal (Units 1 - 4)
Decision Science Group School of Sciences, IGNOU
Indian Institute of Management
Lucknow (UP) Dr. Manish Trivedi (Units 3 & 4)
School of Sciences, IGNOU
Dr. Parmod Kumar (Language Editor)
School of Humanities, IGNOU

Course Coordinator: Mr. Prabhat Kumar Sangal

Programme Coordinator: Dr. Manish Trivedi

Block Production
Mr. Sunil Kumar, AR (P),School of Sciences, IGNOU
CRC prepared by Mr. Prabhat Kumar Sangal, School of Sciences, IGNOU

Acknowledgement: I gratefully acknowledge my colleagues Dr. Manish Trivedi and Mr. Rajesh
Kaliraman, Statistics Discipline, School of Sciences for their great support.

July, 2013
© Indira Gandhi National Open University, 2013
ISBN-978-81-266-
All rights reserved. No part of this work may be reproduced in any form, by mimeograph or any other
means, without permission in writing from the Indira Gandhi National Open University.

Further information on the Indira Gandhi National Open University may be obtained from University’s
Office at Maidan Garhi, New Delhi-110068 or visit University’s website http://www.ignou.ac.in

Printed and published on behalf of the Indira Gandhi National Open University, New Delhi by the
Director, School of Sciences.

Printed at:
STATISTICAL INFERENCE
In MST-003, the basic rules of probabilities and various probability
distributions such as Bernoulli, Binomial, Poisson, Uniform, Normal,
Exponential, etc. have been discussed. Now, the concepts of probability theory
and probability distributions are used in drawing inference about the population
parameters.
Generally, population parameters are unknown and when the population is too
large or the units of the population are destructive in nature or there is a limited
resources and manpower available then it is not possible practically to examine
each and every unit of the population to obtain the population parameters. In
such situations, one can draw sample from the population under study and
utilize sample observations to draw reliable conclusions about the population
parameters. The results obtained from the sample are projected in such a way
that they are valid for the entire population. This technique is known as
statistical inference. The statistical inference may be divided into two areas or
parts:
(i) The population parameters are unknown and we may want to guess the
true value of the unknown parameters on the basis of a random sample
drawn from the population. This type of problem is known as
“Estimation”.
(ii) Some information is available about the population and we may like to
verify whether the information is true or not on the basis of a random
sample drawn from the population. This type of problem is known as
“Testing of hypothesis”.
This course is divided into four blocks and each block consists of four units.
In Block 1, we shall be discussing the concept of sampling distribution of a
statistic. The sampling distributions of sample mean, sample proportion,
sample variance, difference of two sample means, etc. are also discussed in this
block. Here, we also discuss some important sampling distributions which are
widely used in statistical inference and known as exact sampling distributions
such as χ2, t and F.
In Block 2, we discuss first part of statistical inference called estimation,
through which we find the estimate of unknown parameter on the basis of
sample data. Two types of estimation as point estimation and interval
estimation are discussed in this block. The criteria of good estimator and
different methods of estimation for the unknown parameters are also discussed
in this block.
In Block 3, the second part of statistical inference, that is, testing of hypothesis
is discussed. We discuss how the estimated or assumed or a hypothetical value
of the population parameter can be tested and how we can take the decision
about rejection or non-rejection of the hypothesis or assumed value. Here,
different parametric tests as Z-test, t-test, F-test and χ2-test for testing of
hypothesis about population mean, difference of two population means,
population proportion, difference of two population proportions, population
variance, etc. are also discussed.
Most of the parametric tests are based on the assumptions that the parent
population is normal and the data are measured in interval or ratio scale and
concerned with testing the population mean, proportion, variance, etc. But if
the assumption(s) is (are) not fulfilled then we cannot used the parametric test.
In such situations, non-parametric tests do the same job for us. In Block 4, we
discussed different non-parametric tests i.e. sign test, Wilcoxon signed-rank
test, chi-square test for goodness of fit and independence of attributes,
Kolmogorov-Smirnor test for goodness of fit, Kruskal -Wallis test for one way
analysis of variance with completed randomised design, Friedman test for
randomised block design, etc.
Although the material is self contained and self explained in nature, even
though, if the learners are interested to gain more and want to study the
contents in more detail, you may consult the following books:
 Goon, A.M., Gupta, M.K. and Dasgupta, B.; An Outline of Statistical
Theory, Vol II, World Press, Calcutta.
 Rohatgi, V.K.; An Introduction to Probability and Statistics, Wiley-
India.
 Mood, A.M.; Introduction to the Theory of Statistics, Tata McGraw-
Hill Book Company, New Delhi.
 Gupta, S. C. and Kapoor, V. K.; Fundamentals of Mathematical
Statistics, Sultan Chand & Sons.
 Makhopadhyay. P.; Applied Statistics, New Central Book Agency,
Calcutta.
 Gibbons, J.D.; Nonparametric Statistical Inference, McGraw-Hill Book
Company, New York.
 Daniel Wayne W.; Applied Nonparametric Statistics, Houghton Mifflin
Company.
 Conover, W.J.; Practical Nonparametric Statistics (3rd Ed.), Wiley-
India.
SAMPLING DISTRIBUTIONS
One of the major objectives of statistical analysis is to draw reliable
conclusions about the population on the basis of the analysis of the sample
data. This process is known as statistical inference. For drawing the inference
about the population parameter, we use a function of sample values such as
sample mean, sample proportion, sample standard deviation, etc. which is
known as statistic. If we draw all possible samples of same size and for each
sample we calculate a statistic then the value of statistic may or may not be
varying sample to sample. If we arrange all possible values of that statistic with
its corresponding probabilities then this arrangement is known as sampling
distribution of that statistic.
This is the first block of the course MST-004. The aim of this block is to put a
foundation stone for the next blocks of this course and will provide a platform
to the learners to understand the statistical inference. It comprises four units.
Unit 1: Introduction to Sampling Distribution
This unit explains the basis concept of sampling distribution of a statistic with
examples. Here, we also discussed the most important theorem of Statistics
“Central Limit Theorem” according to this theorem, the sampling distributions
tend to be approximately normal for sample sizes greater than 30. In this unit,
we also discussed the law of large numbers. According to law of large
numbers, we come to know “how we draw a reliable inference about the
population parameter by using a sample of finite size?”
Unit 2: Sampling Distribution(s) of Statistic(s)
This unit explores the sampling distributions of sample mean, sample
proportion, sample variance, difference of two sample means, and difference of
two sample proportions and ratio of two sample variances.
Unit 3: Standard Sampling Distributions-I
Sometimes, statistic may follow a particular sampling distribution such as χ2
(read as chi-square), t, F, etc. which are known as exact sampling distributions.
This unit provides the brief introduction of chi-square and t-distributions.
Properties and applications of these distributions also explained in this unit.
Unit 4: Standard Sampling Distributions-II
Last unit of this block is devoted to the rest important sampling distribution
which is widely used in Statistics i.e. F-distribution. Here, we discuss the
properties and applications of this distribution. The standard sampling
distributions are interacted with each other, therefore, the relations between
them are also explored in this unit. In this unit, we also describe the procedure
of finding the tabulated (critical) value of a variate which follows the exact
sampling distribution such as χ2, t and F.
Notations and Symbols
X1, X2, …, Xn : Random sample of size n
SRSWR : Simple random sampling with replacement
SRSWOR : Simple random sampling without replacement
X : Sample mean
S2 : Sample variance
SE : Standard error
2
N (µ, σ ) : Normal distribution with mean µ and variance σ2
Z ~N (0, 1) : Standard normal variate
P : Population proportion
p : Sample proportion
ν : Degrees of freedom(df)
2  : Chi-square with ν df

: Gamma function
B(a, b) : Beta function
log x : Logarithm base e
α : Area under the curve
t   ,  : Tabulated value of t-statistic with ν df such that the area
under the curve of t-distribution to its right (upper) tail is
equal to α
2 ,  : Tabulated value of χ2-statistic with ν df such that the area
under the curve of χ2-distribution to its right (upper) tail is
equal to α
F 1 , 2 ,  : Tabulated value of F-statistic with (ν1, ν2) df such that the
area under the curve of F-distribution to its right (upper) tail
is equal to α
UNIT 1 INTRODUCTION TO SAMPLING
DISTRIBUTION
Structure
1.1 Introduction
Objectives
1.2 Basic Terminology
1.3 Introduction to Sampling Distribution
1.4 Standard Error
1.5 Central Limit Theorem
1.6 Law of Large Numbers
1.7 Summary
1.8 Solutions /Answers

1.1 INTRODUCTION
In many situations, if we extract some information from all the elements or
items of a large population in general, it may be time consuming and
expensive. Also, if the elements or items of population are destroyed under
investigation then the information gathering from all the elements is not
making a sense because all the elements will be destroyed. Therefore, in most
of the cases in daily life, business and industry the information is gathered by
means of sampling. The results of a properly taken sample enable the
investigator to arrive at generalisations that are valid for entire population. The
process of generalising sample results to the population is called Statistical
Inference.
For drawing inference about the population parameters, we draw all possible
samples of same size and determine a function of sample values, which is
called statistic, for each sample. The values of statistic are generally varied
from one sample to another sample. Therefore, the sample statistic is a random
variable and follows a distribution. The distribution of the statistic is called
sampling distribution of the statistic.
This unit is divided into 8 sections. Section 1.1 is introductive in nature. In
Section 1.2, we defined the basic terminology used in statistical inference. The
introduction and needs of sampling distributions are described in the Section
1.3. In Section 1.4, we explored the concept of standard error. The most
important theorem of Statistics “Central Limit Theorem” is described in
Section 1.5. In Section 1.6, we explored the concept of law of large numbers.
Unit ends by providing summary of what we have discussed in this unit in
Section 1.7 and solution of exercises in Section 1.8.
Objectives
After studying this unit, you should be able to:
 define statistical inference;
 define the basic terms as population, sample, parameter, statistic, estimator,
estimate, etc. used in statistical inference;
 explain the concept of sampling distribution;
 explore the importance and needs of sampling distribution;
7
Sampling Distributions  explain the concept of standard error;
 describe the most important theorem of Statistics “Central Limit Theorem”;
 explain the concept of law of large numbers; and
 determine adequate sample size required for sample survey.

1.2 BASIC TERMINOLOGY

Before discussing the sampling distribution of a statistic, we shall be discussing
basic definitions of some of the important terms which are very helpful to
understand the fundamentals of statistical inference.
Population
In general sense “population” means a large group of people who live in a
particular geographical area. For example, the group of people who live in New
Delhi, the group of people working in IGNOU, students enrolled in PGDAST
programme in IGNOU, etc.
In Statistics, population has a much broader meaning. It does not only refer to
The group of individuals / people but also the group of elements or units under consideration by the
items / units/ observations analyst. Thus, population is the collection or group of individuals /items /units/
under study is known as
observations under study. For example, the collection of books in a library, the
population.
particles in a salt bag, the rivers in India, the students in a classroom, etc. are
considered as populations in Statistics.
The total number of elements / items / units / observations in a population is
known as population size and denoted by N. The characteristic under study
may be denoted by X or Y.
We may classify the population into two types as:
(1) Finite and Infinite Population
Population If a population contains finite number of units or observations, it is called a
finite population. For example, the population of students in a class, the
Based Based
on on population of bolts produced in a factory in a day, the population of electric
number subject bulbs in a lot, the population of books in a library, etc. are the finite
populations because in these examples the number of units in the population is
Finite Infinite Real Hypothetical finite in numbers therefore these are all the examples of finite population.
If a population contains an infinite (uncountable) number of units or
observations, it is called an infinite population. For example, the population of
particles in a salt bag, the population of stars in the sky, the population of hairs
on human body, etc. are infinite populations because in these examples the
number of units in the population is not finite therefore these are all the
examples of infinite population.
But theoretically sometimes, populations of too large in size are assumed
infinite.
(2) Real and Hypothetical Population
The population can also be classified as real and hypothetical. A population
comprising the items or units which are all physically present is known as real
population. All of the examples given above are examples of a real population.
If a population consists the items or units which are not physically present but
the existence of them can only be imagined or conceptualised then it is known
8
as hypothetical population. For example, the population of heads or tails in Introduction to Sampling
successive tosses of a coin a large number of times is considered as Distribution
hypothetical population.
Sample
To extract the information from all the elements or units or items of a large
population may be, in general, time consuming, expensive and difficult. And
also if the elements or units of population are destroyed under investigation
then gathering the information from all the units is not make a sense. For
example, to test the strength of the chalk, to test the quality of crackers, etc. the
chalks and crackers are destroyed under testing procedure. In such situations, a
small part of population is selected from the population which is called a
sample. Thus, the sample can be defined as below:
“A sample is a part / fraction / subset of the population.”
The procedure of drawing a sample from the population is called sampling.
You will learn about sampling in detail in Block 1 of course MST-005. The
number of units selected in the sample is known as sample size and it is
denoted by n.
Complete Survey and Sample Survey
Statistical data may be collected by two methods:
(1) Complete Survey or Complete Enumeration or Census
(2) Sample Survey or Sample Enumeration
(1) Complete Survey or Complete Enumeration or Census
When each and every element or unit of the population is investigated or
studied for the characteristics under study then we call it complete survey or
census. For example, suppose we want to find out the average height of the
students of a study centre then if we measure the height of each and every
student of this study centre to find the average height of the students then such
type of survey is called complete survey.
(2) Sample Survey or Sample Enumeration
When only a part or a small number of elements or units (i.e. sample) of
population are investigated or studied for the characteristics under study then
we call it sample survey or sample enumeration. In the above example, if we
select some students of this study centre and measure the height to find average
height of the students then such type of survey is called sample survey.
Simple Random Sampling or Random Sampling
The simplest and most common method of sampling is simple random
sampling. In simple random sampling, the sample is drawn in such a way that
each element or unit of the population has an equal and independent chance of
being included in the sample. If a sample is drawn by this method then it is
known as a simple random sample or random sample. The random sample
of size n is denoted by X1 ,X2 , ...,Xn or Y1 , Y2 , ..., Yn and the observed value of
this sample is denoted by x1 , x 2 , ..., x n or y1 , y2 , ..., yn . Some author use
x1 , x 2 , ...,x n to represent the random sample instead of X1 ,X2 , ...,Xn . But
throughout the course, we use capital letters to represent the random sample,
that is, X1 , X2 , ...,Xn and X1 , X2 , ...,X n simply be assumed that these are the
independent identically distributed(iid) random variables.
9
Sampling Distributions In simple random sampling elements or units are selected or drawn one by one
therefore it may be classified into two types as:
(1) Simple Random Sampling without Replacement (SRSWOR)
In simple random sampling, if the elements or units are selected or drawn one
by one in such a way that an element or unit drawn at a time is not replaced
back to the population before the subsequent draws is called SRSWOR. If we
draw a sample of size n from a population of size N without replacement then
total number of possible samples is N C n . For example, consider a population
that consists of three elements, A, B and C. Suppose we wish to draw a random
sample of two elements then N = 3 and n = 2. The total number of possible
random samples without replacement is N Cn  3 C2  3 as (A, B), (A, C) and
(B, C).
(2) Simple Random Sampling with Replacement (SRSWR)
In simple random sampling, if the elements or units are selected or drawn one
In SRSWR each of the by one in such a way that a unit drawn at a time is replaced back to the
1st, 2nd, .., nth draw the
elements are remain population before the subsequent draw is called SRSWR. In this method, the
same N due to same element or unit can appear more than once in the sample and the
replacement so by rule probability of selection of a unit at each draw remains same i.e. 1/N. In this
of multiplication total method, total number of possible samples is Nn. In above example, the total
number of possible
samples is
number of possible random samples with replacement is N n  32  9 as (A, A),
N  N  ...n times  N n .
(A, B), (A, C), (B, A), (B, B), (B, C), (C, A), (C, B) and (C, C).
Note1: The statistical inference is based on the theoretical distribution so in
whole statistical inference, we assume that the random sample is selected from
the infinite population or from a finite population with replacement so that
removal of the sample unit has no appreciable effect on the composition of the
population. That is, we draw the random sample such that:
(a) The n successive sample observations are independent and
(b) The population composition remains constant from draw to draw.
For example, if we draw a random sample X1 , X2 , ..., Xn from normal
population with mean µ and variance σ2 then Xi’s are independent and follow
same normal distribution.
Parameter
A group of elements or units under study can be regarded as a population and
A parameter is a the characteristics of population can be described with some measures such as
function of total numbers of items, average, variance, etc. These measures are known as
population values
which is used to
parameters of the population. Thus, we can define parameter as:
represent the certain A parameter is a function of population values which is used to represent the
characteristic of the
certain characteristic of the population. For example, population mean,
population.
population variance, population coefficient of variation, population correlation
coefficient, etc. are all parameters. Population parameter mean usually denoted
by µ and population variance denoted by σ2.
We know that the population can be described with the help of distribution and
the distribution is fully determined with the help of its constants such as, in
case of normal distribution, we need to know µ and σ2 to determine the normal
distribution, in case of Poisson distribution, we need to know λ, etc. These
constants are known as parameter.

10
 Sample Mean and Sample Variance Introduction to Sampling
Distribution
If X1 ,X2 , ..., Xn is a random sample of size n taken from a population whose
probability density(mass) function f(x, θ) then sample mean is defined as
n
X 1  X 2  ...  X n 1
X
n

n
X
i 1
i

And sample variance is defined as

1 n 2
S2  
n  1 i 1
 Xi  X

n
2
Here, we divide X
i 1
i  X  by (n – 1) rather than n as our definition of the

variance described in Unit 2 of MST-002. The reason for taking (n – 1) in place

of n will become clear in the Section 5.4 of Unit 5 of this course.
Statistic
Any quantity calculated from sample values and does not contain any unknown
parameter is known as statistic. For example, if X1 ,X2 , ..., Xn is a random
sample of size n taken from a population with mean µ and variance σ2 (both are
ple
not 1 n
unknown) then sample mean X   X i is a statistic whereas
n i 1
tion
X   and X /  are not statistics because both are function of unknown
parameters.
Estimator and Estimate
Generally, population parameters are unknown and the whole population is too
large to find out the parameters. Since the sample drawn from a population
always contains some or more information about the population, therefore in
Any statistic used to
such situations, we guess or estimate the value of the parameter under study estimate an unknown
based on a random sample drawn from that population. population parameter is
So if a statistic is used to estimate an unknown population parameter then it is known as estimator
and the particular value
known as estimator and the value of the estimator based on observed value of of the estimator is
the sample is known as estimate of parameter. For example, suppose the known as estimate of
parameter λ of the Poisson population f(x, λ) is unknown so we draw a random parameter. The
sample X1 ,X2 , ..., Xn of size n from this Poisson population to estimate λ. If estimated value of
sample mean and
1 n sample variance are
we use sample mean X   X i to estimate λ then X is called estimator and
n i 1 denoted by x and s2
respectively.
any particular value of this estimator, say, x is called estimate of unknown
parameter λ. Consider another example, if we want to estimate the average
height () of students in a college with the help of sample mean X then X is
the estimator and its particular value, say, 165 cm is the estimate of the
population average height () .
In this course, we use capital letters for estimator and small letters for
estimated value. In general, the estimator is denoted by Tn = t(X1, X2, …, Xn)
where ‘n’ denotes the sample size and the estimator T is a function of the
random sample X1, X2, …, Xn.
Now, it is right place to do some examples.
11
Sampling Distributions Example 1: Categories the following populations as finite and infinite:
(i) Population of newly born babies during a year in a particular hospital.
(ii) Population of real numbers between 1 and 2.
(iii) Population of number of words on this page.
Solution: (i) and (iii) are the finite populations because the number of elements
(babies or words) in both are finite whereas (ii) is infinite because there are
infinite real numbers are possible between 1 and 2.
Example 2: If population size is 6 then how many samples of size of 4 are
possible with replacement?
Solution: Here, we are given that
Population size = N = 6,
Sample size = n = 4
Since we know that all possible samples of size n taken from a population of
size N with replacement are Nn so in our case Nn = 64 = 1296.
Now, you check your understanding of above discussion by answering the
following exercises.
E1) In which situation we use only sample survey:
(i) Blood testing.
(ii) To know the average income of the people living in a colony.
(iii) To know the number of students enrolled in IGNOU.
(iv) To know the average life of electric bulbs.
E2) If X1 , X2 , ..., Xn is a random sample of size n taken from normal
population with mean µ and variance σ2 both are unknown then find the
statistic in the following:
1 n n
(i)  Xi  
n i 1
(ii) X
i 1
i

n n
2
(iii)  Xi / 
i 1
(iv) X
i 1
i

E3) The weights (in kg.) of 5 workers in a factory are 56, 62, 74, 45 and 50.
How many samples of size of 2 are possible with replacement? Also
write all possible samples of size 2.

1.3 INTRODUCTION TO SAMPLING

DISTRIBUTION
As we have discussed in Section 1.1 that if the population is too large or the
units or items of the population are destructive in nature or there is a limited
resources such as men power, money, etc., then it is not possible practically to
examine each and every unit of the population to obtain the necessary
information about the population. For example, Suppose we want to know the
average life of electric bulbs of a certain brand which are manufactured by a
company. The company manufactures a lot of bulbs say 5,000 per day. Then
gathering information about average life of all the bulbs is not making a sense
12
because the bulbs are destroyed under investigation. In another example, if we Introduction to Sampling
want to know the average income of the persons living in a big city then Distribution
collecting the information from each and every person is very much time and
manpower consuming. In such situations, one can draw sample from the
population under study and utilize sample observations to extract the necessary
information about the population. The results obtained from the sample are
projected in such a way that they are valid for the entire population. Therefore,
the sample works like a “Vehicle” to reach (drawing) at valid conclusions
about the population. Thus, statistical inference can be defined as:
“The process of projecting the sample results for the whole population is
known as statistical inference.”
For drawing the inference about the population, we analyse the sample data,
that is, we calculate the sample statistic as sample mean, sample proportion,
sample variance, etc. Generally, if we want to draw the inference about the
population mean we use sample mean, about population variance we use
sample variance, etc.
For example, suppose we want to study the average life of electric bulbs
produced by the company discussed in above example and we decide to
estimate the average life of electric bulbs on the basis of a sample. So we take a
sample, say, 10 bulbs and measure the life of each bulb selected in the sample.
The observed values of life of bulbs are given in column 2 of the Table 1.1
given below:
Table 1.1: Life Times of 10 Bulbs in Sample-I and Sample-II
S.No. Life of Bulbs (in hours)
Sample-I Sample-II
1 2000 400
2 500 0
3 1000 1400
4 1500 1200
5 1200 800
6 200 700
7 700 1200
8 1400 800
9 800 600
10 1100 1500
Total 10400 8600
X 1040 860

From the above table, the average (mean) life of these selected bulbs is 1040
hours. Now, if we are using this sample to make inference about the population
average life then we say that, the sample mean is 1040 hours give an estimate
of the average life of electric bulbs for the whole population. But if we take
another sample of same size 10 from the population and calculate the sample
mean as shown in the above table. The average life of bulbs selected in sample-
II is 860 hours. This is different from the mean of sample-I. Thus, if we are
using this sample mean to estimate the population mean then we gets a
different estimate of the average life of bulbs for the whole population.
Similarly, we can take many samples of same size 10 and for each sample we
get sample mean which may or may not be distinct. From all of the sample
means, we try to estimate the average life of bulbs of the whole population.
13
Sampling Distributions For better understanding of the process of generalising these sample results to
the whole population, we consider the example in which size of the population
is very small.
Consider a population comprising four typists who type the sample page of a
manuscript. The number of errors made by each typist is shown in Table 1.2
given below:
Table 1.2: Number of Error per Typist
Typist Number of Errors
A 4
B 2
C 3
D 1

The population mean (average number of errors) can be obtained as

4  2  3 1
µ  2.5
4
Now, let us assume that we do not know the average number of errors made by
typists. So we decide to estimate the population mean on the basis of sample of
size n = 2. There are Nn = 42 = 16 possible simple random samples with
replacement of size 2. All possible samples of size n = 2 are given below and
for each sample the sample mean is calculated as shown in Table 1.3 given
below:
Table 1.3: Calculation of Sample Mean
Sample Sample in Term of Sample Sample Mean
Number Typist Observation (X)
1 (A, A) (4, 4) 4.0
2 (A, B) (4, 2) 3.0
3 (A, C) (4, 3) 3.5
4 (A, D) (4, 1) 2.5
5 (B, A) (2, 4) 3.0
6 (B, B) (2, 2) 2.0
7 (B, C) (2, 3) 2.5
8 (B, D) (2, 1) 1.5
9 (C, A) (3, 4) 3.5
10 (C, B) (3, 2) 2.5
11 (C, C) (3, 3) 3.0
12 (C, D) (3, 1) 2.0
13 (D, A) (1, 4) 2.5
14 (D, B) (1, 2) 1.5
15 (D, C) (1, 3) 2.0
16 (D, D) (1, 1) 1.0

From the above table, we can see that the value of the sample statistic (sample
mean) is varying from sample to sample and out of 16 samples only 4
samples(4,7,10 and 13) have their means equal to population mean whereas
other 12 samples result in some error in the estimation process. The error is the
difference between the population mean and the sample mean used for
estimate.
14
Now, consider all possible values of sample mean and calculate their Introduction to Sampling
probabilities (using relative frequency approach of probability described in Distribution
Unit 2 of MST-003) of occurrence. Then we arrange every possible value of
sample mean with their respective probabilities in the following Table 1.4
given below:

Table 1.4: Sampling Distribution of Sample Means

S. No. X Frequency(f) Probability(p)

1 1.0 1 1/16 = 0.0625

2 1.5 2 2/16 = 0.1250
3 2.0 3 3/16 = 0.1875
4 2.5 4 4/16 = 0.2500
5 3.0 3 3/16 = 0.1875
6 3.5 2 2/16 = 0.1250
7 4.0 1 1/16 = 0.0625

So the arrangement of all possible values of sample mean with their

corresponding probabilities is called the sampling distribution of mean. Thus,
we can define the sampling distribution of a statistic as:
“The probability distribution of all possible values of a sample statistic that
would be obtained by drawing all possible samples of the same size from the
population is called sampling distribution of that statistic.”
The sampling distribution of sample means itself has mean, variance, etc.
Therefore, the mean of sample means can be obtained by the formula
1 k k
Mean of sample means  X   ii
K i 1
X f where, K  
i 1
fi

1
 1.0  1  1.5  2  ...  4.0  1  2.5  
16
The mean of sample means can also be calculated as
k
1 2 1
E(X)  X   X i pi  1.0   1.5   ...  4.0   2.5
i 1 16 16 16
Thus, we have seen for this population that mean of sample means is equal to
the population mean, that is, X    2.5. The fact that these two means are
equal is not a chance as it can be shown that X equals to population mean for
any population and any given sample size.
But if the population is large say having 1,000 elements and we want to draw
all samples of size 2 then there are (1000)2  1000000 possible simple random
samples with replacement. Then the listing of all such samples and finding the
sampling distribution would be a difficult task empirically. Therefore, we may
consider a theoretical experiment in which we draw all possible sample of a
fixed size and obtain a sampling distribution.
Although in practice only a random sample is actually selected and the concept
of sampling distribution is used to draw the inference about the population
parameters.
15
Sampling Distributions It is now time for you to try the following exercise to make sure that you
understand about sampling distribution.
E4) If lives of 3 Televisions of certain company are 8, 6 and 10 years then
construct the sampling distribution of average life of Televisions by
taking all samples of size 2.

1.4 STANDARD ERROR

As we have seen in the previous section that the values of sample statistic may
vary from sample to sample and all the sample values are not equal to the
population parameter. Now, one can be interested to measure how much the
values of sample statistic vary from the population parameter on average. As
we have already studied in Unit 2 of MST-002 that the standard deviation is
used as a measure of variation.
Thus, for measuring the variation in the values of sample statistic around the
population parameter we calculate the standard deviation of the sampling
distribution. This is known as standard error of that statistic. Thus, standard
Standard deviation of
error of a statistic can be defined as:
sampling distribution “The standard deviation of a sampling distribution of a statistic is known as
of a statistic is called standard error and it is denoted by SE.”
standard error.
Therefore, the standard error of sample mean is given by
k
1 k 2
SE(X)   f i  x i    where, K   f i
K i 1 i 1

In the previous example, we can calculate the standard error of sample mean as
1  2 2 2
SE  X   1  1.0  2.5   2  1.5  2.5   ...  1   4.0  2.5  
16  

1 10
  2.25  2  ...  2.25    0.791
16 16
The computation of the standard error is a tedious process. There is an
alternative method to compute standard error of the mean from a single sample
as:
If X1 ,X2 , ..., Xn is a random sample of size n taken from a population with
mean µ and variance σ2 then the standard errors of sample mean ( X ) is given
by

SE  X  
n
The standard error is used to express the accuracy or precision of the estimate
of population parameter because the reciprocal of the standard error is the
measure of reliability or precision of the statistic. Standard error also
determines the probable limits or confidence limits (described in Units 7 & 8 of
this course) within which the population parameter may be expected to lie with
certain level of confidence. Standard error is also applicable in testing of
hypothesis (described in Block 3 of this course).
16
The standard errors of some other well known statistics are given below: Introduction to Sampling
Distribution
1. The standard error of sample proportion (p) is given by
PQ
SE(p) =
n
where, P is population proportion and Q = 1-P.
 ) is given by
2. The standard error of sample median ( X

)= π σ
SE( X = 1.25331 SE ( X )
2 n
3. The standard error of difference of two sample means is given by

σ12 σ 22
SE  X  Y = 
n1 n 2

where, σ12 and σ 22 are the population variances and n1 and n2 are the sample
sizes of two independent samples.
4. The standard error of difference between two sample proportions is given
by

P1Q1 PQ
SE( p1  p 2 ) =  2 2
n1 n2

where, P1 and P2 are population proportions of two different populations

and Q1= 1-P1 & Q2 = 1-P2.
From all the above formulae, we can understand that standard error is inversely
proportional to the sample size. Therefore, as sample size increases the
standard error decreases.
Note 2: These standard errors discussed above were calculated under the
assumption that sampling is done either from an infinite population or from a
finite population with replacement. But, in real life sampling problems, most
sampling plans do not permit an element to be selected twice in a given sample
(i.e. sampling with replacement). Consequently, if the population is not large in
relation to the size of the sample and sampling is done without replacement
Nn
then we multiply the above standard errors by the correction factor .
N 1
where, N is the population size. This correction factor is known as finite
population correction factor.
Therefore, in this case, the standard error of the sample mean is given by
 Nn
SE  X  =
n N 1
Similarly,
N  n PQ
SE(p) =
N 1 n
In practice, if the sample size n is less than or equal to 10% of the population
size N, this correction factor may be ignored for all practical purposes.
17
Sampling Distributions Let us do one example relating to standard error.
Example 3: Diameter of a steel ball bearing produced by a semi-automatic
machine is known to be distributed normally with mean 12 cm and standard
deviation 0.1 cm. If we take a random sample of size 10 with replacement then
find standard error of sample mean for estimating the population mean of
diameter of steel ball bearing for whole population.
Solution: Here, we are given that
 = 12, σ = 0.1, n = 10
Since the sampling is done with replacement therefore the standard error of
sample mean for estimating population mean is given by
 0.1
SE  x     0.03
n 10
In the same way, you can try the following exercise.
E5) The average weight of certain type of tyres is 200 pounds and standard
deviation is 4 pounds. A sample of 50 tyres is selected. Obtain the
standard error of sample mean.

E6) A machine produces a large number of items of which 15% are found to
be defective. If a random sample of 200 items is taken from the
population, then find the standard error of sampling distribution of
proportion.

1.5 CENTRAL LIMIT THEOREM

The central limit theorem is the most important theorem of Statistics. It was
first introduced by De Movers in the early eighteenth century.
According to the
central limit theorem, According to the central limit theorem, if X1 ,X2 , ..., Xn is a random sample of
the sampling size n taken from a population with mean µ and variance σ2 then the sampling
distribution of the
sample means tends to
distribution of the sample mean tends to normal distribution with mean µ and
normal distribution as variance σ2/n as sample size tends to large (n  30) whatever the form of
sample size tends to parent population, that is,
large ( n  30 ).
 2 
X ~ N  , 
 n 

and the variate

X 
Z ~ N  0,1
/ n
follows the normal distribution with mean 0 and variance unity, that is, the
variate Z follows standard normal distribution.
We do not intend to prove this theorem here, but merely show graphical
evidence of its validity in Fig. 1.1. Here, we will also try to show that how
large must the sample size be for which we can assume that the central limit
theorem applies?

18
 Introduction to Sampling
Distribution

Fig. 1.1: Sampling distribution of sample means for various populations when n = 2,
n = 5 and n = 30
Sampling Distributions In this figure, we are trying to understand the sampling distribution of sample
mean X for different populations and for varying sample sizes. We divide this
figure into four parts A, B, C and D. The part ‘A’ of this figure shows four
different populations as normal, uniform, binomial and exponential. The rest
parts B, C and D represent the shape of sampling distribution of mean of sizes
n = 2, n = 5 and n = 30 respectively drawn from the populations shown in first
row (Part-A). From the first column of this figure, we observed that when the
parent population is normal then all the sampling distributions for varying
sample sizes are also normal, having same mean but their variances decrease as
n increases.
The second column of this figure represents the uniform population. Here, we
observe that the sampling distribution of X is symmetrical when n = 2 and
tends to normal when n = 30.
However, the third column of this figure represents the binomial population
(discrete). Again when n = 2, the sampling distribution of X is symmetrical
and for n = 5 it is quite bell shaped and tends to normal when n = 30. The last
column of this figure represents the exponential population which is highly
skewed. Here, we observe that for n =30, the distribution of X is symmetrical
bell shaped normal.
Here, we conclude that if we draw a random sample of large size n > 30 from
the population then the sampling distribution of X can be approximated by a
normal probability distribution, whatever the form of parent population.
Note 3: The central limit theorem can also be applicable in the same way for
the sampling distribution of sample proportion, sample standard deviation,
difference of two sample means, difference of two sample proportions, etc. that
is, if we take a random sample of large size  n  36  30 from the population
then the sampling distribution of sample proportion, sample standard deviation,
difference of two sample means, difference of two sample proportions, etc.
approximated by a normal probability distribution, whatever the form of parent
population. The mean and variance of these sampling distributions shall be
discussed in the next unit.
Let us do one example based on central limit theorem.
Example 4: Suppose in IGNOU, the average (mean) weight of all male
students is 60 kg and standard deviation is 25 kg. If a sample of 36 male
students is selected at random, find the probability that the male students
having average weight
(i) more than 70 kg
(ii) less than 55 kg
(iii) between 50 kg and 65 kg
Solution: Here, we are given that
 = 60 kg, σ = 25 kg, n = 36
Since, we draw a large sample (n > 30), therefore by central limit theorem the
sampling distribution of the sample means will also follow a normal
distribution with mean
E  X    60
and variance

20
 2 (25)2 Introduction to Sampling
Var  X     17.36 Distribution
n 36
(i) The probability that the male students having average weight more than
70 kg is given by
P X  70 SeeFig. 1.2
To get the value of this probability, we convert the variate X into a
standard normal variate Z by the transformation
X  E(X) X  60 X  60
Z  
Var(X) 17.36 4.17

Therefore, subtract 60 from each term and then divide each term by 4.17
in the above inequality. Thus, the probability becomes
 X  60 70  60 
P   P  Z  2.40
 4.17 4.17 
 0.5  P  0  Z  2.40

 0.5  0.4918  Using table area 

 under normal curve 
 0.0082

X  60 X  70
Z0 Z  2.40
Fig. 1.2
(ii) The probability that the male students having average weight less than 55
kg is given by
 X  60 55  60 
P X  55  P  
 4.17 4.17 

 P  Z  1.20 See Fig. 1.3

 PZ  1.20  normal curve is symmetrical 
 about the line Z  0. 
 0.5  P  0  Z  1.20

 0.5  0.3849  Using table area under 

 normal curve 
 0.1151
Sampling Distributions

X  55 X  60
Z  1.20 Z0
Fig. 1.3
(iii) The probability that the male students having average weight between 50
kg and 65 kg is given by

 50  60 X  60 65  60 
P 50  X  65  P   
 4.17 4.17 4.17 

 P  2.40  Z  1.20 See Fig.1.4

 P 2.40  Z  0  P0  Z  1.20

 P0  Z  2.40  P0  Z  1.20

 0 .4918  0 .3849  Using table area 

 under normal curve 
 0.8767

X  50 X  60 X  65
Z  2.40 Z0 Z  1.20
Fig. 1.4
Now, you can try the following exercise which will make you more user
friendly with the use of central limit theorem.
E7) Average height of the students of science group in a college is 65 inches
with a standard deviation of 2.2 inches. If a sample of 40 students is
selected at random, what is the probability that the average height of
these 40 students lies between 64 and 65.5 inches?
 Introduction to Sampling
1.6 LAW OF LARGE NUMBERS Distribution

We have already discussed in Section 1.2 of this unit that the population
parameters are generally unknown and for estimating parameters, we draw all
possible random samples of same size from the population and calculate the
values of sample statistic such as sample mean, sample proportion, sample
variance, etc. for all samples and with the help of these values we form
sampling distribution of that statistic. Then we draw inference about the
population parameters.
But in real-world the sampling distributions are never really observed. The
process of finding sampling distribution would be very tedious because it
would involve very large number of samples. So in real-world problems, we
draw a random sample from the population to draw inference about the
population parameters. A very crucial question then arises: “Using a random
sample of finite size say n, can we make a reliable inference about population
parameter?” The answer is “yes”, reliable inference about population parameter
can be made by using only a finite sample and we shall demonstrate this by
“law of large numbers”. This law of large numbers can be stated in words as:
“A positive integer can be determined such that if a random sample of size n or
lager is taken from a population with parameter, say, population mean (), the
probability that the sample mean X will deviate from population mean µ can
be made to be as close to 1 as desired.”
This law states that for any two arbitrary small numbers ε and η where ε  0
and 0  η  1 , there exists an integer n such that if a random sample of size n or
larger is drawn from the population and calculate the sample mean X for this
sample, then the probability that X deviates from  by less than  is greater
than 1 η (as close to 1 as desired) , that is, X arbitrarily close to .
Symbolically this can be written as
2
For any ε  0 and 0  η  1 there exists an integer n such that n  , where
2 
σ2 is the finite variance of population then

P  X      1  

or P    X       1    X  a   a  X  a 

The proof of “law of large numbers” is beyond the scope of this course.
Here, with an example we have to try to understand the law of large numbers in
action, as we take more observations the sample mean approaches to the
population mean.
Suppose that the distribution of the weight of all the young men living in a city
is close to normal distribution with mean 65 kg and standard deviation 25 kg.
To understand the law of large numbers, we calculate the sample mean weight
( X ) for varying sample size n = 1, 2, 3, …. Fig. 1.5 shows the behaviour of the
sample mean weight X of men chosen at random from this city. The graph
plots the values of X (along vertical axis and sample size along horizontal axis)
as sample size varying from n =1, 2, 3…

23
Sampling Distributions First we start a sample of size n = 1, that is, we select a man randomly from the
city. Suppose selected man had weight 70 kg, therefore, the line of the graph
starts from this point. Now, select second man randomly and suppose his
weight is 55 kg. So for n = 2 the sample mean is
70  55
X  62.5 kg
2
This is the second point on the graph. Now, select third man randomly from the
city and suppose his weight is 83 kg. Therefore, for n = 3 the sample mean is
70  55  83
X  69.3 kg
3
This is the third point on the graph.

Fig. 1.5
This process will continue. From the graph we can see that the mean of the
sample changes as we make more observations, eventually, however, the mean
of the observations get close and close to the population mean 65 kg as the
sample size increases.
Note 4: The law of large numbers can also be applied for the sample
proportion, sample standard deviation, difference of two sample means,
difference of two sample proportions, etc.
Example 5: A pathologist wants to estimate the mean time required to
complete a certain analysis on the basis of sample study so that he may be 99%
confident that the mean time may remain with  2 days of the mean. As per the
available records, the population variance is 5 days2. What must be the size of
the sample for this study?
Solution: Here, we are given
1    0.99    0.01 ,   2 , σ 2  5
By the law of large numbers, we have
σ2
n
ε 2η
 5 Introduction to Sampling
 2
 125 Distribution
2  0.01
That is, n  125
Hence, at least 125 units must be drawn in a sample.
Example 6: An investigator wishes to estimate the mean of a population using
a sample large enough that the probability will be 0.95 that the sample mean
will not differ from the population mean by more than 25 percent of the
standard deviation. How large a sample should be taken?
Solution: We have
1    0.95    0.05,   0.25
By the law of large numbers, we have
2
n
 2
2
 2  320
 0.25  0.05
That is, n  320
Therefore, at least 320 units must be taken in the sample so that the required
probability will be 0.95.
Now, try the following exercise for your practice.

E8) The mean of a population is unknown and having a variance equal to 2.

Find out that how large a sample must be taken, so that the probability
will be at least 0.95 that the sample mean will lie within the range of 0.5
of the population mean?
We now end this unit by giving a summary of what we have covered in it.

1.7 SUMMARY
In this unit, we have covered the following points:
1. The statistical procedure which is used for drawing conclusions about the
population parameter on the basis of the sample data is called “statistical
inference”.
2. The group of units or items under study is known as “Population”
whereas a part or a fraction of population is known as “sample”.
3. A “parameter” is a function of population values which is used to
represent the certain characteristic of the population and any quantity
calculated from sample values and does not contain any unknown
population parameter is known as “statistic”.
4. Any statistic used to estimate an unknown population parameter is known
as “estimator” and the particular value of the estimator is known as
“estimate” of parameter.
5. The probability distribution of all possible values of a sample statistic that
would be obtained by drawing all possible samples of the same size from
the population is called “sampling distribution” of that statistic.
25
Sampling Distributions 6. The standard deviation of the sampling distribution of a statistic is known
as “standard error”.
7. The most important theorem of Statistics is “central limit theorem”
which state that the sampling distribution of the sample means tends to
normal distribution as sample size tends to large (n > 30).
8. According to law of large numbers, a positive integer can be determined
such that if a random sample of size n or larger is taken from a population
with parameter, say, population mean (µ), the probability that the sample
mean will be very closed to population mean, can be made as close to 1 as
desired.

1.8 SOLUTIONS / ANSWERS

E1) In (i) and (iv) we use only sample survey because the units (persons or
blubs) are destroyed under investigation and the information gathering
from all the elements is not making a sense whereas in (ii) and (iii) we
can use sample as well as complete survey.
E2) (ii) and (iv) are the statistics because both do not contain the unknown
parameters (i.e. µ and σ2) whereas (i) and (iii) are not statistics because
they contain unknown parameters.
E3) Here, we are given that
N = 5, n = 2
All possible samples (with replacement) are Nn = 52 = 25 which are
shown in the Table 1.5 given below:
Table 1.5: Possible samples
Sample Sample Sample Number Sample
Number Observation Observation
1 56, 56 14 74, 45
2 56, 62 15 74, 50
3 56, 74 16 45, 56
4 56, 45 17 45, 62
5 56, 50 18 45, 74
6 62, 56 19 45, 45
7 62, 62 20 45, 50
8 62, 74 21 50, 56
9 62, 45 22 50, 62
10 62, 50 23 50, 74
11 74, 56 24 50, 45
12 74, 62 25 50, 50
13 74, 74

E4) Here, we are given that

N = 3, n = 2
Since we have to estimate the average life of Televisions on the basis of
samples of size n = 2 therefore, all possible samples (with replacement)
are Nn = 32 = 9 and for each sample we calculate the sample mean as
shown in Table 1.6 given below:

26
 Table 1.6: Calculation of Sample Mean Introduction to Sampling
Distribution
Sample Sample Sample
Number Observation Mean ( X )
1 8, 8 8
2 8, 6 7
3 8,10 9
4 6, 8 7
5 6, 6 6
6 6, 10 8
7 10, 8 9
8 10, 6 8
9 10, 10 10

Since the arrangement of all possible values of sample mean with their
corresponding probabilities is called the sampling distribution of mean
thus, we arrange every possible value of sample mean with their
respective probabilities in the following Table 1.7 given below:
Table 1.7: Sampling distribution of sample means
S.No. Sample Mean Frequenc Probability
(X ) y
1 6 1 1/9
2 7 2 2/9
3 8 3 3/9
4 9 2 2/9
5 10 1 1/9

E5) Here, we are given that

 = 200, σ = 4, n = 50
Since the sampling is done with replacement therefore we know that the
standard error of sample mean for estimating population mean is given
by
 4
SE  X     0.57
n 50
E6) Here, we are given that
15
P= = 0.15, n = 200
100
Therefore, we know that the standard error of sample proportion is
given by

PQ 0.15  1  0.15 
SE  p     0.025
n 200
E7) Here, we are given that
 = 65, σ = 2.2, n = 40
Since we draw a large sample n = 40 > 30, therefore, by central limit
theorem, the sampling distribution of the sample mean will follow a
normal distribution with mean
E  X    65

27
Sampling Distributions and variance
2
2  2.2 
Var  X     0.12
n 40
Therefore, the probability that the average height of these 40 students
will lie between 64 and 65.5 inches is given by
P 64  X  65.5 See Fig.1.6
To get the value of this probability, we convert the normal variate X
into a standard normal variate Z by the transformation
X  E X X  65 X  65
Z  
Var  X  0.12 0.35

Therefore, subtract 65 from each term and then divide each term by
0.35 in the above inequality. Thus, the probability becomes
 64  65 X  65 65.5  65 
P    P 2.86  Z  1.43
 0.35 0.35 0.35 
 P 2.86  Z  0  P0  Z  1.43
 P0  Z  2.86  P0  Z  1.43
= 0.4979 + 0.4236 = 0.9215

X  64 X  65 X  65.5
Z  2.86 Z0 Z  1.43
Fig. 1.6
E8) Here, we are given that
2  2,   0.5, 1    0.95    0.05
By the law of large numbers, we have
2
n
2
2
 2
 160
0.5  0.05
That is, n  160
Therefore, at least 160 units must be taken in the sample so that the
probability will be at least 0.95 that the sample mean will lie within 0.5
of the population mean.
UNIT 2 SAMPLING DISTRIBUTION(S) OF
STATISTIC(S)
Structure
2.1 Introduction
Objectives
2.2 Sampling Distribution of Sample Mean
2.3 Sampling Distribution of Difference of Two Sample Means
2.4 Sampling Distribution of Sample Proportion
2.5 Sampling Distribution of Difference of Two Sample Proportions
2.6 Sampling Distribution of Sample Variance
2.7 Sampling Distribution of Ratio of Two Sample Variances
2.8 Summary
2.9 Solutions/Answers

2.1 INTRODUCTION
In previous unit, we have discussed the concept of sampling distribution of a
statistic. There are so many problems in business and economics where it
becomes necessary to draw inference about population parameters such as
population mean, population proportion, population variance, difference of
two population means, etc. For example, one may want to estimate the average
height of the students in a college, a businessman may want to estimate the
proportion of defective items in a production lot, a manufacturer of car tyres
may want to estimate the variance of the diameter of tyres, a pharmacist may
want to estimate the difference of effect of two types of drugs, etc.
Generally, sample mean is used to draw inference about the population mean.
Similarly, sample proportion and sample variance are used to draw inference
about the population proportion and population variance respectively.
Therefore, it becomes necessary to know the sampling distribution of sample
mean, sample proportion and sample variance, etc.
This unit is divided in 9 sections. Section 2.1 is introductive in nature. One of
the most important sample statistics which is used to draw conclusion about
population mean is sample mean. So in Section 2.2, the sampling distribution
of mean is described, whereas in Section 2.3, the sampling distribution of
difference of two sample means is explored. Second useful statistic generally
used in Social and Business problems is sample proportion. The sampling
distribution of sample proportion is described in Section 2.4 and in Section 2.5
sampling distribution of difference of two sample proportions is explored.
Sometimes, it is useful to estimate the variation of population. For this we use
sampling distribution of sample variance. The sampling distribution of sample
variance is described in Section 2.6 whereas the sampling distribution of ratio
of two sample variances is given in Section 2.7. Unit ends by providing
summary of what we have discussed in this unit in Section 2.8 and solution of
exercises in Section 2.9.
Objectives
After studying this unit, you should be able to:
 describe the need of sampling distributions of different statistics;
29
Sampling Distributions  explain the sampling distribution of sample mean;
 explore the sampling distribution of difference of two sample means;
 describe the sampling distribution of sample proportion;
 explore the sampling distribution of difference of two sample proportions;
 explain the sampling distribution of sample variance; and
 describe the sampling distribution of ratio of two sample variances.

2.2 SAMPLING DISTRIBUTION OF SAMPLE

MEAN
One of the most important sample statistics which is used to draw conclusion
about the population mean is sample mean. For example, an investigator may
want to estimate average income of the peoples living in a particular
geographical area, a product manager may want to estimate the average life of
electric bulbs manufactured by a company, a pathologist may want to estimate
the mean time required to complete a certain analysis, etc. In the above cases
an estimate of the population mean is required and one may estimate this on
the basis of a sample taken from that population. For this, sampling
distribution of sample mean is required.
We have already given you the flavor of sampling distribution of sample mean
with the help of an example in Section 1.3 of previous unit in which we draw
all possible samples of same size from the population and calculate the sample
mean for each sample. After calculating the value of sample mean for each
sample we observed that the values of sample mean vary from sample to
sample. Then the sample mean is treated as random variable and a probability
distribution is constructed for the values of sample mean. This probability
distribution is known as sampling distribution of sample mean. Therefore, the
sampling distribution of sample mean can be defined as:
The sampling distribution
of sample means is a “The probability distribution of all possible values of sample mean that would
theoretical probability be obtained by drawing all possible samples of the same size from the
distribution of sample population is called sampling distribution of sample mean or simply says
means that would be sampling distribution of mean.”
obtained by drawing all
possible samples of the Now, the question may arise in your mind “what is the shape of sampling
same size from the distribution of mean, it is normal, exponential, binomial, Poisson, etc.?”
population.
With the help of Fig. 1.1 of previous unit, we have described the shape of the
sampling distribution of mean for different populations and for varying sample
sizes. From this figure, we observed that when the parent population is normal
then all the sampling distributions for varying sample sizes are also normal
whereas when parent population is uniform, binomial, exponential then the
shapes of the sampling distributions of mean are not in the form of specify
distribution when sample size is small (n = 2 or 5).
Generally, when samples are drawn non-normal populations then it is not
possible to specify the shape of the sampling distribution of mean when the
sample size is small. Although when sample size is large (  30) then we
observed that sampling distribution of mean converges to normal distribution
whatever the form of the population i.e. normal or non-normal.
After knowing the shapes of the sampling distribution of mean in different
situations, you may be interested to know the mean and variance of the
sampling distribution of mean when samples are drawn from normal
population.
30
In practice, only one random sample is actually selected and the concept of Sampling Distribution(s)
sampling distribution is used to draw the inference about the population of Statistic(s)
parameters. If X1 , X 2 , ..., X n is a random sample of size n taken from a normal
population with mean µ and variance σ2 then it has also been established that The statistical inference is
sampling distribution of sample mean X is also normal. The mean and based on the theoretical
variance of sampling distribution of X can be obtained as distribution so if we say that
a random sample, say,
 X  X2  ...  X n  X1 , X 2 ,..., X n is taken from a
Mean of X  E  X   E  1   Bydefination of X 
 n population its means that the
random sample is selected
1 from the infinite population
 [E(X1 )  E(X 2 )  ...  E(X n )]
n or from a finite population
with replacement such that
Since X1, X2,…,Xn are randomly drawn from same population so they also X i ’s are independent and
follow the same distribution as the population. Therefore, followed same population
distribution.
E(X 1 )  E(X 2 )  ...  E(X n )  E(X)  
If X and Y are two
and independent random
Var(X1 )  Var(X 2 )  ...  Var(X n )  Var(X)   2 variables and a & b are two
constant then by the addition
Thus, theorem of expectation, we
have
1 
E  aX  bY   aE  X   bE  Y 
E(X)      ...   
n  
n  times

 and
Var  aX  bY 
1
  n     a 2 Var  X   b 2 Var  Y 
n
E X  

and variance
1 
Var(X)  Var  (X1  X2  ...  Xn )
n 
1
 2  Var(X1 )  Var(X 2 )  ...  Var(X n ) 
n
1  2  1
 2    2  ...   2   2  n2 

n  n  times  n
2
Var  X  
n
Hence, we conclude that if the samples are drawn from normal population
with mean µ and variance σ2 then the sampling distribution of mean X is also
normal distribution with mean µ and variance σ2/n, that is,
 2 
If X i ~ N , 2  then X ~ N   , 
 n 
The standard error (SE) of sample mean can be obtained by the definition of
SE as

SE  X   SD  X   Var  X  
n
Note 1: The statistical inference is based on the theoretical distribution so in
whole statistical inference, we assume that the random sample is selected from
the infinite population or from a finite population with replacement.
31
Sampling Distributions If the sampling is done without replacement from a finite normal population of
size N then the sample mean is distributed normally with mean
E X   
and variance
N  n 2
Var  X  
N 1 n
The proof of this formula is beyond the scope of this course.
Therefore, standard error of sample mean is given by

N  n 2
SE  X   Var  X  
N 1 n
If the sample is drawn from any population other than normal then by central
limit theorem, the sampling distribution of X tends to normal as the sample
size n increases, that means, the sample mean X is normally distributed with
mean µ and variance σ2/n when the sample size is large (  30).
Let us see an application of the sampling distribution of mean with the help of
an example.
Example 1: Diameter of a steel ball bearing produced on a semi-automatic
machine is known to be distributed normally with mean 12 cm and standard
deviation 0.1 cm. If we take a random sample of size 10 then find
(i) Mean and variance of sampling distribution of mean.
(ii) The probability that the sample mean lies between 11.95 cm and 12.05
cm.

Solution: Here, we are given that

 = 12, σ = 0.1, n = 10
(i) Since the population is normally distributed therefore the sampling
distribution of the sample mean also follows a normal distribution with
mean
E  X     12

and variance
 2 (0.1)2
Var  X     0.001
n 10
(ii) The probability that the sample mean lies between 11.95 cm and 12.05
cm is given by
P 11.95  X 12.05  See Fig.2.1
To get the value of this probability, we convert X into a standard normal
variate Z by the transformation
X  E X X  12 X  12 If X ~
Z  
0.03 X
Var  X  0.001 Z

normal
Therefore, by subtracting 12 from each term and then dividing each term ‘0’ and
by 0.03 in the above inequality, we get required probability as X
Z
32 
Simila
X~N
  11.95  12 X  12 12.05  12  Sampling Distribution(s)
P   of Statistic(s)
 0.03 0.03 0.03 

 P 1.67  Z  1.67

 P 1.67  Z  0  P0  Z  1.67
 P0  Z  1.67  P0  Z  1.67
 2P0  Z  1.67

 2  0.4525  Using table for area 

 under normal curve 
= 0.9050
Therefore, the probability that the sample mean lies between 11.95 and
12.05 cm is 0.9050.

X  11.95 X  12 X  12.05
Z  1.67 Z0 Z  1.67

Fig. 2.1
Now, continuing our discussion about the sampling distribution of mean, we
consider another situation.
In the cases described in last two pages, we assume that variance σ2 of the
normal population is known but in general it is not known and in such a
situation the only alternative left is to estimate the unknown σ2. The value of
sample variance (S2) is used to estimate the σ2 where,
1 n1 2
S2  
n  1 i1
 Xi  X 

Thus, in this case, it has been established that the sampling distribution of
X 
mean is not normal and the variate is known as t-variate and follows t-
S/ n
distribution with (n−1) df instead of normal distribution. This distribution has
wild applications in Statistics therefore it will be described in detail in Unit 3
of this course.
If n is sufficiently large   30  then we know that almost all the distributions
are very closely approximated by normal distribution. Thus in this case, t-
distribution is also approximated normal distribution. So for large sample size
  30  , the sampling distribution of mean also follows normal distribution
with mean µ and variance S2/n.
Sampling Distributions Now, try some exercises for your practice.
E1) If parent population is normal with mean  and variance 2 then, what
is sampling distribution of sample means?
E2) The weight of certain type of a truck tyre is known to be distributed
normally with mean 200 pounds and standard deviation 4 pounds. A
random sample of 10 tyres is selected.
(i) What is the sampling distribution of sample mean? Also obtain the
mean and variance of this distribution.
(ii) Find the probability that the mean of this sample is greater than or
equal to 202 pounds.
E3) The mean life of electric blubs produced by a company is 2550 hours.
A random sample of 100 bulbs is selected and the standard deviation is
found to be 54 hours. Find the mean and variance of the sampling
distribution of mean.

2.3 SAMPLING DISTRIBUTION OF DIFFERENCE

OF TWO SAMPLE MEANS
There are so many problems in business and economics where someone may
be interested to draw the inference about the difference of two population
means. For example, two manufacturing companies of blubs are produced
same type of bulbs and one may be interested to know which one is better than
The sa
the other, an investigator may want to know the difference of average income
of diffe
of the peoples living in a two cities, say, A and B, two different types of drugs, means
were tried on certain number of patients for controlling blood pressure and one sampli
may be interested to know which one has better effect on controlling blood differe
pressure, etc. Therefore, in such situation, to draw the inference we require the ( x  y)
sampling distribution of difference of two sample means. obtaine
possibl
Let the same characteristic measures from two populations be represented by the pop
X and Y variables and the variation in the values of these constitute two
population, say, population-I for variation in X and population-II for variation
in Y. Suppose population-I having mean 1 and variance 12 and population-II
having mean  2 and variance 22 . Then we take all possible samples of same
size n1 from population-I and then the sample mean, say, X is calculated for
each sample. Similarly, all possible samples of same size n2 are taken from the
population-II and the sample mean, say, Y is calculated for each sample.
Then we consider all possible differences of means X and Y. The difference
of these means may or may not differ from sample to sample and so we
construct the probability distribution of these differences. The probability
distribution thus obtained is known as sampling distribution of the difference
of sample means. Therefore, the sampling distribution of difference of sample
means can be defined as:
“The probability distribution of all values of the difference of two sample
means would be obtained by drawing all possible samples from both the
populations is called sampling distribution of difference of two sample
means.”

34
If both the parent populations are normal, that is, Sampling Distribution(s)
of Statistic(s)
  
X ~ N 1 , 12 and Y ~ N  2 ,  22 
then as we discussed in previous section that
 2   2 
X ~ N  1 , 1  and Y ~ N   2 , 2 
 n1   n2 
Thus, as we have discussed in Unit 4 of MST-003 that if two independent
random variables X and Y are normally distributed then the difference
(X  Y) of these also normally distributed. Therefore, in our case, the
sampling distribution of difference of two sample means (X  Y) also follows
normal distribution with mean
If X and Y are independent
E  X  Y   E  X   E  Y  = µ1 − µ2 random variables then
and variance E  X  Y   E  X  E  Y 
12  22 and
Var  X  Y   Var  X   Var  Y   
n1 n 2 Var  X  Y 
Therefore, standard error of difference of two sample means is given by  Var  X   Var  Y 
12  22
SE  X  Y   Var  X  Y   
n1 n 2
If the parent populations are not normal but n1 and n2 are large (  30) , then by
central limit theorem, the sampling distribution of (X  Y) is very closely
 2 2 
normally distributed with mean (µ1− µ2) and variance  1  2  .
 n1 n 2 
Let us see an application of the sampling distribution of difference of two
sample means with the help of an example.
Example 2: Electric CFL manufactured by company A have mean lifetime of
2400 hours with standard deviation 200 hours, while CFL manufactured by
company B have mean lifetime of 2200 hours with standard deviation of 100
hours. If random samples of 125 electric CFL of each company are tested, find
(i) The mean and standard error of the sampling distribution of the difference
of mean lifetime of electric CFLs.
(ii) The probability that the CFLs of company A will have a mean lifetime at
least 160 hours more than the mean lifetime of the CFLs of company B.
Solution: Here, we are given that
1 = 2400, σ1 = 200, 2 = 2200, σ2 = 100
n1 = n2 = 125
(i) Let X and Y denote the mean lifetime of CFLs taken from companies A
and B respectively. Since n1 and n2 are large (n1, n2 > 30) therefore, by the
central limit theorem, the sampling distribution of (X  Y) follows
normal distribution with mean
E(X  Y)  1  2  2400  2200  200
and variance
12 22
Var(X  Y)  
n1 n 2
35
Sampling Distributions 2002  1002
  400
125 125
Therefore, the standard error is given by

SE  X  Y   Var  X  Y   400  20

(ii) Here, we want to find out the probability which is given by

P  X  160  Y   P (X  Y)  160  See Fig. 2.2 
To get the value of this probability, we convert variate (X  Y) into a
standard normal variate Z by the transformation
(X  Y)  E(X  Y) (X  Y)  200 (X  Y)  200
Z  
Var(X  Y) 400 20

Therefore, by subtracting 200 from each term and then dividing each term
by 20 in the above inequality, we get the required probability
 (X  Y)  200 160  200 
P
20

20   PZ  2
 
 P  2  Z  0  P 0  Z  
 P 0  Z  2  P 0  Z  
= 0.4772 + 0.5 = 0.9772

X  Y  200 X  Y  160
Z  2.40 Z0
Fig. 2.2
Now, continuing our discussion about the sampling distribution of difference
of two means, we consider another situation.
If population variances 12 &  22 are unknown then we estimate 12 &  22 by the
values of the sample variances S12 &S22 of the samples taken from the first and
second population respectively. And for large sample sizes n1 and n2   30  ,
the sampling distribution of (X  Y) is very closely normally distributed with

 S2 S2 
mean (µ1− µ2) and variance  1  2  .
 n1 n 2 
If population variances 12 &  22 are unknown and 12  22   2 then σ2 is Sampling Distribution(s)
of Statistic(s)
estimated by pooled sample variance S2p where,
1
S2p   n1S12  n 2S22 
n1  n 2  2

and variate

t
 X  Y    1  2 
~ t (n1  n 2  2)
1 1
Sp 
n1 n 2
follows t-distribution with (n1 + n2 − 2) degrees of freedom. As similar as
sampling distribution of mean, for large sample sizes n1 and n2   30  the
sampling distribution of (X  Y) is very closely normally distributed with
1 1
mean (µ1− µ2) and variance S2p    .
 n1 n 2 
Now, you can try the following exercise.
E4) The average height of all the workers in a hospital is known to be 68
inches with a standard deviation of 2.3 inches whereas the average
height of all the workers in a company is known to be 65 inches with a
standard deviation of 2.5 inches. If a sample of 35 hospital workers and
a sample of 50 company workers are selected at random, what is the
probability that the sample mean of height of hospital workers is at
least 2 inch greater than that of company workers?

2.4 SAMPLING DISTRIBUTION OF SAMPLE

PROPORTION
In Section 2.2, we have discussed the sampling distribution of sample mean.
But in many real word situations, in business and other areas where the data
collected in form of counts or the collected data classified into two categories
or groups according to an attribute. For example, the peoples living in a
colony may be classified into two groups (male and female) with respect to the
characteristic sex, the patients in a hospital may be classified into two groups
as cancer and non-cancer patients, the lot of articles may be classified as
defective and non-defective, etc.
Generally, such types of data are considered in terms of proportion of
elements / individuals / units / items possess (success) or not possess (failure)
a given characteristic or attribute. For example, the proportion of female in the
population, proportion of cancer patents in a hospital, proportion of defective
articles in a lot, etc.
In such situations, we deal with population proportion instead of population
mean.
When population proportion is unknown and the whole population is too large
to find out the proportion. In such a situation, to draw the inference about the
population proportion, we require the sampling distribution of sample
proportion.

37
Sampling Distributions For sampling distribution of sample proportion, we draw all possible samples
from the population and for each sample we calculate the sample proportion p
as
X
p 1
n
where, X is the number of observations /individuals / items / units in the
sample which have the particular characteristic under study and n is the total
number of observations in the sample i.e. sample size.
For better understanding of the process, we consider the following example in
which size of the population is very small:
Suppose, there is a lot of 3 cartons A, B & C of electric bulbs and each carton
contains 20 bulbs. The number of defective bulbs in each carton is given
below:
Table 2.1: Number of Defective Bulbs per Carton
Carton Number of
Defectives Bulbs
A 2
B 4
C 1

The population proportion of defective bulbs can be obtained as

2  4 1 7
P 
20  20  20 60
Now, let us assume that we do not know the population proportion of
defective bulbs. So we decide to estimate population proportion of defective
bulbs on the basis of samples of size n = 2. There are Nn = 3 2 = 9 possible
samples of size 2 with replacement. The all possible samples and their
respective proportion defectives are given in the following table:
Table 2.2: Calculation of Sample Proportion
Sample Sample Sample Sample
Carton Observation Proportion(p)

1 (A, A) (2, 2) 4/40

2 (A, B) (2, 4) 6/40
3 (A, C) (2, 1) 3/40
4 (B, A) (4, 2) 6/40
5 (B, B) (4, 4) 8/40
6 (B, C) (4, 1) 5/40
7 (C, A) (1, 2) 3/40
8 (C, B) (1, 4) 5/40
9 (C, C) (1, 1) 2/40

From the above table, we can see that value of the sample proportion is
varying from sample to sample. So we consider all possible sample
proportions and calculate their probability of occurrence. Since there are 9
possible samples therefore the probability of selecting a sample is 1/9. Then
we arrange the possible sample proportion with their respective probability in
Table 2.3 given in next page:

38
 Table 2.3: Sampling Distribution of Sample Proportion Sampling Distribution(s)
of Statistic(s)
S.No. Sample Frequency Probability
Proportion(p)
1 2/40 1 1/9
2 3/40 2 2/9
3 4/40 1 1/9
4 5/40 2 2/9
5 6/40 2 2/9
6 8/40 1 1/9

This distribution is called the sampling distribution of sample proportion.

Thus, we can define the sampling distribution of sample proportion as: The sampling
distribution of the
“The probability distribution of all values of the sample proportion that sample proportion is a
obtained by drawing all possible samples of same size from the population is theoretical probability
called the sampling distribution of the sample proportion.” distribution of sample
proportions that would
This distribution also has its mean. Therefore, the mean of sampling be obtained by drawing
distribution of sample proportion can be obtained as all possible samples of
the same size from the
1 k k
population.
p 
K i1
p if i where, K   f i
i1

1 2 3 8  7
   1   2  ...   1 
9  40 40 40  60
Thus, we have seen that mean of sample proportion is equal to the population
proportion.
As we have already mentioned in the previous unit that finding mean, variance
and standard error from this process is tedious so we calculate these by
another short-cut method when population proportion is known.
As we discussed in Unit 5 of course MST-003 that if a population whose
elements are divided into two mutually exclusive groups− one containing the
elements which possess a certain attribute (success) and other containing
elements which do not possess the attribute(failure), then number of
successes(elements possess a certain attribute) follows a binomial distribution
with mean
E(X)  nP
and variance
Var(X)  nPQ where Q  1  P
where, P is the probability or proportion of success in the population.
Now, we can easily find the mean and variance of the sampling distribution of
sample proportion by using the above expression as
X
E(p)  E  
n
1 1
 E(X)  nP
n n
E ( p)  P
and variance
39
Sampling Distributions X
Var(p)  Var  
n
1
 Var(X)  Var  aX   a 2Var  X  
n2
1
 nPQ  Var  X   nPQ 
n2
PQ
Var(p) 
n
Also standard error of sample proportion can be obtained as

PQ
SEp  Var( p) 
n
If the sampling is done without replacement from a finite population then the
mean and variance of sample proportion is given by
E p  P

and variance
N  n PQ
Var  p  
N 1 n
where, N is the population size and the factor (N-n) / (N-1) is called finite
population correction.
If sample size is sufficiently large, such that np > 5 and nq > 5 then by central
limit theorem, the sampling distribution of sample proportion p is
approximately normally distributed with mean P and variance PQ/n where,
Q = 1‒ P.
Let us see an application of the sampling distribution proportion with the help
of an example.
Example 3: A machine produces a large number of items of which 15% are
found to be defective. If a random sample of 200 items is taken from the
population and sample proportion is calculated then find
(i) Mean and standard error of sampling distribution of proportion.

(ii) The probability that less than or equal to 12% defectives are found in the
sample.
Solution: Here, we are given that
15
P= = 0.15, n = 200
100
(i) We know that when sample size is sufficiently large, such that np > 5 and
nq > 5 then sample proportion p is approximately normally distributed
with mean P and variance PQ/n where, Q = 1– P. But here the sample
proportion is not given so we assume that the conditions of normality
hold, that is, np > 5 and nq > 5. So mean of sampling distribution of
sample proportion is given by
E ( p )  P  0 .15
40
 and variance Sampling Distribution(s)
of Statistic(s)
PQ 0.15  0.85
Var(p)    0.0006
n 200
Therefore, the standard error is given by
SE  p   Var  p   0.0006  0.025

(ii) The probability that the sample proportion will be less than or equal to
12% defectives is given by
P  p  0.12  See Fig. 2.3
To get the value of this probability, we can convert the random variate p
into standard normal variate Z by the transformation
p  E  p p  0.15 p  0.15
Z  
Var  p  0.0006 0.025

Therefore, by subtracting 0.15 from each term and then dividing each
term by 0.025 in the above inequality, we get required probability
 p  0.15 0.12  0.15 
P   PZ  1.20
 0.025 0.025 
 0.5  P 1.20  Z  0
 0.5  P 0  Z  1.20
 0.5  0.3849
= 0.1151

p  0.12 p  0.15
Z  1.20 Z0
Fig. 2.3
Now, it is time for you to try the following exercise to make sure that you
have learnt the sampling distribution of sample proportion.
E5) A state introduced a policy to give loan to unemployed doctors to start
own clinic. Out of 10000 unemployed doctors 7000 accept the policy
and got the loan. A sample of 100 unemployed doctors is taken at the
time of allotment of loan. What is the probability that sample
proportion would have exceeded 60% acceptance.
Sampling Distributions
2.5 SAMPLING DISTRIBUTION OF DIFFERENCE
OF TWO SAMPLE PROPORTIONS
In Section 2.3, we have discussed the sampling distribution of difference of
two sample means. In some cases, we are interested to comparative study that
the proportions of an attributes in the two different populations or group are
equal or not. For example, one may wish to test whether the proportions of
alcohol drinkers in the two cities are same, one may wish to compare the
proportions of literates between two groups of people, one may estimate the
decrease in the proportions of consumption of tea after the increase in excise
duty on it, etc. Therefore, we require the sampling distribution of difference of
two sample proportions for getting the inference about two populations.
Suppose, there are two populations, say, population-I and population-II under
study and the population-I having population proportion P1 and population-II
having population proportion P2 according to an attribute. For describing the
sampling distribution of difference of two sample proportions, we consider all
possible samples of same size n1 taken from population-I and for each sample
calculate the sample proportion p1 of success. Similarly, determine the sample
proportion p2 of success by considering all possible sample of same size n2
from population-II. Then we consider all possible differences of proportions
p1 and p2 . The difference of these proportions may or may not be differ so we
construct the probability distribution of these differences. The probability
distribution thus obtained is called the sampling distribution of the difference
of sample proportions. Therefore, the sampling distribution of difference of The sa
two sample proportions can be defined as: differe
propor
“The probability distribution of all values of the difference of two sample sampli
proportions that have been obtained by drawing all possible samples of same differe
sizes from both the populations is called sampling distribution of difference propor
between two sample proportions.” would
drawin
As we have seen in case of single proportion described in previous section that from b
if sample size is sufficiently large, such that np > 5 and nq > 5 then by central
limit theorem, the sampling distribution of sample proportion p is
approximately normally distributed with mean P and variance PQ/n where,
Q = 1‒ P. Therefore, if n1 and n2 are sufficiently large, such that n1p1  5,
n1q1  5, n2p2  5 and n2q2  5 then

 PQ   PQ 
p1 ~ N  P1, 1 1  and p2 ~ N  P2 , 2 2 
 n1   n2 

where, Q1 = 1‒ P1 and Q2 = 1‒ P2.

Also, by the property of normal distribution described in Unit 13 of MST 003,
the sampling distribution of the difference of sample proportions follows
normal distribution with mean
E(p 1-p2) = E(p1)-E(p 2) = P1-P2
and variance
P1Q1 P2Q2
Var(p1-p 2) = Var(p 1)+Var(p 2)  
n1 n2

That is,
42
  PQ P Q  Sampling Distribution(s)
p1  p 2 ~ N P1  P2 , 1 1  2 2  of Statistic(s)
 n1 n2 

Thus, standard error is given by

P1Q1 P2Q2
SE  p1  p2   Var  p1  p2   
n1 n2

Let us see an application of the sampling distribution of difference of two

sample proportions with the help of an example.
Example 4: In one population, 30% persons had blue-eyed and in second
population 20% had the same blue-eye. A random sample of 200 persons is
taken from each population independently and calculate the sample proportion
for both samples, then find the probability that the difference in sample
proportions is less than or equal to 0.02.
Solution: Here, we are given that
P1 = 0.30, P2 = 0.20,
n1 = n2 = 200
Let p 1 and p2 be the sample proportions of blue-eye persons in the samples
taken from both the populations respectively. We know that when n1 and n2
are sufficiently large, such that n1p1  5, n1q1  5, n2p2  5 and n2q2  5 then
sampling distribution of the difference between two sample proportions is
approximately normally distributed. But here the sample proportions are not
given so we assume that the conditions of normality hold. So mean of
sampling distribution of the difference between two sample proportions is
given by
E  p1  p 2   P1  P2  0.30  0.20  0.10

and variance
P1Q1 P2Q 2
Varp1  p 2   
n1 n2

0.30  0.70 0.20  0.80

 
200 200
 0.0019
The probability that the difference in sample proportions is less than or equal
to 0.02 is given by
P  p1  p 2  0.02  See Fig.2.4 

To get the value of this probability, we convert variate (p1−p2) into a standard
normal variate Z by the transformation
(p1  p 2 )  E  p1  p 2  (p1  p 2 )  0.10 (p1  p 2 )  0.10
Z  
Var  p1  p 2  0.0019 0.04

Therefore, by subtracting 0.10 from each term and then dividing each term by
0.04 in the above inequality, we get the required probability as

43
Sampling Distributions  p  p 2   0.10 0.02  0.10 
P 1   PZ  2.0
 0.04 0.04 
 0.5  P 2.0  Z  0
 0.5  P0  Z  2.0
 0.5  0.4772  0.0228

p1  p 2  0.02 p1  p2  0.10
Z  2.0 Z0
Fig. 2.4
Now, try to solve the following exercises to ensure that you understand this
section properly.
E6) In city A, 12% persons were found to be alcohol drinkers and in
another city B, 10% persons were found alcohol drinkers. If 100
persons of city A and 150 persons of city B are selected randomly,
what is the probability that the difference in sample proportions lies
between 0.01 and 0.08?

2.6 SAMPLING DISTRIBUTION OF SAMPLE

VARIANCE
In Sections 2.2 and 2.4, we have discussed the sampling distributions of
sample mean and sample proportion respectively. But many practical
situations concerned with the variability. For example, a manufacturer of steel
ball bearings may want to know about the variation of diameter of steel ball
bearing, a life insurance company may be interested in the variation of the
number of polices in different years, etc. Therefore, we need information
about the sampling distribution of sample variance.
For describing the sampling distribution of the sample variance, we consider
all possible sample of same size, say, n taken from the population having
variance σ2 and for each sample we calculate sample variance S2. The values
of S2 may vary from sample to sample so we construct the probability
distribution of sample variances. The probability distribution thus obtained is
known as sampling distribution of the sample variance. Therefore, the
sampling distribution of sample variance can be defined as:
“The probability distribution of all values of the sample variance would be
obtained by drawing all possible sample of same size from the parent
population is called the sampling distribution of the sample variance.” The
the sam
Let the characteristic measures from the population be represented by X Sampling Distribution(s)
variable and the variation in the values of X constitute a population. Let of Statistic(s)
X1 , X 2 , ..., X n be a random sample of size n taken from normal population
with mean µ and variance σ2. W.S. Gosset first described the sampling
distribution of sample variance S2. Here, it is to be noted that S2 always be
positive, therefore S2 does not follow normal distribution because the normal
variate takes positive as well as negative values. The distribution of sample
variance can not be obtained directly therefore in this case, some
transformation is made by multiplying S2 by (n −1) and then dividing the
product by σ2. The obtained new variable is denoted by the symbol 2 read as For degrees of
chi-square and follows the chi-square distribution with (n −1) degrees of freedom, please see
freedom as, Section 3.2 of next
unit.

2 
 n  1 S2 ~ 2
 n 1
2
where, S2 is called sample variance and given below
1 n 2
S2  
n  1 i1
 Xi  X  for all i 1, 2,..., n

The chi-square distribution has wild applications in statistics, therefore, it will

be described in detail in the next unit of this course.
For solving the problems relating to sampling distribution of variance, we
convert the variate S2 in a 2-variate by above transformation. For better
understanding, we explain it by taking an example given below.
Example 5: A manufacturer of steel ball bearings has found the variance of
diameter of ball bearings 0.0018 inches2. What is the probability that a random
sample of 25 ball bearings will result in a sample variance at least 0.002
inches2 ?
Solution: Here, we are given that
σ2 = 0.0018, n = 25
The probability that the sample variance is at least 0.002 inches2 is given by
P S2  0.002 

To get the value of this probability, we convert variate S2 into 2-variate which
follows the chi-square distribution with n-1 = 24 degrees of freedom by the
transformation
(n  1)S2 24S2
2n1  
2 0.0018
Therefore, multiplying each term by 24 and then dividing each term by 0.0018
in the above inequality. We get the required probability as
 24S2 24  0.002  2
P
0.0018

0.0018   P  24  26.67 
 

From the chi-square table, we see that P  2 24   15.66   0.90 and

 
P  24  33.20   0.10.
2

45
Sampling Distributions

Fig. 2.5
Therefore, the above probability must lie between 0.10 and 0.90, that is,
0.10  P  2  26.67  P  S2  0.002  0.90

Now, you can try the following exercises.

E7) The weight of certain type of truck tyres has a variance of 11 pounds2.
A random sample of 20 tyres is selected. What is the probability that
the variance of this sample is greater than or equal to 16 pounds2?

2.7 SAMPLING DISTRIBUTION OF RATIO OF

TWO SAMPLE VARIANCES
In previous section, we have discussed about the sampling distribution of
sample variance where we were dealing with one population. Now, one may
be interested to persue for the comparative study of the variances of the two
normal populations as we have done for the means and the proportions. For
example, suppose a quality control engineer wants to determine whether or not
the variance of the quality of the product is changing over the time, an
economist may wish to know whether the variability in incomes differs across
the two populations, etc. For solving such types of problems we require the
sampling distribution of the ratio of sample variances.
Suppose there are two normal populations, say, population-I and population-II
under study, the population-I has variance 12 and population-II has variance
22 . For describing the sampling distribution for ratio of population variances,
we consider all possible samples of same size n1 from population-I and for
each sample we calculate sample variance S12 . Similarly, calculate sample
variance S22 from each sample of same size n2 drawn from population-II.

Then we consider all possible values of the ratio of the variances S12 and S 22 ,
The values of S12 / S 22 may vary from sample to sample so we construct the
probability distribution of the ratio of the sample variances. The probability
distribution thus obtained is known as sampling distribution of the ratio of
sample variances. Therefore, the sampling distribution of ratio of sample
variances can be defined as:
“The probability distribution of all values of the ratio of two sample variances Sampling Distribution(s)
would be obtained by drawing all possible samples from both the populations of Statistic(s)
is called sampling distribution of ratio of sample variances.”
The sampling distribution of ratio of variances is given by Prof. R. A. Fisher
in 1924. According to Prof. R. A. Fisher, the ratio of two independent chi-
square variates when divided by their respective degrees of freedom follows
F-distribution as
2n1 1 /  n1  1
F
2n2 1 /  n 2  1

Since 2 
 n  1 S2 therefore,
2

F
 n1  1 S12 / 12  n1  1
 n 2  1 S22 / 22  n 2  1
S12 / 12
F ~ F n1 1, n2 1
S22 / 22

If 12  22 then

S12
F ~ F n1 1, n2 1
S22
Therefore, the sampling distribution of ratio of sample variances follows
F-distribution with (n1-1, n2-1) degrees of freedom. The F-distribution is
described in detail in fourth unit of this block.
With this we end this unit. We now summarise our discussion.

2.8 SUMMARY
In this unit, we have discussed following points:
1. The need of sampling distributions of different statistics.
2. The sampling distribution of sample mean.
3. The sampling distribution of difference between two sample means.
4. The sampling distribution of sample proportion.
5. The sampling distribution of difference of two sample proportions.
6. The sampling distribution of sample variance.
7. The sampling distribution of ratio of two sample variances.

2.9 SOLUTIONS / ANSWERS

E1) Since parent population is normal so sampling distribution of sample
means is also normal with mean  and variance 2/n.
E2) Here, we are given that
 = 200, σ = 4, n = 10
(i) Since parent population is normal so sampling distribution of
sample means is also normal. Therefore, the mean of this
47
Sampling Distributions distribution is given by
E  X     200

and variance
2
2  4  16
Var  X      1.6
n 10 10
(ii) Now, the probability that the sample mean X is greater than or
equal to 202 pounds is given by
P  X  202  See Fig. 2.6 
To get the value of this probability, we convert X into a standard
normal variate Z by the transformation
X  EX  X  200 X  200
Z  
Var  X  1.6 1.26

Therefore, by subtracting 200 from each term and then dividing

each term by 1.26 in the above inequality, we get required
probability as
 X  200 202  200 
P   P  Z  1.59 
 1.26 1.26 
 0.5  P  0  Z  1.59 
 0.5  0.4441 = 0.0559

X  200 X  202
Z0 Z  1.59
Fig. 2.6
E3) Here, we are given that
 = 2550, n = 100, S = 54
First of all, we find the sampling distribution of sample mean. Since
sample size is large (n = 100 > 30) therefore, by the central limit
theorem, the sampling distribution of sample mean follows normal
distribution. Therefore, the mean of this distribution is given by
E  X     2550
and variance
2
S2  54  2916
Var  X      29.16
n 100 100
E4) Here, we are given that Sampling Distribution(s)
of Statistic(s)
1 = 68, σ1 = 2.3, n1 = 35
2 = 65, σ2 = 2.5, n2 = 50
To find the required probability, first of all we find the sampling
distribution of difference of two sample means. Let X and Y denote
the mean height of workers taken from hospital and company
respectively. Since n1 and n2 are large (n1, n2 > 30) therefore, by the
central limit theorem, the sampling distribution of (X  Y) follows
normal distribution with mean
E  X  Y   1  2  68  65  3

and variance
12 22 2.3 2.5
2 2
Var  X  Y     
n1 n 2 35 50
 0.1511  0.1250  0.2761
The probability that the sample mean of height of hospital workers is 2
inch less than that of company workers is given by
P  X  Y   2  See Fig. 2.7 
To get the value of this probability, we convert variate  X  Y  into a
standard normal variate Z by the transformation

Z
 X  Y  E  X  Y  X  Y  3  X  Y  3
Var  X  Y  0.2761 0.53

Therefore, by subtracting 3 from each term and then dividing each

term by 0.53 in the above inequality, we get required probability
  X  Y   3 2  3
P    P Z  1.89  0.5  P 1.89  Z  0
 0.53 0.53 

 0.5  P 0  Z  1.89

 0.5  0.4706  0.0294

XY  2 X Y  3
Z  1.89 Z0

Fig. 2.7
Sampling Distributions E5) Here, we are given that
X 7000
N =10000, X = 7000  P    0.70 & n = 100
N 10000
First of all, we find the sampling distribution of sample proportion.
Here, the sample proportion is not given and n is large so we can
assume that the conditions of normality hold. So the sampling
distribution is approximately normally distributed with mean
E ( p )  P  0.70
and variance
PQ 0.70  0.30
Var(p)    0.0021  Q  1  P
n 100
Since the sample size n is less than 10% of the population size N,
Nn
therefore, correction factor is ignored.
N 1
The probability that the sample proportion would have exceeded 60%
acceptance is given by
P  p  0.60 See Fig. 2.8
To get the value of this probability, we can convert the random variate
p into standard normal variate Z by the transformation
p  E  p p  0.70 p  0.70
Z  
Var  p  0.0021 0.046

Therefore, by subtracting 0.70 from each term and then dividing each
term by 0.046 in the above inequality, we get the required probability
as
 p  0.70 0.60  0.70 
P   PZ  2.17
 0.046 0.046 
 P 2.17  Z  0  0.5
 P0  Z  2.17  0.5
 0.4850  0.5
= 0.9850

p  0.60 p  0.70
Z  2.17 Z0
Fig. 2.8
E6) Here, we are given that Sampling Distribution(s)
of Statistic(s)
12 10
P1   0.12, P2   0.10, n1 =100, n2 = 150
100 100
Let p 1 and p 2 be the sample proportions of alcohol drinkers in two
cities A and B respectively. Here the sample proportions are not given
and n1 and n2 are large  n1 , n 2  30  so we can assume that conditions
of normality hold. So the sampling distribution of difference of
proportions is approximately normally distributed with mean
Ep1  p 2   P1  P2  0.12  0.10  0.02
and variance
P1Q1 P2Q 2
Varp1  p 2   
n1 n2
0.12  0.78 0.10  0.90
   Q  1  P
100 150
 0.0009  0.0006  0.0015
The probability that the difference in sampling proportions lies
between 0.01 and 0.08 is given by
P 0.01  (p1  p2 )  0.08 See Fig. 2.9
To get the value of this probability, we convert variate (p1  p 2 ) into a
standard normal variate Z by the transformation
(p1  p 2 )  E  p1  p 2  (p1  p 2 )  0.02 (p1  p 2 )  0.02
Z  
Var  p1  p 2  0.0015 0.039

Therefore, by subtracting 0.02 from each term and then dividing each
term by 0.039 in the above inequality, we have required probability
 0.01  0.02 p1  p 2   0.02 0.08  0.02 
P  
 0.039 0.039 0.039 
 P 0.26  Z  1.54
 P 0.26  Z  0  P0  Z  1.54
 P0  Z  0.26  P0  Z  1.54
 0.1026  0.4382 = 0.5408

p1  p2  0.01 p1  p2  0.08
Z   0.26 Z  1.54
Fig. 2.9
Sampling Distributions E7) Here, we are given that
σ2 = 11, n = 20
The probability that the variance of the given sample is greater than 16
pounds2 is given by
P S2  16 
To get the value of this probability, we convert variate S2 into
2-variate which follows the chi-square distribution with n-1 = 19
degrees of freedom by the transformation

2n1 
 n  1 S2
19S2

2 11
Therefore, multiplying each term by 19 and then dividing each term by
11 in the above inequality, we get required probability as
19S2 19 16  2
P
11

11   P 19  27.64 
 
From the chi-square table, we see that P  219   27.20  0.10 and
 
P   219   30.14   0.05.

Fig. 2.10
Therefore, the above probability must lie between 0.05 and 0.10, that
is,
0.05  P  2  29.09  P  S2  16  0.10.
UNIT 3 STANDARD SAMPLING
DISTRIBUTIONS-I
Structure
3.1 Introduction
Objectives
3.2 Introduction to 2-distribution
3.3 Properties of 2-distribution
Probability Curve of 2-dsistribution
Mean and Variance of 2-distribution
3.4 Applications of 2-distribution
3.5 Introduction to t-distribution
3.6 Properties of t-distribution
Probability Curve of t-distribution
Mean and Variance of t-distribution
3.7 Applications of t-distribution
3.8 Summary
3.9 Solutions /Answers

3.1 INTRODUCTION
In Unit 1 of this block, we have discussed the fundamentals of sampling
distributions and some basic definitions related to it. In Unit 2, we have
discussed the sampling distributions of various sample statistics such as
sample mean, sample proportion, sample variance, etc. Sometimes statistics
such as sample mean, sample proportion, sample variance, etc. may follow a
particular sampling distribution. Based on this fact Prof. R.A. Fisher, Prof.
G. Snedecor and some other statisticians worked in this area and found some
exact sampling distributions which are often followed by some of the
statistics. In present and next unit of this block, we shall discuss some
important sampling distributions such as 2(read as chi-square), t and F.
Generally, these sampling distributions are named on the name of the
originator, that is, Fisher’s F-distribution is named on Prof. R.A. Fisher.
This unit is divided in 9 sections. Section 3.1 is introductive in nature. In
Section 3.2 the brief introduction of 2-distribution is given. The different
properties and probability curves for various degrees of freedom n = 1, 4, 10
and 22 are described along with the mean and variance in Section 3.3. The
distribution of  2 is used in testing the goodness of fit, independence of
attributes, etc. Therefore, in Section 3.4 the important applications of 2-
distribution are listed. In Section 3.5, the brief introduction of t-distribution
is described. Properties of t-distribution, probability curve, the general
formula for central moment, mean and variance are described in Section 3.6.
Different applications of t-distribution are listed in Section 3.7. Unit ends by
providing summary of what we have discussed in this unit in Section 3.8 and
solution of exercises in Section 3.9.
Objectives
After studying this unit, you should be able to:
 explain the 2-distribution;
53
Sampling Distributions  describe the properties of 2-distribution;
 explain the probability curve of χ2-distribution;
 find the mean and variance of 2-distribution;
 explore the applications of 2-distribution;
 explain the t-distribution;
 describe the properties of t-distribution;
 describe the probability curve of t-distribution;
 find the mean and variance of t-distribution; and
 explore the applications of t-distribution.

3.2 INTRODUCTION TO χ2-DISTRIBUTION

Before describing the chi-square (χ2) distribution, first we will discuss very
useful concept “degrees of freedom”. The exact sampling distributions are
described with the help of degrees of freedom.
Degrees of Freedom (df)
The term degree of freedom (df) is related to the independency of sample
observations. In general, the number of degree of freedom is the total number
of observations minus the number of independent constraints or restrictions
imposed on the observations.
For a sample of n observations, if there are k restrictions among observations
(k < n), then the degrees of freedom will be (n–k).
For example, suppose there are two observations X1, X2 and a restriction
(condition) is imposed that their total should be equal to 100. Then we can
arbitrarily assign value to any one of these two as 30, 98, 52, etc. but the
value of rest is automatically determined by a simple adjustment like
X 2  (100  X1 ). Therefore, we can say that only one observation is
independent or there is one degree of freedom.
Consider another example, suppose there are 50 observations as X1, X2, …,
X50 and the restriction (condition) are imposed on these like as:
50 50
2
(i)  Xi  2000 (ii)
i 1
X
i 1
i  10000 (iii) X 5  2X 4

The restriction (i) reveals sum of observations equal to 2000 (ii) reveals sum
of square of observations equal to 10000 and restriction (iii) implies fifth
observation double to the fourth observation.
With these three restrictions (i), (ii) and (iii), there shall be only 47
observations independently chosen and rest three could be obtained by
solving equations or by adjustments. The degrees of freedom of this set of
observations of size 50 is now (50 –3) = 47.
Chi-square Distribution
The chi-square distribution is first discovered by Helmert in 1876 and later
independently explained by Karl- Pearson in 1900. The chi-square
distribution was discovered mainly as a measure of goodness of fit in case of
frequency.
54
If a random sample X1, X2, …, Xn of size n is drawn from a normal Standard Sampling
population having mean  and variance σ2 then the sample variance can be Distributions -I
defined as
1 n 2
S2  
n  1 i 1
 Xi  X

or
n
2
X
i 1
i  X    n  1 S2  S2

where, ν = n −1 and the symbol ν read as ‘nu’. Generally, ν symbol is

used to represent the
 S2 general degrees of
Thus, the variate 2  , which is the ratio of sample variance multiplied freedom. Its value may
2 be n, n – 1, etc.
by its degrees of freedom and the population variance follows the 2-
distribution with ν degrees of freedom.
The probability density function of 2-distribution with ν df is given by
1 2   / 2 1
f  2   e  /2
 
2
; 0  2   … (1)
/2 
2
2
where, ν = n −1.
The chi-square distribution can also be we described as:
If random variable X follows normal distribution with mean µ and variance
X 
σ2, then Z  ~ N 0,1 , known as standard normal variate, follows

normal distribution with mean 0 and variance 1. The square of a standard
normal variate Z follows chi-square distribution with 1 degree of freedom
(df), that is,
X 
If X ~ N  ,  2 , then Z  ~ N 0,1 and

2
 X 
Z2   2
 ~  (1)
  
where, (21) read as chi-square with one degree of freedom.

In general, if Xi’s (i =1, 2,…, n) are n independent normal variates with

means µi and variances 2i i  1, 2,..., n  then the sum of squares of n standard
normal variate follows chi-square distribution with n df i.e.
Xi  i
If X i ~ N  i , 2i  , then Zi  ~ N 0,1
i

n n 2
 X i  i 
Therefore,  2   Z2i     ~  2( n )
i 1 i 1   i 

The probability density function of 2-distribution with n df is given by

55
Sampling Distributions 1 2  n / 2 1
f  2   e  /2
 
2
; 0  2   … (2)
n/2 n
2
2
Try the following exercise to make sure that you have understood the chi-
square distribution.

E1) Write down the pdf of chi-square distribution in each of the following
cases:
(i) 6 degrees of freedom
(ii) 10 degrees of freedom
E2) Below, in each case, the pdf of chi-square distribution is given.
Obtain the degrees of freedom of each chi-square distribution:
1  2 / 2 2 3
(i) f  2   e   ; 0  2  
96
1 2 / 2
(ii) f  2   e ; 0  2  
2

3.3 PROPERTIES OF χ2-DISTRIBUTION

In previous section, we have discussed the 2-distribution and its probability
density function. Now, in this section, we shall discuss some important
properties of 2-distribution as given below:
1. The probability curve of the chi-square distribution lies in the first
quadrant because the range of  2 -variate is from 0 to ∞.

2. Chi-square distribution has only one parameter n, that is, the degrees of
freedom.
3. Chi-square probability curve is highly positive skewed.
4. Chi-square-distribution is a uni-modal distribution, that is, it has single
mode.
5. The mean and variance of chi-square distribution with n df are n and 2n
respectively.
After deliberating the properties of 2-distribution above, some of these are
discussed in detailed in the subsequent sub-sections.

3.3.1 Probability Curve of 2-distribution

The probability distribution of all possible values of a statistic is known as
the sampling distribution of that statistic. The shape of probability
distribution of a statistic can be shown by the probability curve. In this sub-
section, we will discuss the probability curve of chi-square distribution.
A rough sketch of the probability curve of chi-square distribution is shown in
Fig. 3.1 at different values of degrees of freedom as n = 1, 4, 10 and 22.

56
 Standard Sampling
Distributions -I

Fig. 3.1: Chi-square probability curves for n = 1, 4, 10 and 22

After looking the probability curves of 2-distribution for n = 1, 4, 10 and 22,

one can understand that probability curve takes shape of inverse J for n = 1.
The probability curve of chi-square distribution gets skewed more and more
to the right as n becomes smaller and smaller. It becomes more and more
symmetrical as n increases because as n tends to ∞, the 2-distribution
becomes normal distribution.
3.3.2 Mean and Variance of χ2-distribution
In previous sub-section, the probability curve of χ2-distribution is discussed
with some of its properties. Now in this sub-section, we will discuss the
mean and variance of the χ2-distribution. The mean and variance of
χ2-distribution is derived with the help of the moment about origin. As we
have discussed in Unit 3 of MST-002 that the first order moment about
origin is known as mean and central second order moment is known as
variance of the distribution. So we first obtain the rth order moment about
origin of χ2-distribution as
r  r
r  E    2       2  f   2  d 2
  o

2 r 1 2  n / 2  1

0
  n
e /2
 
2
d2
2n / 2
2
1 n 
1   2   r 1

n 0
e 2
  2  2  d 2
2n / 2
2 The Gamma function
is discussed in Unit 15
n of MST-003,
r    ax n 1
1 2 n according to this
 n  0 e x dx  n  … (3)
n
n/2 n 1 2
r
 a  

e  ax x n 1dx 
2 0 an
2 2
 
Now, if we put r = 1 in the above expression given in equation (3), we get
the value of first order moment about origin which is known as mean i.e.
Sampling Distributions n
1 1
Mean  1  2
n
n/2 n 1 2 1
2  
2 2
 
n n
1 2 2 n
  n  1  n n 
n
n 1 2 1  
n/2
2  
2 2
 
Similarly, if we put r = 2 in the formula of r given in equation (3) then we
get the value of second order moment about origin i.e.
n
1 2
2  2
n
n/2 n 1 2 2
2  
2 2
 

 n  n  n
1
1  2   2  2  n  2   n  1 n  1
 2  
n 1    n  1 n n 
2  
2
 n  n 
 2 2   1    n  n  2 
 2  2 
Now, we obtain the value of variance by putting the value of first order and
second order moments about origin in the formula given below
2
Variance  2  1 
2
 n  n  2    n   n 2  2n  n2  2n

So the mean and variance of the chi-square distribution with n df are given
by
Mean = n and Variance = 2n
Hence, we have observed that the variance is twice of the mean of chi-square
distribution.
Similarly, on putting r = 3, 4,… in the formula given in equation (3) one may
get higher order moments about origin such as 3 , 4 ,...
Example 1: What are the mean and variance of chi-square distribution with
5 degrees of freedom?
Solution: We know that the mean and variance of chi-square distribution
with n degrees of freedom are
Mean = n and Variance = 2n
In our case, n = 5, therefore,
Mean = 5 and Variance = 10
To evaluate your understanding try to answer the following exercises.
58
E3) What are the mean and variance of chi-square distribution with 10 Standard Sampling
Distributions -I
degrees of freedom?
E4) What are the mean and variance of chi-square distribution with pdf
given below
1 2 / 2 2 3
f  2   e   ; 0  2  
96

3.4 APPLICATIONS OF χ2-DISTRIBUTION

The applications of chi-square distribution are very wide in Statistics. Some
of them are listed below. The chi-square distribution is used:
1. To test the hypothetical value of population variance.
2. To test the goodness of fit, that is, to judge whether there is a discrepancy
between theoretical and experimental observations.
3. To test the independence of two attributes, that is, to judge whether the
two attributes are independent.
The first application listed above shall be discussed in detail in Unit 12 of
this course and the remaining two applications shall be discussed in detail in
Unit 16 of this course.
Now, try to write down the applications of chi-square distribution by
answering the following exercise.
E5) List the applications of chi-square distribution.

3.5 INTRODUCTION TO t-DISTRIBUTION

The t-distribution was discovered by W.S. Gosset in 1908. He was better
known by the pseudonym ‘Student’ and hence t-distribution is called
‘Student’s t-distribution’.
If a random sample X1, X2, …, Xn of size n is drawn from a normal
population having mean  and variance σ2 then we know that the sample
mean X is distributed normally with mean  and variance 2 / n , that is, if
 
Xi ~ N , 2 then X ~ N ,  2 / n  , and also the variate

X 
Z
/ n
is distributed normally with mean 0 and variance 1, i.e. Z ~ N 0, 1 .
In general, the standard deviation σ is not known and in such a situation the
only alternative left is to estimate the unknown σ2. The value of sample
variance (S2) is used to estimate the unknown σ2 where,
1 n1 2
S2  
n  1 i 1
 Xi  X 

X 
Thus, in this case the variate is not normally distributed whereas it
S/ n

59
Sampling Distributions follows t-distribution with (n−1) df i.e.
X 
t ~ t n1 … (4)
S/ n  
n
1 2
where, S2    Xi  X 
 n  1 i1
The t-variate is a widely used variable and its distribution is called student’s
t-distribution on the pseudonym name ‘Student’ of W.S. Gosset. The
probability density function of variable t with (n-1) = ν degrees of freedom
is given by
1
f  t   1 / 2
;  t  … (5)
2
 1   t 
 B  ,  1  
 2 2  

1 
where, B ,  is known as beta function and
2 2
1 
1   a b
B ,   2 2  B  a, b   
 2 2   1  a b
2
Later on Prof. R.A. Fisher found that t-distribution can also be applied to test
the regression coefficient and other practical problems. He proposed t-variate
as the ratio of standard normal variate to the square root of an independent
chi-square variate divided by its degrees of freedom. Therefore, if Z is a
standard normal variate with mean 0 and variance 1 and χ2 is an independent
chi-square variate with n df i.e.
Z ~ N 0, 1 and χ 2 ~ χ 2n 

Then Fisher’s t-variate is given by

Z
t ~ t  n … (6)
2 / n
That is, the t-variate follows the t-distribution with n df and has the same
probability density function as student t-distribution with n df. Therefore,
1
f  t   n 1 / 2
;  t   … (7)
2
 1 n  t 
nB  ,   1  
 2 2  n
After describing the t-distribution, we try to calculate the value of t-variate as
in the given example.
Example 2: The life of light bulbs manufactured by the company A is
known to be normally distributed. The CEO of the company claims that an
average life time of the light bulbs is 300 days. A researcher randomly
selects 25 bulbs for testing the life time and he observed the average life time
of the sampled bulbs is 290 days with standard deviation of 50 days.
Calculate value of t-variate.

60
Solution: Here, we are given that Standard Sampling
Distributions -I
  300, n  25, X  290 and S  50
The value of t-variate can be calculated by the formula given below
X 
t
S/ n

Therefore, we have
290  300 10
t   1
50/ 25 10
Now, try the following exercises for your practice.

E6) Write down the pdf of t-distribution in each of the following cases:
(i) 3 degrees of freedom
(ii) 9 degrees of freedom
E7) Obtain the degrees of freedom of t-distribution whose pdf is given
below:
1
f t  ;  t
2 3
 1 5   t 
5 B  ,  1  
 2 2  5
E8) If the scores on an IQ test of the students of a class are assumed to be
normally distributed with a mean of 60. From the class, 15 students
are randomly selected and an IQ test of the similar level is conducted.
The average test score and standard deviation of test scores in the
sample group are found to be 65 and 12 respectively. Calculate the
value of t-variate.

3.6 PROPERTIES OF t-DISTRIBUTION

In previous section, we have discussed the t-distribution briefly. Now, we
shall discuss some of the important properties of the t-distribution.
The t-distribution has the following properties:
1. The t-distribution is a uni-modal distribution, that is, t-distribution has
single mode.
n
2. The mean and variance of the t-distribution with n df are zero and if
n2
n > 2 respectively.
3. The probability curve of t-distribution is similar in shape to the standard
normal distribution and is symmetric about t = 0 line but flatter than
normal curve.
4. The probability curve is bell shaped and asymptotic to the horizontal
axis.
The properties deliberated above are some of the important properties of the
t-distribution. We shall discuss some of these properties such as probability
curve, mean and variance of the t-distribution in detail in the subsequent
sub-sections.
61
Sampling Distributions 3.6.1 Probability Curve of t-distribution
The probability curve of t-distribution is bell shaped and symmetric about
t = 0 line. The probability curves of t-distribution is shown in Fig. 3.2 at two
different values of degrees of freedom as at n = 4 and n = 12.

Fig. 3.2: Probability curves for t-distribution at n = 4, 12 along with standard normal
curve
In the figure given above, we have drawn the probability curves of
t-distribution at two different values of degrees of freedom with probability
curve of standard normal distribution. By looking at the figure, one can
easily understand that the probability curve of t-distribution is similar in
shape to that of normal distribution and asymptotic to the horizontal-axis
whereas it is flatter than standard normal curve. The probability curve of the
t-distribution is tending to the normal curve as the value of n increases.
Therefore, for sufficiently large value of sample size n(> 30), the t-
distribution tends to the normal distribution.
3.6.2 Mean and Variance of t-distribution
In previous sub-section, we have discussed the probability curve of
t-distribution. After discussing the probability curve it came to our
knowledge that t-distribution is symmetrical about t = 0 line. Since f (t) is
symmetrical about t = 0 line, therefore, all moments of odd order about
origin vanish or become zero, i.e.
 2r 1  0 for all r  0, 1, 2, 3... … (8)
In particular, for r = 0, we get
1  0  Mean
That is, the mean of the t-distribution is zero. Similarly, one can show that
other odd moments about origin are zero by putting r = 1, 2,….
The moments of even order are given by

2 r   t 2 r f t dt

Now, we can also write If f(x
 abou
2r  2  t 2r f  t  dt
0

1  t 2r
2   n 1 / 2
dt
1 n 0
 t2 
n B ,  1 
2 2  n 

 2 r Standard Sampling
2  t  … (9) Distributions -I

1 n 0
 t2 
 n 1 / 2 dt
n B ,  1 
2 2  n 

t2 1 1 
Putting 1    t 2  n  1
n y y 
n n n
Therefore, 2t dt   2
dy  2dt   2 dy  dy
y ty 1 
n   1 y 2
y 
Also t  0  y  1 and t    y  0
By putting these values in equation (9), we get
r
  1 
 n  y  1 
1 0
   n
2r     n 1 / 2
. dy
1 n 1
1 1  2
n B ,  n   1 y
2 2 y
  y   0 f  x  dx   a f  x  dx 
 a 0 
1
r
n 1
n   1 
1 2
2
2
   n   1  y dy
 1 n  0   y 
B , 
2 2
1
r n 1
nr  1 y 
1 2
2
2
 0  y  y dy
1 n
B , 
2 2
n 1 1
nr 1 r
1 2 r 

1 n 0 1  y  2 y 2 2
dy
B , 
2 2
n
nr 1  r 1 1
r  1
  y2 1  y  2 dy
1 n 0
B , 
2 2
 For a  0 & b  0
nr n 1  1 x a 1 1  x  b 1 dx 
 0
 B   r, r  
1 n 2 2 
B ,     B  a, b  
2 2
1 n n 1
 r r
 nr 2 2 2 2 ; n  r  0  B  a, b   a b 
 
1 n n 1 2  ab 
rr
2 2 2 2
n 1
r r
2r  n r 2 2; n  2r …(10)
1 n
2 2
63
Sampling Distributions Now, if we put r = 1 in the above expression given in equation (10), we get
the value of second order moment about origin i.e.
n 1
1 1
2  n 2 2; n2
1 n
2 2
n 1 1
1
n 2 2 2  n  1  n n 
1 n  n  
 1 1
2 2  2
n
 ; n2
n  2
Now, we obtain the value of variance by putting the value of first order and
second order moments about origin in the formula given below
2
Variance  2  1 

n 2
  0
 n  2 
n
 ; n2
 n  2
So the mean and variance of the t-distribution with n df are given by
n
Mean = 0 and Variance  ; n2
 n  2
Now, you can try an exercise to see how much you learn about the properties
of t-distribution.
E9) What are the mean and variance of t-distribution with 20 degrees of
freedom?

3.7 APPLICATIONS OF t-DISTRIBUTION

In previous section of this unit, we have discussed some important properties
of t-distribution and derived its mean and variance. Now, you may be
interested to know the applications of the t-distribution. In this section, we
shall discuss the important applications of the t-distribution. The
t-distribution has wide number of applications in Statistics. The t-distribution
is used:
1. To test the hypothesis about the population mean.
2. To test the hypothesis about the difference of two population means of
two normal populations.
3. To test the hypothesis that population correlation coefficient is zero.
The applications of the t-distribution listed above are discussed further in
Unit 12 of this course.

64
Now, it is time to write down the main applications of t-distribution by Standard Sampling
answering the following exercise. Distributions -I

E10) Write any three applications of t-distribution.

We now end this unit by giving a summary of what we have covered in it.

3.8 SUMMARY
In this unit, we have covered the following points:
1. The chi-square distribution.
2. The properties of χ2-distribution.
3. The probability curve of χ2-distribution.
4. Mean and variance of χ2-distribution.
5. Applications of χ2-distribution.
6. Student’s-t and Fisher’s t-distributions.
7. Properties of t-distribution.
8. The probability curve of t-distribution.
9. Mean and variance of t-distribution.
10. Applications of t-distribution.

3.9 SOLUTIONS / ANSWERS

E1) We know that the probability density function of 2-distribution with
n df is given by
1 2  n / 2  1
f  2   e  /2
  2
; 0  2  
n
2n / 2
2
(i) In our case, n = 6, therefore, for n = 6 df the pdf of 2-distribution
is given by

1 2  6 / 2 1
f  2   e  /2
  2
; 0  2  
6
26 / 2
2

1  2 / 2 2 2

8 3
e  
1 2 / 2 2 2

16
e   ; 0  2    3  2  2
 

(ii) Similarly, for n = 10 df the pdf of 2-distribution is given by

1 2 10 / 2  1
f  2   e  /2
 2
; 0  2  
10
210 / 2
2

1 2 4
 e   / 2  2 
32  5
65
Sampling Distributions 1 2 / 2 2 4

768
e   ; 0  2    5  4  24 

E2) We know that the probability density function of 2-distribution with

n df is given by
1 2  n / 2 1
f  2   e  /2
 
2
; 0   2   … (11)
n
2n / 2
2
(i) Here, we are given that
1  2 / 2 2 3
f  2   e   ; 0  2  
96
The given pdf can be arranged as
8
1 2
2 2 1
 8/ 2 8

f  2   e  /2
   96  16  6  2 
8  2
28 / 2
2
Comparing with equation (11), we get n = 8. So degree of
freedom of chi-square distribution is 8.
(ii) Here, we are given that
1  2 / 2
f  2   e ; 0  2  
2
The given pdf can be arranged as
2
1  2 / 2 2 2 1
 2/ 2 2

f  2
 e    2  2  1  2 
2  2
22 / 2
2
Comparing with equation (11), we get n = 1. So degree of
freedom of chi-square distribution is 1.
E3) We know that the mean and variance of chi-square distribution with n
degrees of freedom are
Mean = n and Variance = 2n
In our case, n = 10, therefore,
Mean = 10 and Variance = 20
E4) Here, we are given that
1 2 / 2 2 3
f  2   e   ; 0  2  
96
As similar as (i) of E2), we get n = 8, therefore,
Mean = n = 8 and Variance = 2n = 16.
E5) Refer Section 3.4.

E6) We know that the probability density function of t-distribution with n

df is given by

66
 1 Standard Sampling
f  t   n 1 / 2
;  t   Distributions -I
2
 1 n  t 
nB  ,   1  
 2 2  n
(i) In our case, n = 3, therefore, for n = 3 df the pdf of t-distribution
is given by
1
f  t  31 / 2
;  t 
2
 1 3  t 
3 B  ,  1  
 2 2  3

1 3

 2 2
2
 1 3   t2 
3  1  
 2 2  3

2  n n  n 
 2     1  1
 1 1 1   t2   2 2  2 
3   1  
 2 2 2  3

2  1 
 2
;    t      and 2  1 
 t2   2 
3 1  
 3
(ii) Similarly, for n = 9 df the pdf of t-distribution is given by
1
f  t   91 / 2
;  t 
2
 1 9  t 
9 B  ,  1  
 2 2  9

1 9

 2 2
5
 1 9  t2 
3  1  
 2 2  9

5  n n  n 
 5     1   1
 1 7 5 3 1 1   t2   2 2  2 
3   1  
 2 2 2 2 2 2  9

24  1 
 5    and 5  4  24 
 2
2
 105   t  
3   1  
 16   9

128
 5
;  t 
 t2 
105  1  
 9

E7) We know that the probability density function of t-distribution with n

df is given by
67
Sampling Distributions 1
f  t   n 1 / 2
;  t   … (12)
2
 1 n  t 
nB  ,  1  
 2 2  n
Here, we are given that
1
f t  3
;  t
 1 5  t2 
5 B  ,  1  
 2 2  5

1
 51 / 2
;  t  
2
 1 5  t 
5 B  ,  1  
 2 2  5
Comparing with equation (12), we get n = 5. So degrees of freedom
of t-distribution are 5.
E8) Here, we are given that
  60, n  15, X  65 and S  12
We know that the t-variate is
X
t
S/ n
Therefore, we have
65  60 5
t   1.14
12 / 15 4.39
E9) We know that the mean and variance of t-distribution with n degrees
of freedom are
n
Mean = 0 and Variance  ; n2
 n  2
In our case, n = 20, therefore,
20 10
Mean = 0 and Variance  
 20  2  11
E10) Refer Section 3.7.

68
UNIT 4 STANDARD SAMPLING
DISTRIBUTIONS-II
Structure
4.1 Introduction
Objectives
4.2 Introduction to F-distribution
4.3 Properties of F-distribution
Probability Curve of F-distribution
Mean and Variance of F-distribution
4.4 Applications of F-distribution
4.5 Relation between t, χ2 and F-distributions
4.6 Tabulated Values of t, χ2 and F
Tabulated Value of t-variate
Tabulated Value of χ2-variate
Tabulated Value of F-variate
4.7 Summary
4.8 Solutions /Answers

4.1 INTRODUCTION
In Unit 3, we have discussed chi-square and t-distributions in detail with
their properties and applications. As we know that the sampling distributions
χ2, t and F are mostly in use. In this unit, we shall discuss the F-distribution
in brief with their properties and applications. As we have said that F-
distribution was introduced by Prof. R. A. Fisher and it is the ratio of two
independent chi-square variates when divided by its respective degrees of
freedom. Later on the F-distribution was also described by the square of
t-variate.
As usual after giving a brief introduction of this unit in this Section 4.1, we
start our discussion by giving an over view of the F-distribution in Section
4.2. Different properties of F-distribution are explained in Section 4.3. The
probability curves of F-distribution for various degrees of freedom
(1  5, 2  5), (1  5, 2  20)and (1  20, 2  5) are described along with
the rth moment about origin, mean and variance. The distribution of F is used
in testing the equality of two variances, equality of several population means,
etc. Therefore in Section 4.4 the different applications of F-distribution are
explored. As mentioned above, all the standard sampling distribution
discussed in this unit and previous unit are very useful and inter-related.
Therefore, in Section 4.5 the relations between the t, χ2 and F -distributions
are obtained. In Section 4.6, the method of obtaining the tabulated value of a
variate is explained which follows either of t, χ2 and F-distributions. Unit
ends by providing summary of what we have discussed in this unit in Section
4.7 and solution of exercises in Section 4.8.
Objectives
After studying this unit, you should be able to:
 introduce the F-distribution;
 explain the properties of F-distribution;
69
Sampling Distributions  describe the probability curve of F-distribution;
 derive the mean and variance of F-distribution;
 explore the applications of F-distribution;
 describe the relations between t, χ2 and F-distributions; and
 explain the method of obtaining the tabulated value of a variate which
follows either of t, χ2 and F-distributions.

4.2 INTRODUCTION TO F-DISTRIBUTION

As we have said in previous unit that F-distribution was introduced by
Prof. R. A. Fisher and defined as the ratio of two independent chi-square
variates when divided by their respective degrees of freedom. If we draw a
random sample X1 , X 2 ,...,X n1 of size n1 from a normal population with mean
1 and variance σ12 and another independent random sample Y1 , Y2 ,..., Yn2 of
size n2 from another normal population with mean 2 and variance 22
respectively then as we have studied in Unit 3 that 1S12 / 12 is distributed as
chi-square variate with ν1 df i.e.
ν1S12
  2 ~ (21 )
2
1 … (1)
σ1
1 n1 1 n1 2
where, 1  n1  1, X  
n1 i1
X i and S2
1  
n1  1 i1
 Xi  X 

Similarly,  2S 22 /  22 is distributed as chi-square variate with ν 2 df i.e.

ν2S22
22  2
~ χ 2 ν2  … (2)
σ2
1 n2 1 n2 2
where, 2  n2  1, Y  
n2 i1
Yi and S2
2  
n2  1 i1
 Yi  Y 

Now, if we take the ratio of the above chi-square variates given in equations
(1) and (2), then we get
χ12 ν1S12 /σ12

χ 22 ν2S22/σ 22

S12 /σ12 χ12 /ν1

  ~ F ν1 ,ν 2  … (3)
S22 /σ22 χ 22 /ν 2
By observing the above form given in equation (3) we reveal that the ratio of
two independent chi-square variates when divided by their respective degrees
of freedom follows F-distribution where, ν1 and ν 2 are called the degrees of
freedom of F-distribution.
Now, if variances of both the populations are equal i.e. σ12  σ 22 then
F-variate is written in the form of ratio of two sample variances i.e.
S12
F ~ F 1 ,2  … (4)
S22

70
By observing the above form of F-variate given in equation (4) we reveal Standard Sampling
Distributions-II
that under the assumption 12  22 , the ratio of two independent random
sample variances follows F-distribution with ν1 and ν 2 df. The probability
density function of F-variate is given as
ν1 / 2
 ν1/ν 2  F ν1 / 2  1
f(F)   ν1  ν2  / 2
; 0F … (5)
ν ν 
B  1 , 2   1  ν1 F 
 2 2  ν 2 

After describing the F-distribution, we try to calculate the value of F-variate
as in the given example.
Example 1: A statistician selects 7 women randomly from the population of
women, and 12 men from a population of men. The table given below shows
the standard deviation of each sample and population:
Population Sample Standard
Population
Standard Deviation Deviation

Women 40 45
Men 80 75

Compute the value of F-variate.

Solution: Here, we are given that
n1  7, n 2  12, 1  40, 2  80, S1  45, S2  75
The value of F-variate can be calculated by the formula given below
S12 /σ12
F
S22 /σ 22
where, S12 &S22 are the sample variances.
Therefore, we have
2 2
 45 /  40
F 2 2
 75 / 80
1.27
  1.93
0.88
For the above calculation, the degrees of freedom ν1 for women’s data are
7 −1 = 6 and the degrees of freedom ν2 for men’s data are 12 −1 =11.
Now, try to answer the following exercises to see how much you learn about
F-distribution.
E1) Write down the pdf of F-distribution in each of the following cases:
(i) (4, 2) degrees of freedom
(ii) (2, 4) degrees of freedom
E2) If random variable X follows the F-distribution whose pdf is given by
1
f x  2
; 0 x  
1  x 
Obtain the degrees of freedom of this distribution.
71
Sampling Distributions E3) For the purpose of a survey 15 students are selected randomly from
class A and 10 students are selected randomly from class B. At the
stage of the analysis of the sample data, the following information is
available:

Class Population Standard Sample Standard

Deviation Deviation
Class A 65 60
Class B 45 50

Calculate the value of F-variate.

After introducing the F-distribution, one may be interested in knowing the
properties of this distribution. Now, we shall discuss some of the important
properties of F-distribution in the next section.

4.3 PROPERTIES OF F-DISTRIBUTION

The F-distribution has wide properties in Statistics. Some of them are as
follow:
1. The probability curve of F-distribution is positively skewed curve. The
curve becomes highly positive skewed when ν2 is smaller than ν1.
2. F-distribution curve extends on abscissa from 0 to .
3. F-distribution is a uni-modal distribution, that is, it has single mode.
4. The square of t-variate with ν df follows F-distribution with 1 and ν
degrees of freedom.
2
5. The mean of F-distribution with (ν1, ν2) df is for  2  2.
2  2
6. The variance of F-distribution with (ν1, ν2) df is
2 22  1   2  2 
2
for  2  4.
1   2  2    2  4 

7. If we interchange the degrees of freedom ν1 and ν2 then there exists a

very useful relation as
1
F 1 , 2 ,1 
F 2 , 1 , 

Here, we have deliberated some of the important properties of the

F-distribution and some of them have discussed such as probability curve,
mean and variance of F-distribution in Sub-sections 4.3.1 and 4.3.2
respectively.
4.3.1 Probability Curve of F-distribution
In Section 4.2, we have defined the F-variate as the ratio of two sample
variances when population variances are equal and follows F-distribution
with 1 and 2 df i.e.

72
 S12 Standard Sampling
F ~ F 1 ,2  … (6) Distributions-II
S22

where, S12 and S 22 are the sample variances of two normal populations. As you
have studied that the F-variate varies from 0 to , therefore, it is always
positive so probability curve of F-distribution wholly lies in the first
quadrant. These ν1 and ν 2 are the degrees of freedom and are called the
parameters of F-distribution. Hence, the shape of probability curve
F-distribution depends on ν1 and ν 2 . The probability curves of F-
distribution for various pairs of degrees of freedom (1 , 2 ) i.e. (5, 5), (5, 20)
and (20, 5) are shown in Fig. 4.1 given below:

 1  5 ,  2  20 
 1  5 ,  2  5 
 1  20 ,  2  5 

Fig. 4.1: Probability curves of F-distribution for (5, 5), (5, 20) and (20, 5) degrees of
freedom.
After looking on the probability curves of F-distribution drawn above, we
can observe that the probability curve of the F-distribution is uni-model
curve. And by increasing in the first degrees of freedom from
1  5 to 1  20 the mean of the distribution (shown by vertical line) does
not change but probability curve shifted from the tail to the centre of the
distribution whereas increasing in the second degrees of freedom from
 2  5 to  2  20 the mean of the distribution (shown by vertical line)
decrease and the probability curve shifted from the tail to the centre of the
distribution. One can also get an idea about the skewness of the
F-distribution, we have observed from the probability curve of the
F-distribution that it is positive skewed curve and it becomes very highly
positive skewed if ν 2 becomes small.

4.3.2 Mean and Variance of F-distribution

In previous sub-section, the probability curve of F-distribution is discussed
with some of its features. Now in this sub-section, we shall describe the
mean and variance of the F-distribution. The mean and variance of F-
distribution is derived with the help of the moment about origin. As we have
discussed in Unit 3 of MST-002 that the first order moment about origin is
known as mean and central second order moment of is known as variance of
Sampling Distributions the distribution. The rth order moment about origin of F-distribution is
obtained as

r  E F r    F r f F  dF
o

1 / 2  / 2
1
 F 1 / 2  1
  Fr  1   2  / 2 dF … (7)
   0
  
B 1 , 2  1  1 F 
2 2  2 
1  
If we take F  x  F  2 x then dF  2 dx. Also when F  0  x  0
2 1 1
and when F    x  .
Now, by putting these values in above equation (7), we get
  2   21  r 1
 /2   x 
r 
 1 /  2  1   1  
   2 
  1  2  / 2   dx
   0
1  x   1 
B 1 , 2 
 2 2 
r
 2 
 
  x  1 / 2   r 1
  1  1  x  1 / 2   r    2 / 2   r 
dx
   0
B 1 , 2 
2 2
r
 2   For a  0 & b  0 
    x a 1 
  
  1 B  1  r, 2  r   a b
dx 
    2 2  0 1  x  
B 1 , 2   
 2 2    B  a, b  
 
Thus,
1 2
r r r 
  2 2 ab 
r   2  for  2  2r  B  a, b    … (8)
 1  1 2  ab  B a
2 2
Now, if we put r = 1 in the above expression given in equation (8), we get
the value of first order moment about origin which is known as mean i.e.
1 2
  2 1 2
1
Mean  1   2 . for  2  2
 1  1 2
2 2
1 1  2
 2  1
  . 2 2 2
 1  1   2  1   2  1
 r   r  1 r  1
 
2  2  2
2
 for  2  2 … (9)
2  2
74
Similarly, if we put r = 2 in the formula of r given in equation (8) then we Standard Sampling
Distributions-II
get the value of second order moment about origin i.e.
1 2
 2  2  2 2  2
2

2    . for 2  4
 1  1 2
2 2
 1     
2
  1  1  1 2  2
  2  2  2 2
 2  for  2  4
 1  1   2  1    2  2   2  2
  
2 2  2  2
22 1  2
 for 2  4 … (10)
1 2  22  4
Now, we obtain the value of variance by putting the value of first order and
second order moments about origin in the formula given below
2
Variance  2  1 

22  1  2 22
 
1  2  2   2  4   2  2 2

22  1  2   2  2   1  2  4  
 2
1  2  2    2  4 

22  12  2 1  22  4  12  41 

 2
1  2  2    2  4 

2  22  1   2  2 
 2
for  2  4 … (11)
1   2  2    2  4 

From the above derivation of mean and variance, we can conclude that mean
of F-distribution is independent of ν1 .
Similarly, on putting r = 3 and 4 in the formula given in equation (8) one
may get 3 and 4 which are the third and fourth order moments about origin
respectively.
Now, you can try some exercises to see how much you learn about the
properties of F-distribution.
E4) Write any five properties of F-distribution.

E5) What are the mean and variance of F-distribution with 1  5 and
 2  10 degrees of freedom?

4.4 APPLICATIONS OF F-DISTRIBUTION

After discussing the main properties of F-distribution in previous section, we
are now discussing some of the important applications of F-distribution.

75
Sampling Distributions Though F-distribution has many applications but some of them are listed
here.
The F-distribution has the following applications:
1. F-distribution is used to test the hypothesis about equality of the
variances of two normal populations.
2. F-distribution is used to test the hypothesis about multiple correlation
coefficients.
3. F-distribution is used to test the hypothesis about correlation ratio.
4. F-distribution is used to test the equality of means of k-populations, when
one characteristic of the population is considered i.e. F-distribution is
used in one-way analysis of variance.
5. F-distribution is used to test the equality of k-population means for two
characteristics at a time i.e. F-distribution is used in two-way analysis of
variance.
The first application listed above will be discussed in Unit 12 of this course,
fourth and fifth applications will be discussed in Block-2 of MST-005,
second application will be discussed in detail in specialisation courses and
the remaining third application is beyond the scope of this course.
Now, try the following exercise.
E6) Write four applications of F-distribution.

4.5 RELATION BETWEEN t, χ2 AND

F-DISTRIBUTIONS
After describing the exact sampling distributions χ2, t & F, we shall explain
relationship between t & F-distributions and χ2 & F-distributions. These
relations can be described in terms of the following two statements:
Statement 1: If a variate t follows Student’s t-distribution with ν df, then
square of t follows F-distribution with (1, ν) df i.e. if t ~ t(ν) then t2 ~F(1, ν).
Statement 2 : If F-variate follows F-distribution with (ν1, ν2) df and if
ν 2   then  2  1F follows a χ2-distribution with ν1 df.
The proofs of these statements are beyond the scope of this course.
Now, try to answer the following exercises.
E7) If a variate t follows Student’s t-distribution with 4 df, then what is
the distribution of square of t. Also write the pdf of that distribution.

4.6 TABULATED VALUES OF t, χ2 AND F

As we have seen in Unit14 of MST-003 that the area or probability of normal
distribution can be calculated with the help of tables. Similarly, we can also
find the area or probability of a variate which follows either of t, χ2 and
F-distributions. In this section, we shall discuss the method of obtaining the
tabulated values of t, χ2 and F-variates.

76
4.6.1 Tabulated Values of t-Variate Standard Sampling
Distributions-II
A variate which follows the t-distribution is described with the help of
degrees of freedom (ν). Table-II in the Appendix contains the tabulated
(critical) values of t-variate for different degrees of freedom (ν) such that the For level of significance,
area under the probability curve of t-distribution to its right tail (upper tail) is please see Unit 9 of this
equal to α as shown in Fig. 4.2. The critical value of t-variate is represented course.
by t ( ν), α where, ν represents the df and α area under the right tail or level of
significance.
In this table, the first column of left hand side contains values of degrees of
freedom (ν) and the column heading represents the values of α on the right
hand side (tail) of the probability curve of the t-distribution and the entries in
the body of the table are the values of t-variate for each particular value of ν
and α. We discuss the process of finding the values of t-variate by taking an
example as: Fig. 4.2
If we want to find out the value of t-variate for which the area on the right
tail is 0.05 and the degrees of freedom is 8. Then we start with the first
column of the t-table and downward headed ‘ν’ until entry 8 is reached. Then
proceed right to the column headed α = 0.05. For your convenient, part of the
t-table is shown in Table 4.1 given below:
Table 4.1: t-table for Tabulated (Critical) Values of t-variate
One- tail α = 0.10 0.05 0.025 0.01 0.005
ν =1 3.078 6.314 12.706 31.821 63.657
2 1.886 2.920 4.303 6.965 9.925
3 1.638 2.353 3.182 4.541 5.841
4 1.533 2.132 2.776 3.747 4.604
5 1.476 2.015 2.571 3.365 4.032
6 1.440 1.943 2.447 3.143 3.707
7 1.415 1.895 2.365 2.998 3.499
8 1.397 1.860 2.306 2.896 3.355

Thus, we get the required value of t-variate as t(ν) = t(ν), α = t(8), 0.05 = 1.860.
The value of t-variate equal to 1.860 means, the probability that t-variate
would exceed 1.860 is 0.05 as shown in Fig. 4.3.
Since t-distribution is symmetrical about t = 0 line, therefore, it is sufficient Fig. 4.3
to tabulate the values of t-variate for right tail only. Therefore, the value of
t-variate corresponding to a left hand tail for α is − t(ν), α as shown in Fig.
4.4.Thus, in above example, if we want to find the value of t-variate such
that left area is 0.05 for 8 df then due to symmetry of t-distribution the value
of t-variate is −1.860.
If we want to find the values of t-variate when the total area on both tails is α
Fig. 4.4
then by symmetry the half area i.e. α/2 lies in both tails as shown in Fig. 4.5.
In such situation, we find the value of t-variate corresponding to α/2 area and
there are two values of t-variate as ± t(ν), α/2. For example, if we want to find
out the values of t-variate for which the area on the both tails is 0.05 and the
degrees of freedom is 8. Then by symmetry, we find the value of variate t
corresponding to α/2 = 0.05/2 = 0.025 and 8 df. Therefore, from the t-table
we get t(ν), α/2 = t(8), 0.025 = 2.571. So required values of t-variate are ± t(ν), α/2 = Fig. 4.5
± t(8), 0.025 = ± 2.571.
Example 2: Find the value of t-variate in each case for which the area on the
right tail (α) and the degrees of freedom (ν) are given in next page:
Sampling Distributions (i) α = 0.01 and ν = 12

(ii) α = 0.10 and ν = 15

Solution:
(i) Here, we want to find the value of t-variate for
α = 0.01 and ν = 12
Therefore, we start with the first column of t-table given in the
Appendix and downward headed ν until entry 12 is reached. Then
proceed right to the column headed α = 0.01. So we get the required
value of t-variate as t(ν), α = t(12), 0.01 = 2.681.
(ii) Here, we are given that
α = 0.10 and ν = 15
By proceeding same way as above, from t-table, we get the required
value of t-variate as t(ν), α = t(15), 0.10 = 1.341. For your convenient, part
of the t-table is shown in Table 4.2 given below:
Table 4.2: t-Table for Tabulated (Critical) Values of t-variate
One- tail α = 0.10 0.05 0.025 0.01 0.005
ν = 10 1.372 1.812 2.228 2.764 3.169
11 1.363 1.796 2.201 2.718 3.106
12 1.356 1.782 2.179 2.681 3.055
13 1.350 1.771 2.160 2.650 3.012
14 1.345 1.761 2.145 2.624 2.977
15 1.341 1.753 2.131 2.602 2.947
16 1.337 1.746 2.120 2.583 2.921

Example 3: Find the values of t-variate in each part for which the area on the
both tails and the degrees of freedom are
(i) α = 0.01 and ν = 12
(ii) α = 0.10 and ν = 29
Solution:
(i) Here, we want to find the value of t-variate for two-tails and for
α = 0.01 and ν = 12
Since total area on the both tails is 0.01 therefore, by symmetry the area
on the right tail is 0.01/2 = 0.005. Thus, we start with the first column
of t-table and downward headed ν until entry 12 is reached. Then we
proceed right to the column headed α = 0.005. We get t(ν), α/2 = t(12), 0.005
= 3.055. So the required values of t-variate are ± t(ν), α/2 = ± t(12), 0.005 = ±
3.055.
(ii) Here, we are given that
α = 0.10 and ν = 29
By proceeding same way as above, from the t-table, we get the required
values of t-variate as ± t(ν), α/2 = ± t(29), 0.05 = ± 1.699.
Method of Finding the Values of t-variate for Degrees of Freedom which
are not listed in the Table

78
The t-table given in the Appendix does not list values for every possible Standard Sampling
degree of freedom. Therefore, it becomes necessary to know how to find Distributions-II
values of t-variate for degrees of freedom not listed in the table.
We discuss the process of finding the values which are not listed in the table
with the help of an example as:
Suppose we want to find out the value of t-variate for 34 degrees of freedom
such that the area on the right side is equal to 0.05. The t-table in the
Appendix does not list a tabulated value for this degree of freedom, so we
need to interpolate it.
First of all, we read the tabulated values of t-variate that are just greater and
just smaller than the degrees of freedom for our interest. Thus, from the
t-table, we get the values of t-variate for 40 and 30 df and area α = 0.05 as
t(40), 0.05 = 1.684 t(30), 0.05 = 1.697
Note that larger the degrees of freedom smaller the tabulated value of
t-variate.
Now, we calculate how much the t-value changes for each degree of freedom
between these two tabulated values. Here, there is a difference of 10(40 – 30)
degrees of freedom and t-value change of 0.013(1.697 – 1.684).
Thus, to determine how much it changes for each degree of freedom, divide
the difference between the t-values by the difference between the degrees of
freedom as
0.013
 0.0013
10
Since, we have to obtain the value for 34 degrees of freedom, this is either 4
more than 30 or 6 less than 40. Therefore, we can interpolate from either
values. To get from 30 to 34 degrees of freedom there is a difference of
4 (34 – 30). So multiply this difference by the amount by which the t-value
change per degree of freedom i.e. 0.0013. This result as
4  0.0013  0.0052
Since larger the degrees of freedom smaller the tabulated value of t-variate
therefore, subtracting this value 0.0052 from t(30, 0.05) = 1.697 to get the
required value. Thus,
t(34), 0.05 = 1.697 − 0.0052 = 1.6918
Now, if we interpolate from 40 degrees of freedom then the difference
6(40 – 34) multiply by 0.0013 and adding this to 1.684, we get
t (34), 0.05 = 1.684 +6  0.0013 = 1.6918
Thus, we get the same value.
Example 4: Find the value of t-variate in each part for which the area on the
right tail (α) and the degrees of freedom (ν) are given below:
(i) α = 0.01 and ν = 42 (ii) α = 0.01 and ν = 35
Solution:
(i) Here, we want to find the value of t-variate for
α = 0.01 and ν = 42
Since t-table does not have the tabulated value for 42 degrees of
freedom so we need to interpolate it. For this, we find the tabulated
79
Sampling Distributions values of t-variate that are just greater and just less than the degree of
freedom 42. Thus, we get
t(60), 0.01 = 2.390 t(40), 0.01 = 2.423
There is a difference of 20 degrees of freedom between these two and a
difference 0.033(2.423-2.390) in the t-values.
Thus, each degree of freedom has an approximate change in the value
of t-variate is
0.033
 0.00165
20
To get 42 degrees of freedom, multiply 0.000165 by 2 (42−40), we get
0.00165  2  0.003
Since larger the degrees of freedom smaller the tabulated value of t-
variate, therefore, subtracting this from 2.423, we get required values as
t(42), 0.01 = 2.423 – 0.003 = 2.420
(ii) Similarly, we find the tabulated values of t-variate that are just greater
and just less than the degree of freedom 35. Thus, we have
t(40), 0.01 = 2.423 t(30), 0.01 = 2.457
There is a difference of 10 degrees of freedom between these two and a
difference 0.034(2.457−2.423) in the t-values.
Thus, each degree of freedom has an approximate change in the value
of t-variate is
0.034
 0.0034
10
To get 35 degrees of freedom, multiplying 0.0034 by 5 (35−30), we
have
0.0034  5  0.017
Since larger the degrees of freedom smaller the tabulated value of
t-variate, therefore, subtracting this from 2.457, we get required value
as
t(35), 0.01 = 2.457 – 0.017 = 2.440
Now, you can try the following exercise.
E8) Find the values of t-variate for which the area and the degrees of
freedom are
(i) α = 0.01 (one-right tail) and ν = 11 (ii) α = 0.05 (one-left tail)
and ν = 16
(iii) α = 0.05 (two-tails) and ν = 19 (iv) α = 0.05 (two-tails) and
ν = 45

4.6.2 Tabulated Value of χ2-Variate

As similar to t-table, the Table-III in the Appendix contains the tabulated
(critical) values of chi-square variate for different degrees of freedom (ν)
such that the area under the probability curve of χ2-distribution to its right
Fig. 4.6 (upper) tail is equal to α as shown in Fig. 4.6.
In this table, the column headings indicates the area on the upper portion Standard Sampling
(right tail) of the probability curve of chi-square distribution and the first Distributions-II
column on the left hand side indicates the values of degrees of freedom (ν).
Also the entry in the body of the table indicates the value of chi-square
variate. Therefore, (2 ), is the value such that the area under the probability
curve of chi-square distribution with ν df to its right tail is equal to α i.e. the
probability that any value of χ2-variate is greater than or equal to (2 ), is α.
We discuss the process of finding the values of χ2-variate by taking an
example as:
If we want to find out the value of chi-square variate with 7 df for which the
area on the right tail is 0.01 then we start with first column i.e. degrees of
freedom (ν) and proceed downward until 7 is reached then proceed right to
the column headed α = 0.01 which gives the required value of χ2 as shown in
the Table 4.3 given below and is also indicate in Fig. 4.7 as
2  2 ,   27,0.01  18.48. Fig. 4.7
Table 4.3: Chi-square Table

α = 0.995 0.99 0.975 0.95 0.90 0.10 0.05 0.025 0.01 0.005
ν =1 --- --- --- --- 0.02 2.71 3.84 5.02 6.63 7.88
2 0.01 0.02 0.05 0.10 0.21 4.61 5.99 7.38 9.21 10.60
3 0.07 0.11 0.22 0.35 0.58 6.25 7.81 9.35 11.34 12.84
4 0.21 0.30 0.48 0.71 1.06 7.78 9.49 11.14 13.28 14.86
5 0.41 0.55 0.83 1.15 1.61 9.24 11.07 12.83 15.09 16.75
6 0.68 0.87 1.24 1.64 2.20 10.64 12.59 14.45 16.81 18.55
7 0.99 1.24 1.69 2.17 2.83 12.02 14.07 16.01 18.48 20.28

The chi-square distribution is a skewed (not symmetrical) distribution.

Therefore, it is necessary to tabulate values of χ2 for  > 0.50 also.
Note 1: Values of chi-square variate for degrees of freedom which are not
listed in the table are obtained in the similar manner as discussed in case of
t-variate. This is explained in part (iv) of the Example 5 given below.
Now, let us do one example based on the above discussion.
Example 5: Find the values of chi-square variate for which the area and
degrees of freedom are
(i) α = 0.05(one-tail) and ν = 2
(ii) α = 0.99(one-tail) and ν = 15
(iii) α = 0.05(two-tails) and ν = 6
(iv) α = 0.01(one-tail) and ν = 64
Solution:
(i) Here, we want to find the value of χ2-variate for one-tail and for
α = 0.05 and ν = 2
Thus, we first start with the first column of χ2-table given in the
Appendix and downward headed ν until entry 2 is reached. Then
proceed right to the column headed α = 0.05. We get the required value
of χ2-variate as 2( ),  2(2),0.05  5.99.

(ii) Here, we want to find the value of χ2-variate for one-tail and for
Sampling Distributions α = 0.99 and ν = 15
By proceeding same way as above, from the χ2-table, we get the
required value of χ2-variate as 2( ),  (15),0.99
2
 5.23.
(iii) Here, we want to find the value of χ2-variate for two-tails and for
α = 0.05 and ν = 6
Since total area on the both tails is 0.05 therefore, half area lies on both
tails as 0.05/2 = 0.025. In this case, χ2-variate has two values one on
right-tail as 2 / 2  2 ,  / 2  26,0.025 and one on left-tail as
2(1 / 2)  2 ,(1 / 2)  26,0.975. So by proceeding same way as above we
get required values of χ2-variate as 26,0.025  14.45 and 26,0.975  1.24 as
shown in Fig.4.8.
Fig. 4.8 (iv) Here, we want to find the value of χ2-variate for one-tail and for
α = 0.01 and ν = 64
2
Since χ -table does not have the tabulated value for 64 degrees of
freedom so we need to interpolate it. For this we find the tabulated
values of χ2-variate that are just greater and just less than the degree of
freedom 64 with α = 0.01. Thus, we have
270, 0.01  100.42 260, 0.01  88.38

There is a difference of 10 degrees of freedom between these two and a

difference 12.04(100.42−88.38) in the χ2-value. Also note that larger
the degrees of freedom larger the tabulated value of χ2-variate.
Thus, each degree of freedom has an approximate change in the value
of χ2-variate as
12.04
 1.204
10
To get the value of χ2-variate for 64 degrees of freedom, multiplying
1.204 by 4 (64−60), we get
1.204  4  4.816
Since larger the degrees of freedom larger the tabulated value of
χ2-variate so adding 4.816 in 88.38, we get required value as
264, 0.01  88.38  4.816  93.196

Now, you can try the following exercise.

E9) Find the value of χ2-variate for which the area on the right tail and the
degrees of freedom are
(i) α = 0.01 and ν = 11
(ii) α = 0.90 and ν = 19
(iii) α = 0.05 and ν = 44

4.6.3 Tabulated Value of F-Variate

F-distribution is described with the help of two degrees of freedom, one for Standard Sampling
numerator and other for denominator. For each combination of degrees of Distributions-II
freedom, F-variate has different tabulated value. Table-IV in the Appendix
contains the tabulated (critical) values of F-variate for various degrees of
freedom such that the area under the probability curve of F-distribution to its
right (upper) tail is equal to α = 0.10, 0.05, 0.025 and 0.01. A general case is
shown in Fig. 4.9.
The first row of F-table indicates the values of degrees of freedom ( 1 ) for
numerator and first column on the left hand side indicates the values of
degrees of freedom ( 2 ) for denominator. We discuss the process of finding
the tabulated values of F-variate by taking an example as:
Suppose we want to find the value of F-variate for which the area on the
right tail is 0.05 and the degrees of freedom for numerator and denominator
are ν1= 10 and ν2 = 15 respectively. Since α = 0.05 so we select the F-table
for α = 0.05 and start with the first column of this table and proceed down to
this column headed 2 until entry 15 is reached then proceed right to the
Fig. 4.9
column headed 1  10 as shown in the Table 4.4. In this way, we get the
required value of F as F( 1 , 2 ),   F(10,15),0.05  2.54 which is also shown in Fig.
4.10.
Table 4.4: F-Table for α = 0.05
Degrees of
freedom Degrees of freedom for numerator(ν 1)
for
numerator 1 2 3 4 5 6 7 8 9 10 11 12
(ν2 )
2.98
Fig. 4.10
10 4.96 4.10 3.71 3.48 3.33 3.22 3.14 3.07 3.02 2.94 2.91
11 4.84 3.98 3.59 3.36 3.20 3.09 3.01 2.95 2.90 2.85 2.82 2.79
12 4.75 3.89 3.49 3.26 3.11 3.00 2.91 2.85 2.80 2.75 2.72 2.69
13 4.67 3.81 3.41 3.18 3.03 2.92 2.83 2.77 2.71 2.67 2.63 2.60
14 4.60 3.74 3.34 3.11 2.96 2.85 2.76 2.70 2.65 2.60 2.57 2.53
15 4.54 3.68 3.29 3.06 2.90 2.79 2.71 2.64 2.59 2.54 2.51 2.48
16 4.49 3.63 3.24 3.01 2.85 2.74 2.66 2.59 2.54 2.49 2.46 2.42

Note 2: The value of F-variate for ν1 degrees of freedom for numerator and
ν2 for denominator is not same for ν2 degrees of freedom for numerator and ν1
for denominator, that is,
F 1 , 2   F 2 , 1 

The F-distribution is also a skewed (not symmetrical) distribution. Therefore,

it becomes necessary to tabulate values of F-distribution for  > 0.50 also.
The tabulated values of F-distribution for  > 0.50 can be calculated by the
relation as
1
F 1 , 2 , 1   [Recall property listed in Section 4.3]
F 2 , 1 , 

Note 3: Values of F-variate for degrees of freedom which are not listed in
the table are obtained in the similar manner as discussed in case of t-variate.
This is explained in part (iv) of the Example 6 given below.
Now, let us do one example based on the above discussion.
Sampling Distributions Example 6: Find the values of F-variate in each part for which the area (α )
and the degrees of freedom(ν1, ν2) are
(i) α = 0.01(one-tail) and (ν1 = 5, ν2 = 8)
(ii) α = 0.95(one-tail) and (ν1 = 12, ν2 = 16)
(iii) α = 0.05 (two-tails)and (ν1 = 4, ν2 = 10)
(iv) α = 0.01(one-tail) and (ν1 = 10, ν2 = 32)
Solution:
(i) Here, we want to find the value of F-variate for
α = 0.01 and (ν1 = 5, ν2 = 8)
Thus, we select the F-table for α = 0.01 and start with the first column
of this table and downward headed ν2 until entry 8 is reached. Then
proceed right to the column headed ν1 = 5. We get the required value of
F-variate as 6.63.
(ii) Here, we want to find the value of F-variate for
α = 0.95 and (ν1 = 12, ν2 = 16)
Since we have F-tables for α = 0.10, 0.05, 0.025 and 0.01only
therefore, for α = 0.95, we can use the relation
1
F 1 ,2 ,(1) 
F 2 , 1 , 

From the F-table for α = 0.05, we have

F 2 , 1 ,   F16,12,0.05  2.60
Thus,
1 1
F 1 , 2 ,(1 )  F12,16 ,0.95    0.3846
F16,12 ,0.05 2.60

(iii) Here, we want to find the value of F-variate for two-tails and for
α = 0.05 and ν1 = 4, ν2 = 10
Since total area on the both tails is 0.05 therefore, half area lies on both
tails as 0.05/2 = 0.025. In this case, F-variate has two values one on
right-tail as F / 2  F 1 , 2 ,  / 2  F 4,10,0.025 and one on left-tail as
F(1 / 2)  F 1 , 2 ,(1 / 2)  F4,10,0.975. So by proceeding same way as above
we get required values of F-variate as F4,10,0.025  4.47 and
1 1
Fig.4.11 F 4,10,0.975    0.11 as shown in Fig.4.11.
F10,4 ,0.025 8.84

(iv) Here, we want to find the value of F-variate for

α = 0.01 and (ν1 = 10, ν2 = 32)
Since F-table for α = 0.01 does not have the tabulated value for these
degrees of freedom so we need to interpolate it. For this we find the
tabulated values of F that are just greater and just less than the degree
of freedom 32. Thus we have
F(10, 40) , 0.01 = 2.80 F(10, 30) , 0.01 = 2.98
 There is a difference of 10 degrees of freedom between these two and a Standard Sampling
difference 0.18(2.98-2.80) in the F-values. Also note that larger the Distributions-II
degrees of freedom smaller the tabulated value of F-variate.
Thus, each degree of freedom has an approximate change in the value
of F as
0.18
 0.018
10
To get 32 degrees of freedom, multiplying 0.018 by 2 (32-30), we get
0.018  2  0.036
Since larger the degrees of freedom smaller the tabulated value of F-
variate, therefore, subtracting this from 2.98, we get required values as
F(10, 32) = 2.98 – 0.036 = 2.944
Now, try to answer the following exercise.
E10) Find the value of F-variate in each part for which the area on the right
side (α) and the degrees of freedom (ν1, ν2) are
(i) α = 0.05 and (ν1 = 8, ν2 = 12)
(ii) α = 0.99 and (ν1 = 5, ν2 = 10)
(iii) α = 0.05 and (ν1 = 23, ν2 = 5)
Let us end with a brief look at what we have covered in this unit.

4.7 SUMMARY
In this unit, we have covered the following points:
1. The F-distribution.
2. The properties of F-distribution.
3. The probability curve of F-distribution.
4. Mean and variance of F-distribution.
5. The applications of F-distribution.
6. The relation between t, F and χ2-distributions.
7. The method of obtaining the tabulated value of a variate which follows
either of t, χ2 and F-distributions.

4.8 SOLUTIONS / ANSWERS

E1) The probability density function of F-distribution with ν1 and ν 2 df is
given as
ν1 / 2
 ν1/ν 2  F ν1 / 2  1
f(F)   ν1  ν2  / 2
; 0F
ν ν 
B  1 , 2   1  ν1 F 
 2 2  ν 2 

(i) Therefore, for 1  4 and 2  2 df the pdf of F-distribution is given
by
85
Sampling Distributions 4/2
F
4 / 2  1
f(F) 
 4/2 
 4 2  / 2
4 2
B  ,   1  4 F 
2 2  2 

4 2 1 F  a b 
 3  B  a, b   
2 1 1  2F   ab 

8F  3  2  2 
 3
; 0F  
1  2F   & 2  1  1 
(ii) Similarly, for 1  2 and 2  4 df the pdf of F-distribution is given
by
2/2
F
2 / 2  1
f(F) 
 2/4 
 2 4  / 2
2 4
B  ,   1  2 F 
2 2  4 

2 1 1  a b 
 3  B  a, b   
4 21  F  ab 
1  
 2

1  3  2  2
 3
; 0 F  
 F & 2  1  1 
2 1  
 2
Hence, from (i) and (ii) parts, we observed that
F 1 ,  2   F 2 , 1 

E2) If random variable X follows the F-distribution with ν1 and ν 2 df then

probability density function of X is given as
ν1 / 2
 ν1/ν 2  x  ν1 / 2  1
f(x)   ν1  ν2  / 2
; 0  x   … (12)
ν ν 
B  1 , 2   1  ν1 x 
 2 2  ν 2 

Here, we are given that
1
f x   2
; 0 x 
1  x 
This can be written as
2/ 2
x
2 / 2  1
f x  
 2/2  ; 0  x    B 1,1  1
 2 2  / 2
2 2
B  ,   1  2 x 
2 2  2 
Comparing with equation (12), we get 1  2 and 2  2 .
E3) Here, we are given that
n1  15, n2  10, 1  65, 2  45,S1  60, S2  50
86
 The value of F-variate can be calculated as follows: Standard Sampling
Distributions-II
S12 /σ12
F
S22 /σ 22
where, S12 &S22 are the values of sample variances.
Therefore, we have
2 2
 60  /  65
F 2 2
 50 /  45
0.85
  0.69
1.23
E4) Refer Section 4.3.
E5) We know that the mean and variance of F-distribution with 1 and 2
degrees of freedom are
2
Mean  for  2  2
2  2
and
2 22  1   2  2 
Variance  2
for  2  4.
1   2  2    2  4 

In our case, 1  5 and 2  10 , therefore,

2 10 5
Mean   
 2  2 10  2 4
2
2 10   5  10  2  65
Variance  2

5 10  2  10  4  48
E6) Refer Section 4.4.
E7) We know that if a variate t follows Student’s t-distribution with ν df,
then square of t follows F-distribution with (1, ν) df i.e. if t ~ t(ν) then
t2 ~F(1, ν). In our case, ν = 4, therefore square of t follows
F-distribution with (1, 4) df whose pdf is given by
ν1 / 2
 ν1/ν 2  F ν1 / 2  1
f(F)   ν1  ν2  / 2
; 0F
ν ν 
B  1 , 2   1  ν1 F 
 2 2  ν 2 

Therefore, for 1  1 and 2  4 df the pdf of F-distribution is given by
2/2
F
1/ 2  1
f(F) 
1/4 
1 4  / 2
1 4
B  ,   1  1 F 
2 2  4 

87
Sampling Distributions 1
2
1
 2 5/ 2
1 F
4 2 F  1  
2  4

31 1
1
 22 2 5/ 2
 2  1
 
1  F
4 F 1  
2  4
3 1
 5/2
; 0 F
16  F
F 1  
 4
E8)
(i) Here, we want to find the value of t-variate for one-right tail
corresponding
α = 0.01 and ν = 11
Therefore, we start with the first column of t-table given in the
Appendix and downward headed ν until entry 11 is reached. Then
proceed right to the column headed α = 0.01. We get the required
value of t-variate as t(ν), α = t(11), 0.01 = 2.718.
(ii) Here, we want to find the value of t-variate for one-left tail
corresponding
α = 0.05 and ν = 16
By proceeding same way as above, from the t-table, we get the
value of t-variate as 1.746. So the required values of t-variate for
one left-tail is − t(ν), α = − t(16), 0.05 = − 1.746.
(iii) Here, we want to find the values of t-variate for two-tails
corresponding
α = 0.05 and ν = 19
Since total area on the both tails is 0.05 therefore, by symmetry,
the area on the right tail is 0.05/2 = 0.025. Thus, we first start
with the first column of t-table and downward headed ν until
entry 19 is reached. Then proceed right to the column headed
0.025. We get t(ν), α/2 = t(19), 0.025 =2.093. So the required values of
t-variate are ± t(ν), α/2 = ± t(19), 0.025 = ± 2.093.
(iv) Here, we want to find the values of t-variate for two-tails
corresponding
α = 0.05 and ν = 45
Since total area on the both tails is 0.05 therefore, by symmetry
the area on the right tail is 0.05/2 = 0.025 and also t-table does not
have the tabulated value for 45 degrees of freedom with
α = 0.025, so we need to interpolate it. For this, we find the
tabulated values of t-variate that are just greater and just less than
the degrees of freedom 45 at α = 0.025. Thus, we get
t(60) , 0.025 = 2.000 t(40) , 0.025 = 2.021

88
 There is a difference of 20 degrees of freedom between these two Standard Sampling
and a difference 0.021(2.021−2.000) in the t-values. Distributions-II

Thus, each degree of freedom has an approximate change in the

value of t as
0.021
 0.00105
20
To get 45 degrees of freedom, multiply 0.00105 by 5 (45−40), we
get
0.00105  5  0.00525
Now subtracting this from 2.021, we get t-value as
t(45) , 0.025 = 2.021 – 0.00525 = 2.01575
So the required values of t-variate are ± t(ν), α/2 = ± t(45), 0.025 =
± 2.01575.
E9) (i) Here, we want to find the value of χ2-variate for
α = 0.01 and ν = 11
Thus, we start with the first column of χ2-table given in the
Appendix and downward headed ν until entry 11 is reached. Then
proceed right to the column headed α = 0.01. We get the required
value of χ2-variate as 24.72.
(ii) Here, we want to find the value of χ2-variate for
α = 0.90 and ν = 19
By proceeding same way as above, from the χ2-table, we get the
required value of χ2 as 11.65.
(iii) Here, we want to find the value of χ2-variate for
α = 0.05 and ν = 44
2
Since χ -table does not have the tabulated value for 44 degrees of
freedom so we need to interpolate it. For this we find the
tabulated values of χ2-variate that are just greater and just less
than the degrees of freedom 44 with α = 0.05. Thus, we have
250, 0.05  67.50 240, 0.05  55.76

There is a difference of 10 degrees of freedom between these two

and a difference 11.74(67.50 – 55.76) in the χ2-values. Also note
that larger the degrees of freedom larger the tabulated value of χ2.
Thus, each degree of freedom has an approximate change in the
value of χ2 as
11.74
 1.174
10
To get 44 degrees of freedom, multiplying 1.174 by 4 (44−40),
we get
1.174  4  4.696
Now adding this in 55.76, we get required value as

89
Sampling Distributions 244,0.05  55.76  4.696  60.456

E10) (i) Here, we want to find the value of F-variate for

α = 0.05 and (ν1 = 8, ν2 = 12)
Thus, we select the F-table for α = 0.05 and start with the first
column of this table and downward headed ν2 until entry 12 is
reached. Then proceed right to the column headed ν1 = 8. We get
the required value of F as 2.85.
(ii) Here, we are given that
α = 0.99 and (ν1 = 5, ν2 = 10)
Since we have F-tables for α = 0.10, 0.05, 0.025 and 0.01only
therefore, for α = 0.99, we can use the relation
1
F 1 , 2 ,(1) 
F 2 ,1 , 

From the F-table for α = 0.01, we have

F10,5,0.01  4.47

Thus,
1 1
F 1 , 2 ,(1 )  F5,10, 0.99    0.10
F10,5, 0.01 10.05

(iii) Here, we want to find the value of F-variate for

α = 0.05 and (ν1 = 23, ν2 = 5)
Since F-table for α = 0.01 does not have the tabulated value for
these degrees of freedom so we need to interpolate it. For this we
find the tabulated values of F that are just greater and just less
than the degrees of freedom 23. Thus we have
F(24, 5) , 0.05 = 4.53 F(20, 5) , 0.05 = 4.56
Here, there is a difference of 4(24–20) degrees of freedom and F
value changes by 0.03(4.56 – 4. 53). Thus, to determine how
much it approximately changes for each degree of freedom divide
the difference between the F values by the difference between the
degrees of freedom, we get
0.03
 0.0075
4
Since, we have 23 degrees of freedom, this is either 3 more than
20 for 1 less than 24. Therefore, we can interpolate from either
values. To get from 20 to 23 degrees of freedom there is a
difference of 3(23-20). So multiplying this difference by the
amount by which the F-value approximate change per degrees of
freedom i.e. 0.0075. These results in
3  0.0075  0.0225
Since larger the degrees of freedom smaller the tabulated value of
F-variate, therefore, subtracting this value 0.0225 from 4.56 to get
required value. Thus,
90
F(23, 5), 0.05 = 4.56 − 0.0225 = 4.5375 Standard Sampling
Distributions-II

91
 MST-004
STATISTICAL
Indira Gandhi
National Open University INFERENCE
School of Sciences

Block

2
ESTIMATION
UNIT 5
Introduction to Estimation 5
UNIT 6
Point Estimation 37
UNIT 7
Interval Estimation for One Population 55
UNIT 8
Interval Estimation for Two Populations 85
Curriculum and Course Design Committee
Prof. K. R. Srivathasan Prof. Rahul Roy
Pro-Vice Chancellor Math. and Stat. Unit
IGNOU, New Delhi Indian Statistical Institute, New Delhi

Prof. Parvin Sinclair Dr. Diwakar Shukla

Pro-Vice Chancellor Department of Mathematics and Statistics
IGNOU, New Delhi Dr. Hari Singh Gaur University, Sagar

Prof. Geeta Kaicker Prof. Rakesh Srivastava

Director, School of Sciences Department of Statistics
IGNOU, New Delhi M.S. University of Baroda, Vadodara

Prof. Jagdish Prasad Prof. G. N. Singh

Department of Statistics Department of Applied Mathematics
University of Rajasthan, Jaipur I.S.M., Dhanbad

Prof. R. M. Pandey Dr. Gulshan Lal Taneja

Department of Bio-Statistics Department of Mathematics
All India Institute of Medical Sciences M.D. University, Rohtak
New Delhi
Faculty members of School of Sciences, IGNOU
Statistics Mathematics
Dr. Neha Garg Dr. Deepika
Dr. Nitin Gupta Prof. Poornima Mital
Mr. Rajesh Kaliraman Prof. Sujatha Varma
Dr. Manish Trivedi Dr. S. Venkataraman

Block Preparation Team

Dr. Ramkishan (Editor) Mr. Prabhat Kumar Sangal
Department of Statistics School of Sciences, IGNOU
D. A. V. (PG) College
C.C. S. University, Merrut

Dr. Parmod Kumar (Language Editor)

School of Humanities, IGNOU

Course Coordinator: Mr. Prabhat Kumar Sangal

Programme Coordinator: Dr. Manish Trivedi

Block Production
Mr. Sunil Kumar, AR (P),School of Sciences, IGNOU
CRC prepared by Mr. Prabhat Kumar Sangal, School of Sciences, IGNOU

Acknowledgement: I gratefully acknowledge my colleagues Mr. Rajesh Kaliraman and Dr. Neha
Garg, Statistics Discipline, School of Sciences for their great support.

July, 2013
© Indira Gandhi National Open University, 2013
ISBN-978-81-266-
All rights reserved. No part of this work may be reproduced in any form, by mimeograph or any other
means, without permission in writing from the Indira Gandhi National Open University.

Further information on the Indira Gandhi National Open University may be obtained from University’s
Office at Maidan Garhi, New Delhi-110068 or visit University’s website http://www.ignou.ac.in

Printed and published on behalf of the Indira Gandhi National Open University, New Delhi by the
Director, School of Sciences.

Printed at:
ESTIMATION
In Block 1 of this course, you have studied the sampling distributions of
different statistics as sample mean, sample proportion, sample variance, etc.
and standard sampling distributions as χ2, t, F and Z which provide a platform
to the learners how to draw the inference about the population parameter(s) on
the basis of the sample(s).

In present block, we shall be studying the estimation theory, through which we

estimate the unknown parameter on the basis of sample data. Two types of
estimation i.e. point estimation and interval estimation are discussed in this
block. This block comprises four units.
Unit 5: Introduction to Estimation
Estimation admits two problems; the first is to select some criteria or properties
such that if an estimator possesses these properties it is said to be the best
estimator among all possible estimators and the second is to derive some
methods or techniques through which we obtain an estimator which possesses
such properties. This unit is devoted to explain the criteria of good estimator.
This unit also explains different properties of good estimator such as
unbiasedness, consistency, efficiency and sufficiency with different examples.
Unit 6: Point Estimation
This unit explores the basic concepts of point estimation. In point estimation,
we determine a single statistic whose value is used to estimate the value of
unknown parameter. In this unit, we shall discuss some frequently used
methods of finding point estimate such as method of maximum likelihood,
method of moments and method of least squares.
Unit 7: Interval Estimation for One Population
Instead of estimating the population parameter by a single value, an interval is
used for estimating the population parameter within which we can be
reasonably sure that the true value of parameter will lie. This technique is
known as interval estimation. In this unit, we shall discuss the method of
obtaining the interval estimates of population mean, population proportion and
population variance of normal population. Also we shall explorer the interval
estimation for population parameters of non-normal populations.
Unit 8: Interval Estimation for Two Populations
This unit is devoted to describe the method of obtaining the confidence interval
for difference of population means, difference of population proportions and
ratio of population variances of two normal populations.
Notations and Symbols
X1, X2, …, Xn : Random sample
x1, x2, …, xn : Observed value of random sample
Θ : Parameter space and read as big theta
f(x,θ) : Probability density (mass) function
f  x1 , x1 ,..., x n ,  : Joint probability density (mass) function of sample
values
L(θ) : Likelihood function of parameter θ
T  t(X1 , X1,..., X n ) : Estimator
E(T) : Expectation of T
Var(T) : Variance of T
SE : Standard error
e : Efficiency
*
T : Most efficient estimator
MVUE : Minimum variance unbiased estimator
̂ : Estimate of θ

: Partial derivative with respect to θ

X(1), X(2),…, X(n) : Ordered statistic
ML : Maximum Likelihood
Mr : rth sample moment about origin

Mr : rth sample moment about mean

r : rth population moment about origin

r : rth population moment about mean

1−  : Confidence coefficient or Confidence level
Q  q(X 1 , X1,..., X n , ) : Pivotal quantity
E : Sampling error or margin of error
UNIT 5 INTRODUCTION TO ESTIMATION
Structure
5.1 Introduction
Objectives
5.2 Basic Terminology
5.3 Characteristics of Estimators
5.4 Unbiasedness
5.5 Consistency
5.6 Efficiency
Most Efficient Estimator
Minimum Variance Unbiased Estimator
5.7 Sufficiency
5.8 Summary
5.9 Solutions /Answers

5.1 INTRODUCTION
In many real-life problems, the population parameter(s) is (are) unknown and
someone is interested to obtain the value(s) of parameter(s). But, if the whole
population is too large to study or the units of the population are destructive in
nature or there is a limited resources and manpower available then it is not
practically convenient to examine each and every unit of the population to find
the value(s) of parameter(s). In such situations, one can draw sample from the
population under study and utilize sample observations to estimate the
parameter(s).
Every one of us makes estimate(s) in our day to day life. For example, a house
wife estimates the monthly expenditure on the basis of particular needs, a sweet
shopkeeper estimates the sale of sweets on a day, etc. So the technique of
finding an estimator to produce an estimate of the unknown parameter on the
basis of a sample is called estimation.
There are two methods of estimation:
1. Point Estimation
2. Interval Estimation

In point estimation, we determine a appropriate single statistic whose value is

used to estimate the unknown parameter whereas in interval estimation, we
determine an interval that contains true value of the unknown parameter with
certain confidence. The point estimation and interval estimation are briefly
described in Unit 6 and Unit 7 respectively of this block.
Estimation admits two problems; the first is to select some criteria or properties
such that if an estimator possesses these properties it is said to be the best
estimator among possible estimators and the second is to derive some methods
or techniques through which we obtain an estimator which possesses such
properties. This unit is devoted to explain the criteria of good estimator. This
unit is divided into nine sections. Section 5.1 is introductory in nature. The
basic terms used in estimation are defined in Section 5.2. Section 5.3 is devoted
to criteria of good estimator which are explained one by one in subsequent
5
Estimation sections. Section 5.4 is explored the concept of unbiasedness with examples.
Unbiasedness is based on fixed sample size whereas the concept based on
varying sample size, that is, consistency is described in Section 5.5. There
exists more than one consistent estimator of a parameter, therefore in Section
5.6 is explained the next property efficiency. Section 5.7 is devoted to describe
sufficiency. Unit ends by providing summary of what we have discussed in this
unit in Section 5.8 and solution of exercises in Section 5.9.
Objectives
After studying this unit, you should be able to:
 define the parameter space and joint probability density (mass) function;
 describe the characteristics of an estimator;
 explain the unbiasedness of an estimator;
 explain the consistency of an estimator;
 explain the efficiency of an estimator;
 explain the most efficient estimator;
 explain the sufficiency of an estimator; and
 describe the minimum variance unbiased estimator.

5.2 BASIC TERMINOLOGY

Before discussing the properties of a good estimator, we discuss basic
definitions of some important terms. These terms are very useful in
understanding the fundamentals of theory of estimation discussed in this block.
Discrete and Continuous Distributions
In Units 12 and 13 of MST-003, we have discussed standard discrete and
continuous distributions as binomial, Poisson, normal, exponential, etc. We
know that the populations can be described with the help of distributions,
therefore, standard discrete and continuous distributions are used in statistical
inference. Here, we discuss some standard discrete and continuous distributions
in brief as in tabular form:
S. Distribution Parameter(s) Mean Variance
No.

1 Bernoulli (discrete)
1 x p p pq
P  X  x   px 1  p  ; x  0,1

2 Binomial (discrete)
n&P np npq
P  X  x   n Cx pxq n x ; x  0,1,..., n
Poisson (discrete)
3
e  x λ λ λ
P X  x  ; x  0,1,...&   0
x!
Uniform (discrete)
4
1
n 1 n2  1
n
P  X  x   ; x  0,1,..., n 2 12
n
Hypergeometric (discrete)
5 nM NM  N  M  N  n 
M
Cx N  M Cn  x N, M & n
P X  x  N
; x  0,1,..., min M, n N N 2  N  1
Cn

6
 Introduction to Estimation
6 Geometric (discrete) p p
p
P  X  x   pq x ; x  0,1, 2,... q q2

7 Negative Binomial (discrete) rp rp

r&p
P  X  x   x r1 Cr 1prq x ; x  0,1,2, ... q q2

Normal(continuous)
2
1  x  
8 1   
f x  e 2  
;   x   µ & σ2 µ σ2
 2
&   0,      

Standard Normal(continuous)
9
1  12 x 2 -- 0 1
f x  e ; x
2
Uniform (continuous) 2

1 a&b
ab b  a
10 f x  ; a  x  b, b  a 2 12
ba
Exponential (continuous) 1 1
11 θ
f  x   ex ; x  0 &   0  2
Gamma (continuous)
b b
12 ab a&b
f  x   eax x b 1; x  0 &a  0 a a2
b
Beta First Kind (continuous)
1 b1
ab
13 f x   x a 1 1  x  ; 0  x  1 a 2
B  a,b  a&b  a  b  a  b  1
ab
& a  0, b  0

Beta Second Kind (continuous)

14 1 x a 1 a a  a  b  1
f x  ; x0 a&b 2
B  a, b  1  x a  b b 1  b  1  b  2 
& a  0,b  0

Parameter Space
The set of all possible values that the parameter  or parameters 1, 2, …, k
can assume is called the parameter space. It is denoted by Θ and is read as
“big theta”. For example, if parameter  represents the average life of electric
bulbs manufactured by a company then parameter space of  is    :   0,
that is, the parameter average life  can take all possible values greater than or
equal to 0, Similarly, in normal distribution (, σ2), the parameter space of
parameters  and σ2 is   (, 2 ) :     ; 0    .

Joint Probability Density (Mass) Function

If X1 ,X 2 , ..., X n is a random sample of size n taken from a population whose
probability density (mass) function is f(x,θ) where,  is the population
parameter then the joint probability density (mass) function of sample values is
denoted by f  x1 , x1 ,..., x n ,  and defined as

For discrete case,

f  x1 , x1,..., x n ,   P  X1  x 1 , X 2  x 2 ,..., X n  x n 
7
Estimation since X1 ,X 2 , ..., X n are independent, therefore,

f  x1 , x1 ,..., x n ,   P  X1  x1  P  X 2  x 2 ...P  X n  x n 

In this case, the function f  x1 , x1 ,..., x n ,  represents the probability that the
particular sample x1 , x 2 , ..., x n has been drawn for a fixed (given) value of
parameter θ.
For continuous case,
f  x1 , x1,..., x n ,   f  x1,  .f  x 2 ,  ... f  x n , 

In this case, the function f  x1 , x1 ,..., x n ,  represents the probability density

function of the random sample X1 ,X 2 , ..., X n .

The process of finding the joint probability density (mass) function is described
by taking some examples as:
If a random sample X1 ,X 2 , ..., X n of size n is taken from Poisson distribution
whose pdf is given by
e x
P X  x  ; x  0, 1, 2, ... &   0
x!
then joint probability mass function of X1 ,X 2 , ..., X n can be obtained as

f  x1 , x1,..., x n ,    P  X1  x1  P  X 2  x 2 ...P  X n  x n 

e  x1 e x2 e  xn

 . ...
x1! x 2! x n!
 ... 
 
e  x1  x2  ...  xn
n  times


x1 !x 2 ! ... x n !
n

 nλ
 xi
e λi 1
 n
Π x i!
i1

Similarly, if X1 ,X 2 , ..., X n is a random sample of size n taken from

exponential population whose pdf is given by
f  x,   e x ; x  0,   0
then the joint probability density function of sample values can be obtained as
f  x1 , x1 ,..., x n ,    f  x1 ,  .f  x 2 ,  ... f  x n ,  

 ex1 .ex 2 ... ex n


1  ...
1 1
 n  times
e  x1  x 2 ... x n 
n

n
  xi
f  x1 , x1 ,..., x n ,    e i 1

8
Let us check your understanding of above by answering the following Introduction to Estimation
exercises.
E1) What is the pmf of Poisson distribution with parameter λ = 5. Also find
the mean and variance of this distribution.
E2) If  represents the average marks of IGNOU’s students in a paper of 50
marks. Find the parameter space of .
E3) A random sample X1 , X 2 , ..., X n of size n is taken from Poisson
distribution whose pdf is given by
e  x
P X  x  ; x  0, 1, 2, ... &   0
x!
Obtain joint probability mass function of X1 ,X 2 , ..., X n .

5.3 CHARACTERISTICS OF ESTIMATORS

It is to be noted that a large number of estimators can be proposed for an
unknown parameter. For example, if we want to estimate the average income
of the persons living in a city then the sample mean, sample median, sample
mode, etc. can be used to estimate the average income. Now, the question
arises, “Are some of possible estimators better, in some sense, than the others?”
Generally, an estimator can be called good for two different situations:
(i) When the true value of parameter is being estimated is known− An
estimator might be called good if its value close to the true value of the
parameter to be estimated. In other words, the estimator whose sampling
distribution concentrates as closely as possible near the true value of the
parameter may be regarded as the good estimator.
(ii) When the true value of the parameter is unknown− An estimator may
be called good if the data give good reason to believe that the estimate will
be closed to the true value.
In the whole estimation, we estimate the parameter when the true value of the
parameter is unknown. Hence, we must choose estimates not because they are
certainly close to the true value, but because there is a good reason to believe
that the estimated value will be close to the true value of parameter. In this unit,
we shall describe certain properties, which help us in deciding whether an
estimator is better than others. Prof. Ronald. A. Fisher,
a Great English
Prof. Ronald A. Fisher was the man who pushed ahead the theory of estimation Mathematical
and introduced these concepts and gave some properties of good estimator as Statistician.
follows: (1890-1962)

1. Unbiasedness
2. Consistency
3. Efficiency
4. Sufficiency
We shall discuss these properties one by one in the subsequent sections.
Now, give the answer of the following exercise.
E4) Write the four properties of good estimator.
 Estimation
5.4 UNBIASEDNESS
Generally, population parameter(s) is (are) unknown and if the whole
population is too large to study to find the value of unknown parameter(s) then
one can estimate the population parameter(s) with the help of estimator(s)
which is(are) always a function of sample values.
An estimator is said to be unbiased for the population parameter such as
population mean, population variance, population proportion, etc. if and only if
An estimator is said to the average or mean of the sampling distribution of the estimator is equal to the
be unbiased if the true value of the parameter.
expected value of the
estimator is equal to the Mathematically,
true value of the If X1 ,X 2 , ..., X n is a random sample of size n taken from a population whose
parameter being
estimated. probability density (mass) function is f(x,θ) where,  is the population
parameter then an estimator T = t( X1 ,X 2 , ..., X n ) is said to be unbiased
estimator of the parameter  if and only if
E(T)  θ ; for all θ  Θ
This property of estimator is called unbiasedness.
Normally, it is preferable that the expected value of the estimator should be
exactly equal to the true value of the parameter being estimated. But if the
expected value of the estimator does not equal to the true value of parameter,
then the estimator is said to be “biased estimator”, that is, if
E (T )  θ
then estimator T is called biased estimator of .
The amount of biasness is given by
b(θ)  E (T)  θ
If b(θ) > 0 or E(T) > θ, then the estimator T is said to be positively biased for
parameter .
If b(θ) < 0 or E(T) < θ, then the estimator T is said to be negatively biased for
parameter .
If E (T)  θ as n   i.e. if an estimator T is unbiased for a large sample only
then estimator T is said to be asymptotically unbiased for .
Now, we explain the procedure how to show that a statistic is unbiased or not
for a parameter with the help of some examples:
Example 1: Show that sample mean (X) is an unbiased estimator of the
population mean ( ) if it exists.

Solution: Let X1 ,X 2 , ..., X n be a random sample of size n taken from any

population with mean . Then for unbiasedness we have to show that
E X  

Consider,
 X  X 2  ...  X n 
E  X  E  1   By defination of sample mean 
 n
10
 1 Introduction to Estimation
  E  X1   E  X 2   ...  E  X n    E  aX  bY   aE  X   bE  Y 
n
Since X1, X2,…,Xn are randomly drawn from same population so they also If X and Y are two random
follow the same distribution as the population. Therefore, variables and a & b are two
E(X1 )  E(X 2 )  ...  E(X n )  E(X)   constants then by the
addition theorem of
Thus, expectation, we have
E  aX  bY   aE  X   bE  Y 
1 
E(X)       ...   
n  

n  times


1
  n   
n
E  X  

Hence, sample mean  X  is an unbiased estimator of the population mean .

Also if x1 , x 2 , ..., x n are the observed values of the random sample
1 n
X1 ,X 2 , ..., X n then x   x i is unbiased estimate of population mean.
n i1
Example 2: A random sample of 10 cadets of a centre is selected and
measures their weights (in kg) which are given below:
48, 50, 62, 75, 80, 60, 70, 56, 52, 78
Determine an unbiased estimate of the average weight of cadets of the centre.
Solution: We know that sample mean (X) is an unbiased estimator of the
population mean and its particular value is the unbiased estimate of population
mean, therefore,
1 n X  X2  ...  X n
X 
n i1
Xi  1
n

48  50  62  75  80  60  70  56  52  78
  63.10
10
Hence, an unbiased estimate of the average weight of cadets of the centre is
63.10 kg.
Example 3: A random sample X1 ,X 2 , ..., X n of size n taken from a population
whose pdf is given by
1
f ( x, )  e x /  ; x  0 ,  0

Show that sample mean  X  is an unbiased estimator of parameter .

Solution: For unbiasedness, we have to show that

E X  

Here, we are given that

11
Estimation 1
f ( x, )  e x /  ; x  0 ,  0

Since we do not know the mean of this distribution therefore first of all we find
the mean of this distribution. So we consider,

E  X    x f  x,  dx
0

 
1 1
  x ex / dx   x ex / dx
0
 0

1 21  x /  1 2   n 1  ax n
  x e dx     x e dx  n 
0  1/ 2  0 a 

Since X1, X2,…,Xn are randomly drawn from same population having mean θ,
therefore,

E  X1   E  X 2   ...  E  X n   E  X   

Consider,
If X and Y are two random
 X  X 2  ...  X n 
variables and a & b are two E  X  E  1   By defination of sample mean 
constants then by the  n
addition theorem of 1  E  aX  bY  
  E  X1   E  X 2   ...  E  X n   
expectation, we have n   aE  
X  bE   
Y
E  aX  bY   aE  X   bE  Y 
1
 (

 ... )

n n  times

1
  n   
n
Thus, X is an unbiased estimator of .

Note 1: If X1 ,X 2 , ..., X n is a random sample taken from a population with

mean  and variance σ2, then

1 n 2
S2  
n i1
 Xi  X is a biased estimator of σ2.

n 2 1 n 2
whereas, S2 
n 1
S i.e. S2  
n  1 i1
 Xi  X is an unbiased estimator of σ2.

The proof of the above result is beyond the scope of this course. But for your
convenience, we will show this result with the help of the following example.
Example 4: Consider a population comprising three televisions of certain
company. If lives of televisions are 8, 6 and 10 years then construct the
sampling distribution of average life of Televisions by taking samples of size 2
and show that sample mean is an unbiased estimator of population mean life.
Also show that S2 is not an unbiased estimator of population variance whereas
S2 is an unbiased estimator of population variance where,

12
 1 n 2 1 n 2 Introduction to Estimation
S2  
n i1
 Xi  X and S2  
n 1 i1
 Xi  X

Solution: Here, population consists three televisions whose lives are 8, 6 and
10 years so we can find the population mean and variance as
8  6  10
 8
3
1 2 2 2 8
2   8  8    6  8   10  8     2.67
3   3
Here, we are given that
Population size = N = 3 and sample size = n = 2
Therefore, possible numbers of samples (with replacement) that can be drawn
from this population are Nn = 32 = 9. For each of these 9 samples, we will
calculate the values of X,S2 and S2 by the formulae given below:

1 n 1 n 2 1 n 2
X 
n i1
X i , S2
 
n i1
 X i  X  and S2
 
n  1 i1
 Xi  X

and necessary calculations for these results are shown in Table 5.1 given
below:
Table 5.1: Calculation for X, S2 and S2

Sample Sample X 2
2 S2 S2
Observation  Xi  X 
i 1

1 8, 8 8 0 0 0
2 8, 6 7 2 1 2
3 8,10 9 2 1 2
4 6, 8 7 2 1 2
5 6, 6 6 0 0 0
6 6, 10 8 8 4 8
7 10, 8 9 2 1 2
8 10, 6 8 8 4 8
9 10, 10 10 0 0 0
Total 72 12 24

We calculate X,S2 and S2 as

1 1 1
X1  8  8  8, X 2  8  6   7,..., X 9  10  10  10
2 2 2
1 2 2 1 2 2
S12   8  8  8  8   0, S22  8  7    6  7    1,...,
2   2  
1 2 2
S9  10  10   10  10    0
2

2  

13
Estimation 1  2 2 1  2 2
S12 
 8  8   8  8    0, S 22   8  7    6  7    2,...,
2 1 2 1
1  2 2
S92 
 10  10   10  10    0
2 1
Form the Table 5.1, we have
1 k 1
E X   X i   72  8  
k i 1 9
Hence, sample mean is unbiased estimator of population mean.
Also
1 k 2 1
E S2    Si   12  1.33   2
k i 1 9

Therefore, S2 is not an unbiased estimator of σ2 whereas,

1 k 2 1
E S2    Si   24  2.67   2
k i 1 9
S2 is unbiased estimator of σ2.
Remark 1:
1. Unbiased estimators may not be unique. For example, sample mean and
sample median are unbiased estimators of population mean of normal
population.
2. Unbiased estimators do not always exist for all the parameters. For
example, for a Bernoulli distribution (), there is no unbiased estimator for
2. Similarly, for a Poisson distribution (λ), there exists no unbiased
estimator for 1/ λ.
3. If an estimator is unbiased for all types of distribution, then it is called an
absolutely unbiased estimator. For example, sample mean is an absolutely
unbiased estimator of population mean, if population mean exists.
4. If Tn and Tn* are two unbiased estimator of parameter  then aTn+ (1−a) Tn*
is also unbiased estimator of  where, ‘a’ (0  a  1) is any constant.
For the better understanding of the unbiasedness try some exercises.

E5) If X1 ,X 2 , ..., X n is a random sample taken from Poisson distribution

whose probability mass function is given by
e   x
P X  x  ; x  0, 1, 2, ... &   0
x!
then show that the sample mean X is unbiased estimator of λ.
E6) If X1 ,X 2 , ..., X n is a random sample of size n taken from a population
whose pdf is
f x,   e x   ;  x ; 
then show that the sample mean is an unbiased estimator of (1 + ).

14
One weakness of unbiasedness is that it requires only the average value of the Introduction to Estimation
estimator equals to the true value of population parameter. It does not require
those values of the estimator to be reasonably close to the population
parameter. For this reason, we require some other properties of good estimator
as consistency, efficiency and sufficiency which are described in subsequent
sections.

5.5 CONSISTENCY
In previous section, we have learnt about the unbiasedness. An estimator T is
said to be unbiased estimator of parameter, say,  if the mean of sampling
distribution of estimator T is equal to the true value of the parameter . This
concept was defined for a fixed sample size. In this section, we will learn about
consistency which is defined for increasing sample size.
If X1 ,X 2 , ..., X n is a random sample of size n taken from a population whose
probability density (mass) function is f(x,θ) where,  is the population
parameter then consider a sequence of estimators, say, T1 = t1(X1),
T2 = t2(X1, X2), T3 = t3(X1, X2, X3),…, Tn = tn(X1, X2, ..., Xn) . A sequence of
estimators is said to be consistent for parameter  if the deviation of the values
of estimator from the parameter tends to zero as the sample size increases. That
means values of estimators tend to get closer to the parameter  as sample size
increases.
In other words, a sequence {Tn} of estimators is said to be consistent sequence
of estimators of  if Tn converges to  in probability, that is
p
Tn    as n   for every   … (3)

or for every   0

lim P  Tn       1 ... (4)

n 

or for every   0 and   0 there exist n  m such that

P  Tn       1   ; nm … (5)

where, m is some very large value of n. Expressions (3), (4) and (5) are to
mean the same thing.
Generally, to show that an estimator is consistent with the help of above
definition is slightly difficult, therefore, we use sufficient conditions for
consistency which are given below:
Sufficient conditions for consistency
If {Tn}is a sequence of estimators such that for all   

(i) E  Tn    as n  , that is, estimator Tn is either unbiased or

asymptotically unbiased estimator of  and

(ii) Var  Tn   0 as n  , that is, variance of estimator Tn converges to 0 as

n .

15
Estimation Then estimator Tn is a consistent estimator of .
Now, we explain the procedure based on both the criteria (definition and
sufficient condition) to show that a statistic is consistent or not for a parameter
with the help of some examples:
Example 5: Prove that sample mean is always a consistent estimator of the
population mean provided that the population has a finite variance.

Solution: Let X1 ,X 2 , ..., X n be a random sample taken from a population

having mean  and finite variance σ2. By the definition of consistency, we have
lim P  Tn       lim P  X       Here, Tn  X 
n  n 

 X   n 
 lim P   
n 
 / n  

By central limit theorem (described in Unit 1 of this course), we know that the
X 
variate Z  is a standard normal variate for large sample size n.
/ n
Therefore,
  n
lim P  Tn  θ  ε   lim P  Z  
n  n
  

  n  n
 lim P  Z   X  a  a  X  a 
n
    
 n
 b
 
 lim
n   f zdz  P  a  U  b   f  u  du 
 n  a 

 n

1 z2 / 2  1 z2 / 2 
 lim
n   2
e dz  f  z   2  e 
 n  



1  z2 / 2
  e dz
 2

1 z 2 / 2
Since e is the pdf of a standard normal variate Z therefore, the
2π
integration of this in whole range   to  is unity.
Thus,
lim P  Tn       lim P  X       1 as n  
n  n 

Hence, sample mean is a consistent estimator of population mean.

Note 2: This example can also be proved with the help of sufficient conditions
for consistency as shown in next example.

16
Example 6: If X1 ,X 2 , ..., X n is a random sample taken from Poisson Introduction to Estimation

distribution(λ), then show that sample mean  X  is consistent estimator of λ.

Solution: We know that the mean and variance of Poisson distribution (λ) are
E  X    and Var  X   

Since X1 ,X 2 , ..., X n are independent and come from same Poisson distribution,
therefore,
E  X i   E  X    and Var  X i   Var  X    for all i  1, 2, ..., n

Now consider,

1 
E  X   E   X1  X2  ...  Xn    Bydefination of sample mean
n 

1  E  aX  bY  
  E  X1   E  X 2   ...  E  X n  
n   aE  X   bE  Y  
 

1  1
  
    =  n   
 ... 
n n  times  n

Thus, sample mean  X  is unbiased estimator of .

Now consider,

1 
Var  X   Var   X1  X2  ...  Xn  
n 
 If X and Y are two 
1 independent 
 2  Var  X1   Var  X 2   ...  Var  X n    random variable then 
n  Var(aX + bY) 
= a 2 Var(X) + b2 Var(Y) 
 
1  
  
2  
 
 ... 

n  n  times 
1
 nλ 
n2

Var  X    0 as n  
n
Hence, by sufficient condition of consistency, it follows that sample mean (X)
is consistent estimator of λ.

Remark 2:
1. Consistent estimators may not be unique. For example, sample mean and
sample median are consistent estimators of population mean of normal
population.
2. An unbiased estimator may or may not be consistent.

3. A consistent estimator may or may not be unbiased.

17
Estimation 4. If Tn is a consistent estimator of  and f is a continuous function of  then
f  Tn  is consistent estimator of f (). This property is known as invariance
Continuous function is property. For example, if X is a consistent estimator of population mean θ
described in Sections
5.7 and 5.8 of Unit 5 of then eX also consistent estimator of e because e is a continuous function of
MST- 001. θ.
It is now time for you to try the following exercises to make sure that you have
understood consistency.

E7) If X1 ,X 2 , ..., X n is a random sample of size n taken from pdf

1 ;   x    1
f  x,   
0 ; elsewere
then show that the sample mean is an unbiased as well as consistent
 1
estimator of  θ   .
 2

E8) If X1 ,X 2 , ..., X n are n observations taken from geometric distribution

with parameter , then show that X is consistent estimator of 1/. Also
find consistent estimator of e1/.

5.6 EFFICIENCY
In some situations, we see that there are more than one estimators of an
parameter which are unbiased as well as consistent. For example, sample mean
and sample median both are unbiased and consistent for the parameter  when
sampling is done from normal population with mean  and known variance σ2.
In such situations, there arises a necessity of some other criterion which will
help us to choose ‘best estimator’ among them. A criterion which is based on
the concept of variance of the sampling distribution of the estimator is termed
as efficiency.
If T1 and T2 are two estimators of an parameter . Then T1 is said to be more
efficient than T2 for all sample sizes if
Var  T1   Var  T2  for all n

Let us do some examples about efficiency:

Example 7: Show that sample mean is more efficient estimator than sample
median for estimating mean of normal population.
Solution: Let X1 ,X 2 , ..., X n be a random sample taken from normal
population with mean  and variance σ2. Also let X and X  be the sample
mean and sample median respectively. We have seen in Unit 2 that the
sampling distribution of the mean from a normal population follows normal
distribution with means  and variance σ2/n. Similarly, it can be shown that the
sampling distribution of the median from a normal population also follows
 2
normal distribution with mean  and variance . Therefore,
2 n

18
 σ2 Introduction to Estimation
Var  X  
n
2
   πσ
Var  X
2n

 2 2   2    . Thus, we
But   and  1 therefore, Var  X   Var  X
2n n  2 n 
conclude that sample mean is more efficient estimator than sample median.
Example 8: If X1, X2, X3, X4 and X5 is a random sample of size 5 taken from a
population with mean  and variance σ2. The following two estimators are
suggested to estimate 
X1  X 2  X 3  X 4  X 5 X  2X 2  3X 3  4X 4  5X 5
T1  , T2  1
5 15
Are both estimators unbiased? Which one is more efficient?
Solution: Since X1, X2,…,X5 are independent and taken from same population
with mean  and variance σ2 therefore,
E  X i    and Var  X i   2 for all i  1,2,...,5
Consider,
 X  X2  X3  X4  X5 
E  T1   E  1 
 5
1
  E  X1   E  X 2   E  X 3   E  X 4   E  X 5  
5
1
 µ  µ  µ  µ  µ 
5
E T1   µ
Similarly,
 X  2X 2  3X3  4X4  5X5 
E  T2   E  1 
 15
1
  E  X1   2E  X 2   3E  X 3   4E  X 4   5E  X 5  
15 
1
 µ  2µ  3µ  4µ  5µ
15
1
 15 
15
E  T2   

Hence, both the estimators T1 and T2 are unbiased estimators of .

Now for efficiency, we consider
 X  X 2  X 3  X 4  X5 
Var  T1   Var  1 
 5

19
Estimation 1
  Var  X1   Var  X 2   Var  X 3   Var  X 4   Var  X 5  
25 
1 2

25

σ  σ2  σ2  σ2  σ2 
1

25
 52 
1
Var  T1    2
5
Similarly,
 X  2X2  3X3  4X 4  5X5 
Var  T2   Var  1 
 15

1  Var  X1   4Var  X 2   9Var  X 3  

  
225   16Var  X 4   25Var  X5  


1 2 55σ 2

225

σ  4σ 2  9σ 2  16σ 2  25σ 2 
225

11σ 2
VarT2  
45
Since, Var  T1   Var  T2  , therefore, we conclude that estimator T1 is more
efficient than T2.
5.6.1 Most Efficient Estimator
In a class of estimators of a parameter, if there exists one estimator whose
variance is minimum (least) among the class, then it is called most efficient
estimator of that parameter. For example, suppose T1, T2 and T3 are three
estimators of parameter θ having variance 1/n, 1/(n+1) and 5/n respectively.
Since variance of estimator T2 is minimum, therefore, estimator T2 is most
efficient estimator in that class.
The efficiency of an estimator measured with respect to the most efficient
estimator is called “Absolute Efficiency”. If T* is the most efficient estimator
having variance Var(T*) and T is any other estimator having variance Var(T),
then efficiency of T is defined as

Var  T* 
e
Var  T 

Since variance of most efficient estimator is minimum, therefore,

Var  T* 
e 1
Var  T 

5.6.2 Minimum Variance Unbiased Estimator

An estimator T of parameter  is said to be minimum variance unbiased
estimator (MVUE) of  if and only if
20
(i) E (T) =  for all    , that is, estimator T is unbiased estimator of  and Introduction to Estimation

(ii) Var  T   Var  T for all    , that is, variance of estimator T is less
than or equal to variance of any other unbiased estimator T.
The minimum variance unbiased estimator (MVUE) is the most efficient
unbiased estimator of parameter  in the sense that it has minimum variance in
class of unbiased estimators. Some authors used uniformly minimum variance
unbiased estimator (UMVUE) in place of minimum variance unbiased
estimator (MVUE).
Now, you can try the following exercises.

E9) If X1 ,X 2 , ..., X n is a random sample taken from a population having

1 n
mean  and variance σ2, then show that the statistic T   Xi is
n  1 i1
biased but more efficient than sample mean for estimating the
population mean.
E10) Suppose X1 ,X 2 , ..., X n is a random sample taken from normal
population with mean  and variance σ2. The following two estimators
are suggested to estimate  as
X1  X 2  X 3 X1  X 2
T1  and T2   X3
3 2
Are both estimators unbiased? Which one of them is more efficient?
E11) Define most efficient estimator and minimum variance unbiased
estimator.

5.7 SUFFICIENCY
In statistical inference, the aim of the investigator or statistician may be to
make a decision about the value of the unknown parameter (). The
information that guides the investigator in making a decision is supplied by the
random sample X1 ,X 2 , ..., X n . However, in most of the cases the observations
would be to numerous and too complicated. Directly use of these observations
is complicated or cumbersome, therefore, a simplification or condensation
would be desirable. The technique of condensing or reducing the random
sample X1 ,X 2 , ..., X n into a statistic such that it contains all the information
about parameter  that is contained in the sample is known as sufficiency. So
prior to continuing our search of finding best estimator, we introduce the
concept of sufficiency.
A sufficient statistic is a particular kind of statistic that condenses random
sample X1 , X 2 , ..., X n in a statistic T  t(X1 , X 2 , ..., X n ) in such a way that no
information about parameter  is lost. That means, it contains all the
information about  that is contained in the sample and if we know the value of
sufficient statistic, then the sample values themselves are not needed and can
nothing tell you more about . In other words,

21
Estimation A statistic T is said to be sufficient statistic for estimating a parameter  if it
contains all the information about  which are available in the sample. This
property of an estimator is called sufficiency. In other words,
An estimator T is sufficient for parameter  if and only if the conditional
distribution of X1 , X 2 , ..., X n given T = t is independent of .

Mathematically,
f  x1 , x 2 , ..., x n / T  t   g  x1 , x 2 , ..., x n 

where, the function g  x1 , x 2 ,..., x n  does not depend on the parameter .

Note 3: Generally, the above definition is used to show that a particular

statistic is not a sufficient statistic because it may be very tedious task to obtain
the conditional distribution. Hence, we use factorization theorem which
facilitates us to find sufficient statistic without any difficulty which is given
below.
Theorem (Factorization Theorem): Let X1 , X 2 , ..., X n be a random sample
of size n taken from the probability density (mass) function f(x, ). A statistic
or estimator T is said to be a sufficient for parameter  if and only if the joint
density (mass) function of X1 , X 2 , ..., X n can be factored as

f  x1, x 2 , ..., x n ,    g  t(x),  . h  x 1 , x 2 , ..., x n 

where, the function g  t(x),  is a non-negative function of parameter  and

observed sample values (x1, x 2 ,..., x n ) only through the function t(x) and the
function h  x1 , x 2 ,..., x n  is non-negative function of (x1, x 2 ,..., x n ) and does not
involve the parameter .
For applying factorization theorem, we try to factor the joint density (mass)
function as the product of two functions, one of which is function of
parameter(s) and another is independent of parameter(s).
The proof of this theorem is beyond the scope of this course.
Note 4: The factorization theorem should not be used to show that a given
statistic or estimator T is not sufficient.
Now, do some examples to show that a statistic is sufficient estimator for a
parameter by using the factorization theorem:
Example 9: Show that the sample mean is sufficient for the parameter  of the
Poisson distribution.
Solution: We know that the probability mass function of Poisson distribution
with parameter  is
e x
P X  x  ; x  0, 1, 2, ... &   0
x!
Let X1 , X 2 , ..., X n be a random sample taken from Poisson distribution with
parameter . Then joint mass function of X1 , X 2 , ..., X n can be obtained as

f  x1 , x 2 ,..., x n ,    P  X1  x1  . P  X 2  x 2  ... P  X n  x n 

22
 e  x1 e x2 e  xn Introduction to Estimation
 . ...
x1! x 2! x n!
 ... 
 
e n  times
 x1  x 2  ...  x n

x1 !x 2 ! ... x n !
n
 xi
e  n
 i 1 n 
 n i 1
x i ! repersents the 
 xi !  product of x i ! 
i 1

The joint mass function can be factored as

 
 1   1 n 
f  x1 , x 2 ,..., x n ,     e  n
  nnx
  x  n  xi 
  xi !  i 1 
 i1 
 g  t(x),  . h  x1 , x 2 ,..., x n 

where, g  t(x),    e  n  n x is a function of parameter  and the observed

1
sample values x1 , x 2 ,..., x n only through t(x)  x and h  x 1 , x 2 ,..., x n   n is
 xi !
i 1

a function of sample values x 1 , x 2 ,..., x n and is independent of parameter .

Hence, by factorization theorem of sufficiency, X is sufficient statistic for .

Note 4: Since throughout the course we are using capital letter for statistic or
estimator therefore in the last line of above example we use X in place of x.
Thus, in all the examples and exercises relating to sufficient statistic, we are
using similar approach.
Example 10: A random sample X1 ,X 2 , ..., X n of size n is taken from gamma
distribution whose pdf is given by
a b ax b 1
f  x,a, b   e x
b
Obtain sufficient statistic for
(i) ‘b’ when ‘a’ is known
(ii) ‘a’ when ‘b’ is known
(iii) ‘a’ and ‘b’ both are unknown.
Solution: The joint density function can be obtained as
f  x1, x 2 ,..., x n , a, b   f  x1, a, b  f  x 2 , a, b  ...f  x n , a, b 

a b  ax1 b1 a b ax2 b1 a b axn b1

 e x1 . e x 2 ... e x n
b b b
n
b 1
a nb  xi n
a
 n
e i 1   x i 
 b  i 1 
23
Estimation Case I: When ‘a’ is known then we find sufficient statistic for ‘b’
The joint density function can be factored as
 nb n b 1   a x
i
n
Since ‘a
a  
f  x1 , x 2 ,..., x n ,b    x  .e i 1 ‘a’ is tre
 b n  i1 i  
   
constan

 g  t(x), b .h  x1 , x 2 ,...x n 

b 1
a nb  n 
where, g  t(x), b   n   xi 
is a function of parameter ‘b’ and sample
 b 
i 1 

n
n a  xi
values x1 , x 2 ,..., x n only through t(x)   x i and h  x1 , x 2 ,..., x n   e i 1
is a
i 1

function of sample values x1 , x 2 ,..., x n and is independent of parameter ‘b’.

n
Hence, by factorization theorem of sufficiency,  Xi is sufficient statistic for
i 1
parameter ‘b’
Case II: When ‘b’ is known then we find sufficient statistic for ‘a’
The joint density function can be factored as
n
 a
 xi  b 1
n
 nb
f  x1 , x 2 ,..., x n ,a   a e
i 1 . 1   x 
  n  i
Since ‘b’ is known so ‘b’
is treated as a constant.   b
 i1   
 g  t(x), a .h  x1 , x 2 ,...x n 
n
a
 xi
i 1
where, g  t(x),a   a nbe is a function of parameter ‘a’ and sample values
n b 1
1  n 
x1 , x 2 ,..., x n only through t(x)   x i and h  x 1 , x 2 ,..., x n   n   xi 
is a
i1
 
b 
i 1 
function of sample values x1 , x 2 ,..., x n and is independent of parameter ‘a’.
n
Hence by factorization theorem of sufficiency,  Xi is sufficient statistic for
i1
‘a’.
Case III: When ‘a’ and ‘b’ are unknown then we find jointly sufficient
statistics for ‘a’ and ‘b’
The joint density function can be factored as
n
 a
 xi b 1 
1  n 
f  x 1 , x 2 ,..., x n , a, b    a nbe  .1
i 1
x
n  i 1 i 
 
   b  


 gt1 ( x ), t 2 ( x ), a, b. h x1 , x 2 , ..., x n 

24
 n
Introduction to Estimation
a
 xi 1  n 
b 1
i 1
where, g  t1 (x), t 2 (x), a, b   a nbe n   xi 
is a function of
  b 
i 1 
n
parameters ‘a’& ‘b’ and sample values x1 , x 2 ,..., x n only through t1 (x)   x i
i 1
n
and t 2 (x)   x i whereas, h  x1 , x 2 ,...x n   1 and independent of parameters ‘a’
i 1
and ‘b’.
n n
Hence, by factorization theorem,  X and  X are jointly sufficient for
i1
i
i 1
i

parameters ‘a’& ‘b’.

Example 11: If X1 , X 2 , ..., X n is a random sample taken from uniform
distribution U(, ), find the sufficient statistics for  and .
Solution: The probability density function of U(, ) is given by
1
f x, ,   ;  x 

The joint density function can be obtained as
f  x1 , x 2 ,..., x n , ,    f  x1 , ,   .f  x 2 , ,   ... f  x n , ,  
1 1 1
 . ...
β α βα β α
1

β  αn
ics of Since the range of variable depends upon the parameters so we consider
e ordered statistics X 1 , X 2  , ..., X  n  . Therefore, the joint density function can be
re the factored as
aced
1
er of f  x1 , x 2 ,..., x n , ,    n
;   x 1  x 2   ...  x  n   
e are    
 1 
Xn  I x 1 ,  I2 x n  , .1
n 1
     
where, x 1 and x  n  are the minimum and maximum sample observations
respectively and
1; if x 1  
 
I1 x 1 ,   
0; otherwise
 1; if x  n   
 
I2 x n ,   
 0; otherwise
Therefore,
f  x1 , x 2 ,..., x n , ,    g  t1 (x), t 2 (x), ,   .h  x1 , x 2 ,..., x n 

25
Estimation  
1
where, g  t1 (x), t 2 (x), ,     n
I
1 x 1 , 
 I 2 x n ,   is a function of
     
parameters (α, β) and sample values x1 , x 2 ,..., x n only through t1 (x)  x 1 and
t 2 (x)  x  n  whereas, h  x1 , x 2 ,...x n   1 and independent of parameters ‘α’ and
‘β’.
Hence, by factorization theorem of sufficiency, X 1 and X  n  are jointly
sufficient for  and .
Remark 3:
1. A sufficient estimator is always a consistent estimator.
2. A sufficient estimator may be unbiased.
3. A sufficient estimator is the most efficient estimator if an efficient
estimator exists.
4. The random sample X1 , X 2 ,..., X n and order statistics X 1 , X  2  ,..., X  n  are
always sufficient estimators because both are contain all the information
about the parameter(s) of the population.
5. If T is a sufficient statistic for the parameter  and (T) is a one to one
function of T then (T) is also sufficient for θ. For example, if T   X i is
1 T
sufficient statistic for parameter θ then X 
n
 X i  n is also sufficient
T
for θ because X  is a one to one function of T.
n
Now, you will understand more clearly about the sufficiency, when you try the
following exercises.
E12) If X1 , X 2 , ..., X n is a random sample taken from exp () then find
sufficient statistic for .
E13) If X1 , X 2 , ..., X n is a random sample taken from normal population
N(, σ2), then obtain sufficient statistic for  and σ2or both according as
other parameter is known or unknown.
E14) If X1 , X 2 , ..., X n is a random sample from uniform population over the
interval [0, ]. Find sufficient estimator of .
We now end this unit by giving a summary of what we have covered in it.

5.8 SUMMARY
In this unit, we have covered the following points:
1. The parameter space and joint probability density (mass) function.
2. The basic characteristics of an estimator.
3. Unbiasedness of an estimator.
4. Consistency of an estimator.
5. Efficiency of an estimator.
26
6. The most efficient estimator. Introduction to Estimation

7. Minimum variance unbiased estimator.

8. The sufficiency of an estimator.

5.9 SOLUTIONS / ANSWERS

E1) We know that the pmf of Poisson distribution is
e x
PX  x  ; x  0,1, 2,...&   0
x!
and mean and variance of this distribution are
Mean = Variance = λ
In our case, λ = 5, therefore, pmf of Poisson distribution is
e 5 5x
P X  x  ; x  0,1, 2,...
x!
Also mean and variance of this distribution are
Mean = Variance = λ = 5
E2) Since parameter  represents the average marks of IGNOU’s students
in a paper of 50 marks, therefore, a student can take minimum 0 marks
and maximum 50 marks. Thus, the parameter space of  is
   : 0    50 .

E3) The probability mass function of Poisson distribution is

e x
P X  x  ; x  0, 1, 2, ... &   0
x!
The joint probability mass function of X1 , X 2 , ..., X n can be obtained as

f  x1, x 2 ,..., x n ,    P  X1  x1  . P  X 2  x 2  ... P  X n  x n 

e  x1 e  x2 e  xn

 . ...
x1! x 2! x n!
 ... 
 
e n  times
 x1  x2  ...  x n

x1 !x 2 ! ... x n !
n

 nλ
 xi
e λi 1
 n
Π xi!
i 1

E4) Refer Section 5.3.

E5) We know that the mean of Poisson distribution with parameter λ is λ
i.e.
E(X)  

27
Estimation Since X1 , X 2 , ..., X n are independent and come from same Poisson
distribution, therefore,
E(Xi )  E(X)   for all i  1, 2, ..., n
Now consider,
1   Bydefination of 
E(X)  E  (X1  X2  ...  Xn ) 
n  sample mean 

E  aX  bY 
 E(X1 )  E(X 2 )  ...  E(X n ) 
1 
 
n   aE  X   bE  Y  
1  1
  
      n   
 ... 
n n  times  n
Hence, sample mean (X) is unbiased estimator of parameter .
E6) Here, we have to show that
E X  1  

We are given that

f x,   e x  ;   x  ,      
Since we do not know the mean of this distribution therefore first of all
we find the mean of this distribution. So we consider,

E  X    x f  x,  dx



  xe x   dx


Putting x    y  dx  dy. Also when x    y  0 & when

x    y  . Therefore,

E  X     y   e y dy
0

 
  ye dy   e ydy
y

0 0

 
  y2 1 e y dy   y11e  y dy
0 0


 n 1 
 x e  xdx  n 
 2   1  1 
 0 
 and 2  1  1 

Since X1 , X 2 , ..., X n are independent and come from same population,

so
E(X i )  E(X)  1   for all i  1, 2, ..., n

28
 Now consider, Introduction to Estimation

 X  X2  ...  Xn   Bydefination of 
E  X  E  1 
 n sample mean 

1  E  aX  bY  
  E(X1 )  E(X 2 )  ...  E(X n )    aE  X   bE  Y  
n  

1 
  (1   )  (1   )  ...  (1  )

n   
n  times 

1
  n 1    
n
 1 
Thus, sample mean is an unbiased estimator of (1+).
E7) We have
f x, θ   1 ; θ  x  θ 1

This is the pdf of uniform distribution U[, +1] and we know that for
U[a, b]
2

E X 
ab
and Var  X  
b  a
2 12
In our case, a = θ and b = θ + 1, therefore
2

EX 
   1 1
   and Var  X  
   1    1
2 2 12 12
Since X1 ,X 2 , ..., X n are independent and come from same population,
therefore,
1 1
E  Xi   E  X     and Var  X i   Var  X    i  1, 2, ..., n
2 12

 1
To show that X is unbiased estimator of     , we consider
 2
1   Bydefination of 
E(X)  E  (X1  X2  ...  Xn ) 
n  sample mean 

1  E  aX  bY  
  E(X1 )  E(X 2 )  ...  E(X n )
  aE  X   bE  Y  
n  
 
1  1  1  1 
           ...      
n 
2  2  2
 
 n  times 

1   1 
 n    
n  2 

29
Estimation 1
θ
2

 1
Therefore, X is unbiased estimator of     .
 2

For consistency, we have to show that

1
E X    and Var  X   0 as n  
2
If X and Y are two
independent random Now consider,
variables and a & b are
two constants then 1 
Var  X   Var  (X1  X2  ...  Xn ) 
Var(aX + bY) n 
= a 2 Var(X) + b 2 Var(Y)
1
  Var(X1 )  Var(X 2 )  ...  Var(X n )
n2

 
1  1 1 1 
 2  12  12  ...  12 
n   
 n  times 

1n
  
n 2  12 

1
Now, Var  X    0 as n  
12n
1
Thus, E  X     and Var(X)  0 as n  
2

 1
Hence, sample mean X is also consistent estimator of     .
 2

E8) We know that the mean and variance of geometric distribution(θ) are
given by
1 1 
E(X)  and Var  X   2
 

Since X1 , X 2 , ..., X n are independent and come from same geometric

distribution, therefore,

E(X i )  E(X) and Var  X i   Var  X  for all i  1, 2,..., n

First, we show sample mean X is consistent estimator of 1/.

Therefore, we consider,

1   By defination of 
E  X   E  (X1  X2  ...  X n ) 
n  sample mean 

30
 1  E  aX  bY   Introduction to Estimation
  E(X1 )  E(X 2 )  ...  E(X n )   aE  X   bE  Y  
n  
 
11 1 1
    ...  
 
n    

 n  times 
1n 1
  
n  
Now consider,
1 
Var  X   Var  (X1  X2  ...  Xn ) 
n 
1
  Var(X1 )  Var(X 2 )  ...  Var(X n )
n2
 
1  1     1     1   
 2  2    2   ...   2 
n
      
 n  times 
1   1    1  1   
  n  2    n  2 
n2     
1 1  
Var  X    2   0 as n  
n  
1
Since E  X   and Var(X)  0 as n  

Hence, sample mean X is consistent estimator of 1/.
Since e1/  is continuous function of 1/ therefore, by invariance
property of consistency eX is consistent estimator of e1/  .
E9) Since X1 , X 2 , ..., X n is a random sample taken from a population
having mean  and variance σ2.
Therefore,
E  X i    and Var  X i   2 for all i  1, 2,..., n

Consider,

 1 n 
E  T   E   Xi 
 n  1 i 1 
1
 E  X1  X 2  ...  X n 
n 1

 E(X1 )  E(X 2 )  ...  E(Xn )   aE  X   bE Y 

1  E aX  bY

n 1  

31
Estimation
1  
     ...   
n  1  
n  times


1
  n   
n 1
Therefore, T is biased estimator of population mean .
For efficiency, we find variances of estimator T  and sample mean X as

 1 
Var(T)  Var  (X1  X2  ...  Xn )
 n 1 
If X and Y are two 1
independent random  Var(X1 )  Var(X2 )  ...  Var(Xn )
(n  1)2
variables and a & b are
two constants then
1  2 
Var(aX + bY)    2  ...  2
2   
= a 2 Var(X) + b 2 Var(Y) (n  1)  n  times 

1 n2

(n  1) 2
 n 2
 
(n  1)2
Now consider,
1 
Var(X)  Var  (X1  X2  ...  X n ) 
 n 
1
  Var(X1 )  Var(X 2 )  ...  Var(X n )
n2
1  2 
    2  ...  2
2   
n  n  times 
1 2

n2
  n
n 2


Since, Var(T  ) < Var(X) , therefore, T is more efficient estimator than

sample mean.
E10) Since X1 , X 2 , ..., X n are independent taken from normal population
with mean  and variance σ2 therefore,
E  Xi    and Var  X i   2 for all i  1, 2, 3
To check estimators T1 and T2 are unbiased, we find expectations of T1
and T2 as
If X and Y are two  X  X 2  X3 
E  T1   E  1 
independent random  3 
variables and a & b are
two constants then 1
  E  X1   E  X 2   E  X 3  
E(aX + bY) 3
= aE(X) + bE(Y) and
Var(aX + bY) 1
= a 2 Var(X) + b 2 Var(Y)
E  T1        
3
32
 1 Introduction to Estimation
  3   

Now consider,
 X  X2 X3 
E  T2   E  1  
 4 2 
1 1
  E  X1   E  X 2    E  X 3 
4 2
1   
       
4 2 2 2
ET2   µ
Hence, T 1 and T2 both are unbiased estimators of .
For efficiency, we find the variances of T1 and T2 as
 X  X 2  X3 
Var  T1   Var  1 
 3 
1
  Var  X1   Var  X 2   Var  X 3  
9
1 2 3σ 2

9

σ  σ2  σ2 
9

σ2
VarT1  
3
Now consider,
 X  X 2 X3 
Var  T2   Var  1  
 4 2 
1 1
  Var  X1   Var  X 2    Var  X 3 
16 4
1 2 1
     2    2
16 4
σ 2 σ 2 σ 2  2σ 2
  
8 4 8
3σ 2

8
Since Var  T1   Var  T2  therefore, T1 is more efficient estimator of 
than T2.
E11) Refer Sub-sections 5.6.1 and 5.6.2.
E12) Here, we take random sample from exp () whose probability density
function is given by
f  x,    e x ; x 0 &  0

33
Estimation The joint density function of X1 , X 2 , ..., X n can be obtained and can be
factored as
f  x1 , x 2 ,...x n ,   f  x1 ,  .f  x 2 ,  ...  x n , 

 e  x1 .  e  x 2 ...  e  x n
n

n
  xi
 e i 1

n
   x i 
  n e i1 .1
 
 
L    g  t  x  ,  .h  x1, x 2 ,..., x n 
n
  xi
where, g  t  x  ,    n e i 1
is a function of parameter  and sample
n
values x1 , x 2 ,..., x n only through t  x    x i and h  x1 , x 2 ,..., x n   1 ,is
i 1
n
independent of . Hence by factorization theorem of sufficiency,  Xi
i 1

is sufficient estimator of .
E13) Here, we take random sample from N(, σ2) whose probability density
function is
1
1   x   2
f  x, ,  2   e 2 2
;    x  ,      ,   0
2 2
The joint density function of X1 , X 2 , ..., X n can be obtained as

f  x1 ,x2 ,...x n , , 2   f  x1,, 2  .f  x 2 , , 2  ... f  x n , , 2 

1 1 1
1   x1  2 1   x2  2 1   x n  2
2 2 2 2 2 2
 e . e ... e
2  2 2  2 2  2
n
n 1
 1   2σ2    xi  x    x  µ   2 n
n 1
  e i 1
 1   2 2  xi  
 2πσ 2    e i 1
2
 2 
… (1)
n
n 1
1  2   xi  x  x  µ  2
  2σ i 1
  e 2
 add and subtract x 
 2πσ 

n
 x  x  n 1 2 2
  x    2  x i  x  x  
 1   2 2   i 
  e i 1
2
 2  
n n n
n 1  2 2 
 1   2 2   x i  x     x    2 x     xi  x 

 e i 1 i 1 i 1

2 
 2  

34
 n 1 
n
 x i  x 2  n  x  2  0 
 Introduction to Estimation
 1   2 2   n 
   e
i 1 
  x i  x   0, by 
2  i 1
 2    the 
 property of mean 
n
n 1 2 n 2
 1   2 2   x i x   2  x  
f  x1 , x 2 ,...x n , , 2    e i 1 2
… (2)
2 
 2 
Case I: Sufficient statistic for  when σ2 is known
The joint density function given in equation (2) can be factored as
n
n 1
  n
 x  2  1    2 2
 x i  x 
2

22
f  x 1 , x 2 , ...x n ,     e .  .e i1

  2  2  

 gt x , µ .hx1, x 2 ,..., x n 

n n
 1 
  x  2
2 2
where, g  t  x  ,    e . is a function of parameter 
2 
 2  
and sample values x1 , x 2 ,..., x n only through t  x   x , whereas
n
1

2 2
  x i  x 2
h  x1 , x 2 ,..., x n   e i 1
is independent of . Hence, X is
sufficient estimator for  when σ2 is known.
Case II: Sufficient statistic for σ2 when  is known
The joint density function given in equation (1) can be factored as
n
n 1
 1   22
  x i   2
f  x 1 , x 2 , ...x n ,  2     e
i 1

 2  2 
n
n 1 2
 1   2 2
  x i  
  e
i 1
.1
 2  2 
 gtx , .hx1 , x 2 ,..., x n 
n
n 1 2
 1   2 2   x i  
where, g  t  x  ,      e i 1
is a function of parameter
2
 2  
n
2
σ2 and sample values x1 , x 2 ,..., x n only through t  x     x i    ,
i 1
2
whereas h  x1 , x 2 ,..., x n   1 is independent of σ . Hence by
n
2
factorization theorem of sufficiency,   X i    is sufficient estimator
i 1
2
for σ when  is known.
Case III: When both  and σ2 are unknown
The joint density function given in equation (2) can be factored as
1 
 1   n 1s2  n  x  2  
f  x1 , x 2 ,...x n , ,    
2
2
e 2 2 
 .1
 2 
35
Estimation 1 n 2
where, s2    xi   
n  1 i1

f  x1 , x 2 ,...x n , , 2   g  t1  x  , t 2  x  , , 2  .h  x1,x 2 ,..., x n 

where, g  t1  x  , t 2  x  , , 2  is a function of (,σ2) and sample values

x1 , x 2 ,..., x n only through t1  x   x, t 2  x   s 2 , whereas
h  x1 , x 2 , ..., x n  is independent of (, σ2). Hence by factorization
theorem, X and S2 are jointly sufficient for  and σ2.

Note 5: Here, it is remembered that X is not sufficient statistic for  if

σ2 is unknown and S2 is not sufficient for σ2 if  is unknown.
E14) The sample is taken from U [0,] whose probability density function is
1
f x , θ   ;0xθ θ0
θ
The joint density function of X1 , X 2 , ..., X n can be obtained as

f  x1 , x 2 ,...x n ,   f  x1 ,   .f  x 2 ,  ... f  x n ,  

1 1 1 1
 . ...  n
θ θ θ θ
Since range of variable depends upon the parameter , so we consider
ordered statistics X 1  X  2   ...  X  n  .

Therefore, the joint density function can be factored as

1
f  x1 , x 2 ,...x n ,    ; 0  x 1  x 2   ...  x  n   
n
1 
  n Ix n  , .1
 
where,
 1 ; if x n   θ
Ix n  , θ   
 0 ; otherwise
Therefore,
f  x1 , x 2 ,...x n ,   g  t  x  ,  .h  x1 , x 2 ,..., x n 

where, g  t  x  ,  is a function of  and sample values only through

t(x)  x  n  , whereas h  x1, x 2 ,..., x n  is independent of .

Hence, by factorization theorem X  n  is sufficient statistic for .

36
UNIT 6 POINT ESTIMATION
Structure
6.1 Introduction
Objectives
6.2 Point Estimation
Methods of Point Estimation
6.3 Method of Maximum Likelihood
Properties of Maximum Likelihood Estimators
6.4 Method of Moments
Properties of Moment Estimators
Drawbacks of Moment Estimators
6.5 Method of Least Squares
Properties of Least Squares Estimators
6.6 Summary
6.7 Solutions / Answers

6.1 INTRODUCTION
In previous unit, we have discussed some important properties of an estimator
such as unbiasedness, consistency, efficiency, sufficiency, etc. And according
to Prof. Ronald A. Fisher, if an estimator possess these properties then it is said
to be a good estimator. Now, our point is to search such estimators which
possess as many of these properties as possible. In this unit, we shall discuss
some frequently used methods of finding point estimate such as method of
maximum likelihood, method of moments and method of least squares.
This unit is divided into seven sections. Section 6.1 is introductory in nature.
The point estimation and frequently used methods of point estimation are
explored in Section 6.2. The most important method of point estimation i.e.
method of maximum likelihood and the properties of its estimators are
described in Section 6.3. The method of moments with properties and
drawbacks of moment estimators are described in Section 6.4. Section 6.5 is
devoted to the method of least squares and its properties. Unit ends by
providing summary of what we have discussed in this unit in Section 6.6 and
solution of exercises in Section 6.7.
Objectives
After going through this unit, you should be able to:
 define and obtain the point estimation;
 define and obtain the likelihood function;
 explore the different methods of point estimation;
 explain the method of maximum likelihood;
 describe the properties of maximum likelihood estimators;
 discuss the method of moments;
 describe the properties of moment estimators;
 explain the method of least squares; and
 explore the properties of least squares estimators.
37
Estimation
6.2 POINT ESTIMATION
There are so many situations in our day to day life where we need to estimate
The technique of the some unknown parameter(s) of the population on the basis on the sample
estimating the unknown observations. For example, a house wife may want to estimate the monthly
parameter with a single expenditure, a sweet shopkeeper may want to estimate the sale of sweets on a
value is known as point day, a student may want to estimates the study hours for reading of a particular
estimation. unit of this course, etc. This need is fulfilled by the technique of estimation. So
the technique of finding an estimator to produce an estimate of the unknown
parameter is called estimation.
We have already said that estimation is broadly divided into two categories
namely:
 Point estimation and
 Interval estimation
If, we find a single value with the help of sample observations which is taken
as the estimated value of unknown parameter then this value is known as point
estimate and the technique of estimating the unknown parameter with a single
value is known as “point estimation”.
If instead of finding a single value to estimate the unknown parameter if we
find two values between which the parameter may be considered to lie with
certain probability(confidence) is known as interval estimate of the parameter
and this technique of estimating is known as “interval estimation”. For
example, if we estimate the average weight of men living in a colony on the
basis of sample mean, say, 62 kg then 62 kg is called point estimate of average
weight of men in the colony and this procedure is called as point estimation. If
we estimate the average weight of men by an interval, say, [40,110] with 90%
confidence that true value of the weight lie in this interval then this interval is
called interval estimate and this procedure is called as interval estimation.
Now, the question may arise in your mind that “how point and interval
estimates are obtained?” So we will describe some of the important and
frequently used methods of point estimation in the subsequent sections of this
unit and methods of interval estimation in the next unit.

6.2.1 Methods of Point Estimation

Some of the important and frequently used methods of point estimation are:
1. Method of maximum likelihood
2. Method of moments
3. Method of least squares
4. Method of minimum chi-square
5. Method of minimum variance
The method of maximum likelihood, method of moments and method of least
squares will be discussed in detail in subsequent sections one by one and other
methods are beyond the of scope of this course.
Now, try the following exercises.
E1) Find which technique of estimation (point estimation or interval
estimation) is used in each case given below:
38
 (i) An investigator estimates average income Rs. 1.5 lack per annum of Point Estimation
the people living in a particular geographical area, on the basis of a
sample of 50 people taken from that geographical area.
(ii) A product manager of a company estimates the average life of
electric bulbs in the range 800 hours and 1000 hours, with certain
confidence, on the basis of a sample of 20 bulbs.
(iii) A pathologist estimates the mean time required to complete a
certain analysis in the range 30 minutes to 45 minutes, with certain
confidence, on the basis of a random sample of 25.
E2) List any three methods of point estimation.

6.3 METHOD OF MAXIMUM LIKELIHOOD

For describing the method of maximum likelihood, first we have to define
likelihood function.
Likelihood Function
If X 1 , X 2 , ..., X n is a random sample of size n taken from a population with
joint probability density (mass) function f  x1 , x1 ,..., x n ,  of sample values
then likelihood function is denoted by L(θ) and is defined as follows:
L    f  x1 , x1 ,..., x n , 
For discrete case,
L    P  X1  x1  P  X 2  x 2  ...P  X n  x n 
For continuous case,
L    f  x1 ,  .f  x 2 ,  ... f  x n , 
The main difference between the joint probability density (mass) function and
the likelihood function is that in joint probability density (mass) function we
consider the X's as variables and the parameter θ as fixed and consider as a
function of sample observations while in the likelihood function we consider
the parameter θ as the variable and the X's as fixed and consider as a function
of parameter θ.
The process of finding the likelihood function is described by taking an
example.
If X 1 , X 2 , ..., X n is a random sample of size n taken from exponential
distribution (θ) whose pdf is given by
f x,   e  x ; x  0,   0
Then the likelihood function of parameter θ can be obtained as
L    f  x1 ,  .f  x 2 ,  ... f  x n , 

 e  x1 . e  x 2 ... e  x n

1 ...
1 1
 n  times
e   x1  x 2 ... x n 
n

n
  xi
L     e i 1

39
Estimation The likelihood principal states that all the information in a sample to draw the
inference about the value of unknown parameter θ is found in the
corresponding likelihood function. Therefore, the likelihood function gives the
relative likelihoods for different values of the parameters, given the sample
data.
From theoretical point of view, one of the most important methods of point
estimation is method of maximum likelihood because it generally gives very
good estimators as judged from various criteria. It was initially given by Prof.
C.F. Gauss but later on it was used as a general method of estimation by Prof.
Ronald A. Fisher in 1912. The principal of maximum likelihood estimation is
to find /estimate /choose the value of unknown parameter which would most
likely generate the observed data. We know that the likelihood function gives
the relative likelihoods for different values of the parameters for the observed
data. Therefore, we search the value of unknown parameter for which the
likelihood function is maximum corresponding to the observed data. The
concept of maximum likelihood estimation is explained with a simple example
given below:
Suppose, we toss a coin 5 times and we observe 3 heads and 2 tails. Instead of
assuming that the probability of getting head is p = 0.5, we want to find /
estimate the value of p that makes the observed data most likely. Since number
of heads follows the binomial distribution, therefore, the probability (likelihood
function) of getting 3 heads in 5 tosses is given by
2
P  X  3  5 C 3 (p)3 1  p 
Imagine that p was 0.1 then
2
P  X  3  5 C 3 (0.1) 3  0.1  0.0081
Similarly, for different values of p the probability of getting 3 heads in 5 tosses
is given in Table 6.1 given below:
Table 6.1: Probability/Likelihood Function Corresponding to Different Values of p

S. No. p Probability/
Likelihood Function
1 0.1 0.0081
2 0.2 0.0512
3 0.3 0.1323
4 0.4 0.2304
5 0.5 0.3125
6 0.6 0.3456
7 0.7 0.3087
8 0.8 0.2048
9 0.9 0.0729

From Table 6.1, we can conclude that p is more likely to be 0.6 because at
p = 0.6 the probability is maximum or the likelihood function is maximum.
Therefore, principle of maximum likelihood (ML) consists in finding an
estimate for the unknown parameter  within the admissible range of θ, i.e.
within parameter space , which makes the likelihood function as large as
possible, that is, maximize the likelihood function. Such an estimate is known
as maximum likelihood estimate for unknown parameter . Thus, if there exists
40
an estimate, say, θ̂x 1 , x 2 ,..., x n  of the sample values which maximizes Point Estimation

likelihood function L() is known as “maximum likelihood estimate”. That

is, ^ - is used to represent

L ˆ  sup L   the estimate / estimator
 of parameter and read
For maximising the likelihood function, the theory of maxima or minima as cap i.e.
(discussed in Unit 6 of MST-002) is applied i.e. we differentiate L() or L ̂ - means it is the
partially with respect to the parameter  and put it equal to zero. The equation estimate of parameter θ
so obtain is known as likelihood equation, that is, and read as theta cap.


L0

Then we solve the likelihood equation for parameter  which gives the ML
estimate if its second derivative is negative at   ˆ , that is,
2
L 0
2 ˆ

Since likelihood function (L) is the product of n functions so differentiating L

as such is very difficult. Also L is always non-negative and log L remains a
finite value. So log L attains maximum when L is maximum. Hence, we can
consider log L in place of L and we find ML estimate as

 log L   0

provided,
2
 log L   0
 2 ˆ

When there are more than one parameter, say, 1, 2,…,k then ML estimates
of these parameters can be obtained as the solution of k simultaneous
likelihood equations

 log L   0; for all i  1, 2,..., k
i
provided, the matrix of derivatives

2
 log L   0; for all i  j  1, 2,..., k
i j
i ˆ i &  j ˆ j

Let us explain the procedure of ML estimation with the help of some examples.
Example 1: If the number of weekly accidents occurring on a mile stretch of a
particular road follows Poisson distribution with parameter λ then find the
maximum likelihood estimate of parameter λ on the basis of the following data:
Number of Accidents 0 1 2 3 4 5 6
Frequency 10 12 12 9 5 3 1

Solution: Here, the number of weekly accidents occurring on a mile stretch of

a particular road follows Poisson distribution with parameter λ so pmf of the
Poisson distribution is given by
41
Estimation e x
P X  x   ; x  0, 1, 2, ... &   0
x!
First we find the theoretical ML estimate of the parameter λ as:
Let X1 , X 2 , ..., X n be a random sample of size n taken from this Poisson
distribution, therefore, likelihood function of parameter λ can be obtained as
L     L  P  X  x1 . P  X  x 2  ... P  X  x n 

e x1 e x2 e xn

 . ...
x1 ! x2 ! xn !
 ... 
 
e  x1 x 2  ... x n
n  times


x1 !x 2 ! ...x n !
n

xi
e  n i 1
 n
… (1)
 xi !
i1

Taking log on both sides

n n
log L   n   x i log   log  x i ! … (2)
i 1
i 1

Differentiating equation (2) partially with respect to λ, we get

n
 1
 log L    n   x i    0 … (3)
 i 1 
For maxima or minima, that is, for finding ML estimate, we put

 log L   0

1 n
 n   xi  0
 i 1
1 n
 ˆ   xi  x
n i 1
Now, we obtain the second derivative, that is, differentiating equation (3)
partially with respect to λ, we get
2 n
 1
2 
log L  0   xi   2 
 i 1   
Put   x , we have
2 n
 1 
 log L     xi  2   0 for all values of xi’s
 2  x i 1 x 
Therefore, maximum likelihood estimate of parameter λ of Poisson distribution
is sample mean.
Since we observed that the sample mean is the maximum likelihood estimate of
parameter λ so we calculate the sample mean of the given data as
42
 S. No. Number of Accidents(X) Frequency(f) fX Point Estimation
1 0 10 0
2 1 12 12
3 2 12 24
4 3 9 27
5 4 5 20
6 5 3 15
7 6 1 6

N = 52 fX  104
The formula for calculating mean is

1
X
N
 fX where, N is the total number of accidents
1
  104  2
52
Hence, maximum likelihood estimate of λ is 2.
Example 2: For random sampling from normal population N(, σ2), find the
maximum likelihood estimators for  and σ2.
Solution: Let X 1 , X 2 , ..., X n be a random sample of size n taken from normal
population N(, σ2), whose probability density function is given by
1
1   x   2
f  x, , 2   e 2 2
;    x  ,       ,   0
22
Therefore, the likelihood function for parameters  and σ2 can be obtained as
L  , 2   L  f  x1, , 2  .f  x 2 , , 2  ... f  x n , , 2 
1 1 1
1   x1   2 1   x 2   2 1   x n   2
2 2 2 2 2 2
 e . e ... e
22 22 22
n
n/2 1
 1     xi   2
2 2 i 1
 2 
e … (4)
 2  
Taking log on both sides of equation (4), we get
n
n n n 1 2
log L  log1  log  2    log  2   2  x i   … (5)
2 2 2 2 i 1

Differentiating equation (5) partially with respect to  and σ2 respectively,

we get
n
 1
 log L    0  0  0  2 2x i    1 … (6)
 2 i 1

n
 n 1 1 2
 log L    0  0   ( 1)  xi   … (7)
2 2 2 2( 2 )2 i 1

For finding ML estimate of , we put

43
Estimation 
 log L  0

n
1

2σ 2
 2x
i 1
i  µ  1  0

n
  x i  µ   0
i1

n
1 n
  x i  nµ  0  ˆ   xi  x
i 1 n i 1
Thus, the ML estimate for  is the observed sample mean x .
For ML estimate of σ2, we put

log L  0
σ 2
n
n 1
 x  µ  0
2
 2
 ( 1) i
2σ 2(σ 2 ) 2 i 1

n
 nσ 2   x i  µ 
2

i 1
 0
2σ 4
n
  nσ 2   x i  µ   0
2

i1

1 n 2
 ˆ 2    x i  ˆ 
n i 1
1 n 2
 ˆ 2    x i  x   s2
n i 1
Thus, the ML estimates for  and σ2 are x and s2 respectively.
Hence, ML estimators for  and σ2 are X and S2 respectively.
Note 1: Since throughout the course we are using capital letters for estimators
therefore in the last line of above example we use capital letters for ML
estimators for  and σ2.
Note 2: Here, the maxima and minima method is used to obtain the ML
estimates when the range of random variable is independent of parameter .
Whereas when the range of random variable is involved or depends on
parameter  then this method fails to find the ML estimates. In such cases, we
use order statistics to maximize the likelihood function. Following example
will explain this concept.
Example 3: Obtain the ML estimators for  and  for the uniform or
rectangular population whose pdf is given by
 1
 ; α x β
f x, α, β    β  α
 0 ; elsewhere

Solution: Let X 1 , X 2 , ..., X n be a random sample of size n taken from uniform

population U(, ). Therefore, likelihood function can be obtained as
44
 L  f  x1, ,   .f  x 2 , ,   ... f  x n , ,   Point Estimation

 1  1   1 
 ...
             

n  times

n
 1 
 
 
 log L  n log β  α  … (8)
Differentiating equation (8) partially with respect to α and β respectively, we
get likelihood equations for  and  as
 n
 log L   0  0
 
and
 n
 log L   0  0
 
Both the equations give an inadmissible solution for α & β as  =  = . So the
method of differentiation fails. Thus, we have to use another method to obtain The order statistics of
the desired result. a random sample
In such situations, we use basic principal of maximum likelihood, that is, we X1 , X 2 , ..., X n are the
choose the value of the parameters α & β which maximize the likelihood sample values placed
function. If x 1  x 2   ...  x n  is an ascending ordered arrangement of the in ascending order of
magnitude. These are
observed sample, then   x 1  x 2   ...  x  n   . Also, it can be seen
denoted by
  x  n  and   x 1 .   x  n  means, β takes values greater than or equal to x  n  X 1  X  2   ...  X  n 

and least value of β is x n . Similarly,   x 1 means, α takes values less than or
equal to x 1 and maximum value of α is x 1 . Now, likelihood function will be
maximum when α is maximum and β is minimum. Thus, the minimum possible
value of  consistent with the sample is x(n) and the maximum possible value of
 consistent with the sample is x(1). Hence, L is maximum if  = x(n) and
 = x(1).
Thus, ML estimates for  and  are given by
ˆ  x 1  Smallest sample observation
and
ˆ  x  n   Largest sample observation

Hence, ML estimators for  and  are X 1 and X  n  respectively.

6.3.1 Properties of Maximum Likelihood Estimators

The following are the properties of maximum likelihood estimators:
1. A ML estimator is not necessarily unique.
2. A ML estimator is not necessarily unbiased.
3. A ML estimator may not be consistent in rare case.
4. If a sufficient statistic exists, it is a function of the ML estimators.
45
Estimation 5. If T = t(X1, X2, …,Xn) is a ML estimator of  and  () is a one to one
function of , then (T) is a ML estimator of (). This is known as
invariance property of ML estimator.
6. When ML estimator exists, then it is most efficient in the group of such
estimators.
It is now time for you to try the following exercises to make sure that you get
the concept of ML estimators.

E3) Prove that for the binomial population with density function
P  X  x   n C x p x q n x ; x  1, 2, ..., n ,q  1 − p

the maximum likelihood estimator for p is X / n .

E4) Obtain the ML estimate of  for the following distribution
1
f x, θ   ; 0  x  θ, θ  0
θ
If the sample values are1.5, 1.0, 0.7, 2.2, 1.3 and 1.2.
E5) List any five properties of maximum likelihood estimators.

6.4 METHOD OF MOMENTS

The method of moments is the oldest but simple method for determining the
point estimate of the unknown parameters. It was discovered by Karl Pearson
in 1894. The application of this method was invariable continued until Prof.
Ronald A. Fisher introduced maximum likelihood estimation method. The
principle of this method consists of equating the sample moments to the
corresponding moments of the population, which are the function of unknown
population parameter(s). We equate as many sample moments as there are
unknown parameters and solve these simultaneous equations for estimating
unknown parameter(s). This method of obtaining the estimate(s) of unknown
parameter(s) is called “Method of Moments”.
Let X1 , X 2 , ..., X n be a random sample of size n taken from a population whose
probability density (mass) function is f(x, ) with k unknown parameters, say,
1, 2,…,k. Then the rth sample moment about origin is
1 n r
Mr   Xi
n i 1

and about mean is

1 n r
Mr  
n i1
 Xi  X 

While the rth population moment about origin is

r
r  E  X

and about mean is

r
r  E  X   

46
Generally, the first moment about origin (zero) and rest central moments (about Point Estimation
mean) are equated to the corresponding sample moments. Thus, the equations
are
1  M1
r  M r ; r  2, 3, ..., k

By solving these k equations for unknown parameters, we get the moment

estimators.
Let us explain the concept of method of moments with the help of some
examples.
Example 4: Find the estimator for λ by the method of moments for the
exponential distribution whose probability density function is given by
1
f x , λ   e x / λ ; x  0, λ  0
λ
Solution: Let X 1 , X 2 , ..., X n be a random sample of size n taken from
exponential distribution whose probability density function is given by
1 x / 
f  x,   e ; x  0,   0

We know that the first moment about origin, that is, population mean of
exponential distribution with parameter  is
µ1  λ
and the corresponding sample moment about origin is
1 n
M1   Xi  X
n i 1
Therefore, by the method of moments, we equate population moment with
corresponding sample moment. Thus,
1  M1

1 n
ˆ   X i  X
n i 1
Hence, moment estimator for  is X.
Example 5: If X 1 , X 2 , ..., X m is a random sample taken from binomial
distribution (n, p) where, n and p are unknown, obtain moment estimators for
both n and p.
Solution: We know that the mean and variance of binomial distribution (n, p)
are given by
µ1  np and µ 2  npq
Also the corresponding first sample moment about origin & second sample
moment about mean (central moment) are
1 m 1 m 2
M 1  
m i 1
X i  X and M 2  
m i 1
 Xi  X 

Therefore, by the method of moments, we equate population moments with

corresponding sample moments. Thus,
47
Estimation 1  M1
1 m
np   Xi  X
m i 1
… (9)

and
2  M2
1 m 2
npq  
m i 1
 X i  X   S2 … (10)

We solve above equations (9) & (10) for n and p by dividing equation (10) by
equation (9), we get
S2
q̂ 
X
Since p = 1-q, therefore the estimator of p is
S2 X  S2
pˆ  1  qˆ  1  
X X
Put the value of p in equation (9), we get
 X  S2 
n X
 X 
X2
n̂ 
X  S2
X  S2 X2
Hence, moment estimators for p and n are and respectively
X X  S2
where,
1 m 1 m 2
X  X i and S2    X i  X  .
m i 1 m i 1
Example 6: Show that moment estimator and maximum likelihood estimator
are same of the parameter  of the geometric distribution G() whose pmf is
x
P  X  x   1   ;   0, x  0, 1, 2,...

Solution: Let X1 , X 2 , ..., X n be a random sample of size n taken from G()

whose probability mass function is given by
x
P  X  x    1    ; x  0, 1, 2, ...

Here, we first find the maximum likelihood estimator of . The likelihood

function for parameter θ can be obtained as
L     P  X  x1  . P  X  x 2  ... P  X  x n 

 θ1  θ  1 .θ1  θ  2 ...θ1  θ 

x x xn

 n 1  
xi
i 1

Taking log on both sides, we get

n
log L  n log θ   x i log 1  θ  … (11)
i1

48
Differentiating equation (11) partially with respect to  and equating to zero, Point Estimation
we get
 n n 1
 log L     x i .  1  0
  i1 1  
n n

n i1  xi
n i 1
xi
  0 
θ 1 θ θ 1 θ
n n
 n 1  θ   θ x i  n  nθ  θ  x i  0
i 1 i 1

 n 
 θ  x i  n   n
 i1 
n 1
 θ̂  n

x 1
x i 1
i n

Also, it can be seen that the second derivative, i.e.

2
 log L  0
 2 
1
x 1

1
Therefore, the ML estimator of  is .
X 1
Now, we find the moment estimator of .
We know that the first moment about origin, that is, mean of geometric
distribution is
1 θ
µ1 
θ
and the corresponding sample moment is
1 n
M1   Xi  X
n i 1
Therefore, by the method of moments, we have
1  M1

1 
 X

  X  1    X    1
1
   X  1  1  ˆ 
 X  1
1
Thus, moment estimator of θ is .
X 1
Hence, the maximum likelihood estimator and moment estimator both are same
in case of geometric distribution.
49
Estimation 6.4.1 Properties of Moment Estimators
The following are the properties of moment estimators:
1. The moment estimators can be obtained easily.
2. The moment estimators are not necessarily unbiased.
3. The moment estimators are consistent because by the law of large numbers
a sample moment (raw or central) is a consistent estimator for the
corresponding population moment.
4. The moment estimators are generally less efficient than maximum
likelihood estimators.
5. The moment estimators are asymptotically normally distributed.
6. The moment estimators may not be function of sufficient statistics.
7. The moment estimators are not unique.
6.4.2 Drawbacks of Moment Estimators
The following are the drawbacks of moment estimators:
1. This method is based on equating population moments with sample
moments. But in some situations, like as cauchy distribution, the population
moment does not exist therefore in such situations this method cannot be
used.
2. This method does not, in general, give estimators with all the desirable
properties of a good estimator.
3. The property of efficiency is not possessed by these estimators.
4. The moment estimators are not unbiased in general.
5. Generally, the moment estimators and the maximum likelihood estimators
are identical. But if they do differ, then ML estimates are usually preferred.
Now, try to solve the following exercises to ensure that you have understood
method of moments properly.
E6) Obtain the estimator of parameter  when sample is taken from a
Poisson population by the method of moments.
E7) Obtain the moment estimators of the parameters  and  2 when the
sample is drawn from normal population.
E8) Describe the properties and drawbacks of moment estimators.

6.5 METHOD OF LEAST SQUARES

The idea of least squares estimation emerges from the concept of method of
maximum likelihood. Consider the maximum likelihood of the parameter 
when σ2 is known on the basis of a random sample Y1, Y2,…,Yn of size n taken
from a normal population (, σ2). The density function of normal population is
given by
1
1   y  2
f  y, , 2   e 2 2
;    y  ,       ,   0
22
Then likelihood function for  and σ2 is
50
 L  , 2   L  f  y1, , 2  .f  y 2 , , 2  ...f  y n , , 2  Point Estimation

n
1
 1 
n/2 
2 σ2
  y i µ  2
i1
 2 
e
 2 πσ 
Taking log on both sides, we have
n 1 n 2
log L   log  22   2   y i   
2 2 i1
By principle of maximum likelihood estimation, we have to maximize log L
n
with respect to , and log L is maximum when  yi  µ  is minimum, i.e. sum
2

i1
n
of squares  yi  µ  must be least.
2

i1

The method of least squares is mostly used to estimate the parameter of linear
function. Now, suppose that the population mean  is itself a linear function of
parameters 1, 2, …, k, that is,
  x11  x 22  ...  x k k
k
  x ii
j1

where, xi’s are not random variables but known constant coefficients of
unknown parameter i’s for forming a linear function of i’s.
We have to minimize
k k 2
 
E    y i   x i θ i  with respect to i.
i 1  i 1 
Hence, method of least squares gets its name from the minimization of a sum
of squares. The principle of least squares states that we choose the values of
unknown population parameters 1, 2,…, k, say, θ̂1 , θ̂ 2 , ..., θ̂ k on the basis of
observed sample observations y1, y2,…,yn which minimize the sum of squares
k k 2
 
of deviations   y i   x i i  .
i 1  i 1 
Note: The method of least squares has already been discussed in Unit 5 of
MSL-002 and further application of this method in estimating the parameters of
regression models is discussed in specialisation courses.
6.5.1 Properties of Least Squares Estimators
Least squares estimators are not so popular. They possess some properties
which are as follows:
1. Least squares estimators are unbiased in case of linear models.
2. Least squares estimators are minimum variance unbiased estimators
(MVUE) in case of linear models.
Now, try the one exercise.

E9) Describe the two properties of least squares estimators.

We now end this unit by giving a summary of what we have covered in it.
51
Estimation
6.6 SUMMARY
After studying this unit, you must have learnt about:
1. The point estimation.
2. Different methods of finding point estimators.
3. The method of maximum likelihood and its properties.
4. The method of moments, its properties and drawbacks.
5. The method of least squares and its properties.

6.7 SOLUTIONS /ANSWERS

E1)
(i) Here, investigator estimates the average income Rs. 1.5 lack i.e. he /
she estimate average income with a single value, therefore, the
investigator used point estimation technique.
(ii) The product manager estimates the average life of electric bulbs
with the help of two values 800 hours and 1000 hours, therefore, the
product manager used interval estimation technique.
(iii) A pathologist estimates the mean time required completing a certain
analysis with the help of two values 30 minutes and 45 minutes,
therefore, he/she used interval estimation technique.
E2) Refer Sub-section 6.2.1.
E3) Let X 1 , X 2 , ..., X m be a random sample of size m taken from B(n, p)
whose pmf is given by
P  X  x   n C x p x q n  x ; x  0,1,..., n & q  1  p

The likelihood function for p can be obtained as

L  p   L  P  X  x1  . P  X  x 2  ... P  X  x m 

 n C x1 p x1 q n x1 . n C x2 p x2 q n x2 ... n C xm p xm q n x m

m m
m  xi   n  x i 
  n C xi   p i 1 q i1
i 1

Taking log on both sides, we have

m m m
log L  log   n

C x i   x i log p    n  x i  log(1  p)  q  1  p 
i 1 i 1 i 1
Differentiate partially with respect to p and equating to zero, we have
m
 1 m  1 
 log L   0   xi      n  x i     1  0
p i1  p  i 1 1 p 
m m

 xi
i1
 n  x 
i 1
i
 
p 1 p
52
 m m Point Estimation
x
i 1
i nm   x i
i 1
 
p 1 p
m m m
  x i  p x i  nmp  p x i
i 1 i1 i1

x i
x  1 m 
 p̂  i 1
nm

n  x   xi
m i 1 

Also, it can be seen that the second derivative, i.e.
2
 log L   0
p 2 p
x
n

Hence, ML estimator for parameter p is X / n.

E4) Let X1, X2, …,Xn be a random sample taken from given U(0, )
distribution. The likelihood function for θ is given by
L    L  f  x1,  .f  x 2 ,  ... f  x n , 
n
 1  1   1   1 
     ...     
        
Taking log on both sides, we have
logL   n log θ … (12)
Differentiate equation (12) partially with respect to θ, we get
 n
 log L    0, has no solution for .
 
So ML estimate cannot be found by differentiation. Therefore, by the
principle of ML estimation, we choose the value of  which maximize
likelihood function. Hence, we choose  as small as possible.
If x(1), x(2), …, x(n), is an ordered sample from this population, then
0  x1  x2  ...  xn  . Also, it can be seen that θ  x  n  which
means that θ takes values greater than or equal to x  n  and minimum
value of θ is x  n . Now, likelihood function will be maximum when  is
minimum. The minimum value of  is x(n). Therefore, ML estimate of 
is the maximum observation of the sample, that is, ˆ  x n 
Here, the given random sample is 1.5, 1.0, 0.7, 2.2, 1.3 and 1.2.
Therefore, the ordered sample is 0.7 < 1.0< 1.1 < 1.3 < 1.5 < 2.2. Here,
the maximum observation of this sample is 2.2, therefore maximum
likelihood estimate of  is 2.2.
E5) Refer Section 6.3.1.
E6) Let X 1 , X 2 , ..., X n be a random sample of size n taken from Poisson
population whose probability mass function is given by

53
Estimation e  x
P[X  x]  ; x  0,1, 2, ... &   0
x!
We know that for Poisson distribution
µ1  λ
and the corresponding sample moment is
1 n
M1   Xi  X
n i 1
Therefore, by the method of moments, we equate population moment
with corresponding sample moment. Thus,
1  M1
1 n
  Xi  X
n i 1
Hence, moment estimator for  is X.
E7) Let X1 , X 2 , ..., X n be a random sample of size n taken from normal
population N(, σ2), whose probability density function is given by
1
1   x 2
f  x, , 2   e 2 2
;    x  ,      ,   0
22
We know that for N(, σ2)
µ 1  µ and µ 2  σ 2
and the corresponding sample moments are
1 n 1 n 2
M1   X i  X and M 2    X i  X 
n i 1 n i 1
Therefore by the method of moments, we equate population moments
with corresponding sample moments. Thus,
1  M1
1 n
ˆ   Xi  X
n i 1
and
2  M2
1 n 2
 ˆ 2  
n i 1
 X i  X   S2

Hence, moment estimators for  and 2 are X and S2 respectively.

E8) Refer Sub-sections 6.4.1and 6.4.2.
E9) Refer as Section 6.5.1.

54
UNIT 7 INTERVAL ESTIMATION FOR ONE
POPULATION
Structure
7.1 Introduction
Objectives
7.2 Interval Estimation
Confidence Interval and Confidence Coefficient
One-Sided Confidence Intervals
7.3 Method of Obtaining Confidence Interval
7.4 Confidence Interval for Population Mean
Confidence Interval for Population Mean when Population Variance is Known
Confidence Interval for Population Mean when Population Variance is Unknown
7.5 Confidence Interval for Population Proportion
7.6 Confidence Interval for Population Variance
Confidence Interval for Population Variance when Population Mean is known
Confidence Interval for Population Variance when Population Mean is Unknown
7.7 Confidence Interval for Non-Normal Populations
7.8 Shortest Confidence Interval
7.9 Determination of Sample Size
7.10 Summary
7.11 Solutions / Answers

7.1 INTRODUCTION
In the previous unit, we have discussed the point estimation, under which we
learn how one can obtain point estimate(s) of the unknown parameter(s) of the
population using sample observations. Everything is fine with point estimation
but it has one major drawback that it does not specify how confident we can be
that the estimated close to the true value of the parameter.
Hence, point estimate may have some possible error of the estimation and it
does not give us an idea of how these estimates deviate from the true value of
the parameter being estimated. This limitation of point estimation is over come
by the technique of interval estimation. Therefore, instead of making the
inference of estimating the true value of the parameter through a point estimate
one should make the inference of estimating the true value of parameter by a
pair of estimate values which are constituted an interval in which true value of
parameter expected to lie with certain confidence. The technique of finding
such interval is known as “Interval Estimation”.
For example, suppose that we want to estimate the average income of persons
living in a colony. If 50 persons are selected at random from that colony and
the annual average income is found to be Rs. 84240 then the statement that the
average annual income of the persons in the colony is between Rs. 80000 and
Rs. 90000 definitely more likely to be correct than the statement that the annual
average income is Rs. 84240.
This unit is divided into eleven sections. Section 7.1 is introductory in nature.
The confidence interval and confidence coefficient are defined in Section 7.2.
The general method of finding the confidence interval is explored in Section
55
Estimation 7.3. Confidence interval for population mean in different cases as population
variance is known and unknown are described in Section 7.4 whereas in
Section 7.5, the confidence interval for population proportion is explained. The
confidence interval for population variance in different cases when population
mean is known and unknown are described in Section 7.6. Section 7.7 is
devoted to explain the confidence interval for non-normal populations. The
concept of shortest confidence interval and determination of sample size are
explored in Sections 7.8 and 7.9 respectively. Unit ends by providing summary
of what we have discussed in this unit in Section 7.10 and solution of exercises
in Section 7.11.
Objectives
After studying this unit, you should be able to:
 need of interval estimation;
 define the interval estimation;
 describe the method of obtaining the confidence interval;
 obtain the confidence interval for population mean of a normal population
when population variance is known and unknown;
 obtain the confidence interval for population proportion;
 obtain the confidence interval for population variance of a normal
population when population mean is known and unknown;
 obtain the confidence intervals for population parameters of a non-normal
populations;
 explain the concept of the shortest interval; and
 determination of sample size.

7.2 INTERVAL ESTIMATION

In previous section, we introduced you with the interval estimation and after
that we can conclude that if we find two values with the help of sample
observations and constitute an interval such that it contain the true value of
parameter with certain probability then it is known as interval estimate of the
parameter. This technique of estimation is known as “Interval Estimation”.
In this section, we will formally define:
 Confidence Interval and Confidence Coefficient
 One-sided Confidence Intervals
in the following two sub-sections.
7.2.1 Confidence Interval and Confidence Coefficient
Let X1 , X 2 , ..., X n be a random sample of size n taken from a population
whose probability density (mass) function is f(x, ). Let T1 = t1( X1 , X 2 , ..., X n )
and T2 = t2( X 1 , X 2 , ..., X n ) where T1  T2  be two statistics such that the
probability that the random interval [T1, T2] includes the true value of
population parameter  is (1− α), that is,
Fig. 7.1
P  T1    T2   1  
as shown in the Fig.7.1, where,  does not depend on .
Then the random interval [T 1, T2] is known as (1-) 100% confidence interval Interval Estimation
for unknown population parameter  and (1-) is known as confidence for One Population
coefficient or confidence level. The above probability statement may be
explained with the help of an example:
Suppose we say that the probability that the random interval contains the true
value of parameter θ is 0.95 then by this statement we simple mean that if 100
samples of same size, say, ‘n’ are drawn from the given population f(x,) and
the random interval [T1, T2] is computed for each sample, then 95 times out of
100 intervals, the random interval [T1, T2] contains the true value of parameter
. Hence, higher the probability (1-), more the confident we will have that
the random interval [T1, T2] will actually includes the true value of . The
statistics T1 and T2 are known as lower and upper confidence limits or
confidence bounds or fiducial limits, respectively for . And the interval is
known as two-sided confidence interval. The length of confidence interval is
defined as
L = Upper confidence limit − Lower confidence limit
i.e. L = T2-T1
The confidence interval may also be one-sided so in the next sub-section we
define one-sided confidence intervals.
7.2.2 One-Sided Confidence Intervals
In some situations, we may be interested in finding an upper bound or lower
bound but not both for population parameter with a given confidence. For
example, one may be interested to obtain a bound such that he /she is 95%
confident that the average life of the electric bulbs of a company is no less than
one years. In such cases, we construct one-sided confidence intervals.
Let X1 , X 2 , ..., X n be a random sample of size n taken from the population
having probability density(mass) function f(x, ) and also let T1 be a statistic
such that
Fig. 7.2
P    T1   1  
as shown in Fig. 7.2, then statistic T1 is called a lower confidence bound for
parameter θ with confidence coefficient (1-) and [T1, ) is called a lower one
sided (1-) 100% confidence interval for parameter θ.
Similarly, let T2 be a statistic such that
P    T2   1  
as shown in Fig. 7.3, then statistic T2 is called an upper confidence bound for Fig. 7.3
parameter θ with confidence coefficient (1-) and (−,T2] is called a upper
one sided (1-) 100% confidence interval for parameter θ.
Note 1: Generally, one-sided confidence intervals are rarely used so we focus
on two-sided confidence interval in this course. Generally, confidence
interval means two-sided confidence interval unless or otherwise it is
stated as one-sided.
Now, you can try the following exercises.
E1) Find the length of the following confidence intervals:
(i) P   1.65    3.0   0.95 (ii) P  1.68    2.70  0.95
(iii) P  1.70    2.54   0.95 (iv) P  1.96    1.96  0.95
Estimation E2) Find the lower and upper confidence limits and also confidence
coefficient of the following confidence intervals:
(i) P  0    1.5  0.90 (ii) P  1    2  0.95
(iii) P  2    2   0.98 (iv) P  2.5    2.5  0.99

7.3 METHOD OF OBTAINING CONFIDENCE

INTERVAL
After knowing about the confidence interval, the question may arise in your
mind that “how confidence intervals are obtained?” There are following two
methods are generally used for obtaining confidence intervals:
1. Pivotal quantity method
2. Statistical method
The statistical method is beyond of scope of this course so we will keep our
focus only on pivotal-quantity method which is also known as general method
of interval estimation. Before going to describe this method, we first define the
pivotal quantity as:
Pivotal Quantity
Let X1 , X 2 , ..., X n be a random sample of size n taken from a population
having probability density (mass) function f (x, ). If quantity
Q = q (X1 , X 2 , ..., X n , ) is a function of X1 , X 2 , ..., X n and parameter  such
that its distribution does not dependent on unknown parameter  then the
quantity Q is known as a pivotal quantity.
For example, if X1 , X 2 , ..., X n is a random sample taken from normal
population with mean  and variance 4, i.e. N(, 4) where, the parameter µ is
unknown then we know that the sampling distribution of sample mean is also
normal with mean  and variance 4/n, that is, X ~ N  , 4 / n  and the sampling
X
distribution of variate Z  is N  0, 1 . Since distribution of X depends
4/n
on the parameter  to be estimated, therefore, it is not a pivotal quantity
whereas the distribution of variate Z is independent of parameter  so it is a
pivotal quantity.
Pivotal Quantity Method
The pivotal quantity method for confidence interval has following steps:
Step 1: First of all, we search the statistic for unknown parameter which can
be used to estimate the parameter, say, θ, preferably a sufficient
statistic, whose distribution is completely known. After that we find
the function based on that statistic whose distribution does not
dependent on parameter  which is to be estimated i.e. we find the
pivotal quantity Q.
Step 2: Introduce two constants, say, ‘a’ and ‘b’, depending on α but not on
unknown parameter θ, such that
P  a  Q  b  1  

58
Step 3: Since pivotal quantity is a function of parameter, therefore, we convert Interval Estimation
above interval for parameter  as for One Population

P T1    T2   1  
where, T1 and T2 are functions of sample values and a & b.
Step 4: Determine constants ‘a’ and ‘b’ by minimizing the length of the
interval
L  T2  T1
With the help of the pivotal quantity method, we will find the confidence
intervals for population mean, proportion, variance which will describe one by
one in subsequent sections.
Now, you can try the following exercise.
E3) Describe the general method of constructing confidence interval for
population parameter.

7.4 CONFIDENCE INTERVAL FOR POPULATION

MEAN
There are so many problems in real life where it becomes necessary to obtain
the confidence interval of population mean. For example, an investigator may
interested to find the interval estimate of average income of the people living in
a particular geographical area, a product manager may want to find the interval
estimate of average life of electric bulbs manufactured by a company, a
pathologist may want to obtain the interval estimate of the mean time required
to complete a certain analysis, etc.
For describing confidence interval for population mean, let X1 , X 2 , ..., X n be a
random sample of size n taken from normal population having mean  and
variance σ2. We can determine confidence interval for population mean 
under following two cases:
1. When population variance σ2 is known
2. When population variance σ2 is unknown.
These two cases are discussed one by one in subsequent Sub-sections 7.4.1 and
7.4.2 respectively.
7.4.1 Confidence Interval for Population Mean when
Population Variance is Known
Let X1 , X 2 , ..., X n be a random sample of size n taken from normal population
N(, σ2) when σ2 is known, that is, σ2 has a specify value, say,  02 .
To find out the confidence interval for population mean, first of all we search
the statistic for estimating µ whose distribution is completely known.
Generally, we use the value of statistic (sample mean) X to estimate the
population mean  and also it is a sufficient statistic for parameter θ. Therefore,
we use X to make the pivotal quantity.
We know that when parent population is normal N(, 2) then sampling
distribution of sample mean X is normally distributed with mean  and
variance σ2/n, that is, if
59
Estimation X i ~ N  ,  2 
then
X ~ N  ,  2 / n 
and the variate
X 
Z ~ N(0, 1)
/ n
follows the normal distribution with mean 0 and variance unity. Therefore, the
probability density function of Z is
1  12 z2
f z  e ; z 
2
Since distribution of Z is independent of the parameter to be estimated i.e. ,
therefore, Z can be taken as pivotal quantity. So we introduce two constants,
say, z/2 and z1-/2= -z/2(since distribution of Z is symmetrical about Z = 0
line see Fig. 7.4) such that
P  z  / 2  Z  z  / 2   1  
where, z/2 is the value of the variate Z having an area of /2 under the right
tail of the probability curve of Z as shown in Fig. 7.4.
By putting the value of Z in above equation / probability statement, we get
 X 
P  z  / 2   z/2   1  
 / n 
Now, for converting this interval for parameter , we multiply each term in
Fig. 7.4 above inequality by σ / n , we get
   
P z / 2  X    z / 2  1 
 n n 
Now, subtracting X from each term in above inequality then we get
   
P  X  z / 2     X  z  / 2  1 
 n n 
Multiplying each term by (-1) in above inequality, we get
     by multiplying  1 
P X  z / 2    X  z / 2 1 
 n n   the inequality isreversed 

This can be rewritten as

   
P X  z / 2    X  z/ 2  1 
 n n 
Hence, (1-) 100% confidence interval of population mean is given by
   
 X  z  / 2 n , X  z  / 2 n  … (1)

and corresponding limits are given by


X  z/ 2 … (2)
n
Note 2: The value of z/2 can be obtained by the method described in Unit 14 Interval Estimation
of MST-003. In interval estimation generally, we have to find the 90%, 95%, for One Population
98% and 99% confidence intervals therefore, the corresponding values of z/2
are summaries in the Table 7.1 given below and we will use these values
directly when needed:
Table 7.1: Communally Used Values of Standard Normal Variate Z

1–α 0.90 0.95 0.98 0.99

α 0.10 0.05 0.02 0.01
α/2 0.05 0.025 0.01 0.005
z/2 1.645 1.96 2.33 2.58

For example, if we want to find the 99% confidence interval (two-sided) for 
then
1    0.99    0.01
For  = 0.01, the value of z  / 2  z0.005 is 2.58 therefore, 99% confidence
interval for  is given by
   
 X  2.58 , X  2.58 
 n n
Application of the above discussion can be seen in the following example.
Example 1: The mean life of the tyres manufactured by a company follows
normal distribution with standard deviation 3200 kms. A sample of 250 tyres is
taken and it is found that the average life of the tyres is 50000 kms with a
standard deviation of 3500 kms. Establish the 99% confidence interval within
which the mean life of tyres of the company is expected to lie.
Solution: Here, we are given that
n  250,   3200, X  50000, S  3500
Since population standard deviation i.e. population variance σ2 is known,
therefore, we use (1-) 100% confidence limits for population mean when
population variance is known which are given by

X  z / 2
n
where, z/2 is the value of the variate Z having an area of /2 under the right
tail of the probability curve of Z and for 99% confidence interval, we have
1    0.99    0.01. For  = 0.01 we have, z / 2  z0.005  2.58.
Therefore, the 99% confidence limits are

X  2.58
n
By putting the values of n, X and σ, the 99% confidence limits are
3200
50000  2.58 
250
50000  522.20  49477.80 and 50522.20
Hence, 99% confidence interval within which the mean life of tyres of the
company is expected to lie is
 49477.80, 50522.20
61
Estimation 7.4.2 Confidence Interval for Population Mean when
Population Variance is Unknown
In the cases, described in previous sub-section we assume that variance σ2 of
the normal population is known but in general it is not known and in such a
situation the only alternative left is to estimate the unknown σ2. The value of
sample variance (S2) is used to estimate the σ2 where,
1 n1 2
S2  
n  1 i1
 Xi  X 

In this situation, we know that the variate

X 
t ~ t ( n 1)
S/ n
follows t-distribution (described in Unit 3 of this course) with (n−1) df.
Therefore, probability density function of statistic t is given by
n
1  t2 2
f (t )  1   ;   t 
 n 1 1  n  1 
B ,  n 1 
 2 2
Since distribution of statistic t is independent of the parameter to be estimated,
therefore, t can be taken as pivotal quantity. So we introduce two constants
t n 1,  / 2 and t  n 1, 1 / 2   t  n 1,  / 2 (since t-distribution is symmetrical about
t = 0 line see Fig. 7.5) such that
P   t  n 1,  / 2  t  t  n 1,  / 2   1   … (3)

where, t  n 1,  / 2 is the value of the variate t with n–1 df having an area of /2
under the right tail of the probability curve of variate t as shown in Fig. 7.5.
By putting the value of variate t in equation (3), we get
 X  
P   t  n 1,  / 2   t  n 1,  / 2   1  
Fig. 7.5  S/ n 
Now, for converting the above interval for parameter , we multiply each term
in above inequality by S / n then we get
 S S 
P   t  n 1,  / 2  X   t  n 1,  / 2 1 
 n n 
After subtracting X from each term in above inequality, we get
 S S 
P   X  t  n 1,  / 2     X  t  n 1,  / 2  1 
 n n 
Now, by multiplying each term by (-1) in above inequality, we get
 S S   by multiplying  1 
P  X  t  n 1,  / 2    X  t  n 1,  / 2  1 
 n n   the inequality is reversed 

This can be rewritten as

 S S 
P  X  t  n 1,  / 2    X  t  n 1,  / 2  1 
 n n 
Hence, when variance σ2 is unknown then (1-) 100% confidence interval for Interval Estimation
population mean of normal population is given by for One Population

 S S 
 X  t  n 1,  / 2 n , X  t  n 1,  / 2 n  … (4)

and corresponding limits are given by

S
X  t  n 1,  / 2 … (5)
n
Note 2: For different confidence intervals and different degrees of freedom, the
values of t n 1, α / 2 are different. Therefore, for given values of  and n we read
the tabulated value of t-statistic from the table of t-distribution (t-table) given
in Appendix (at the end of Block 1 of this course) by using the method
described in Unit 4 of this course.
For example, if we want to find the 95% confidence interval for  when n = 8
then we have
1    0.95    0.05
From t-table, for  = 0.05 and ν = n ‒1 = 7, we have the value of
t  n 1,  / 2  t  7 ,0.025  2.365.

As we have seen in t-table of the Appendix that when sample size is greater
than 30 (n > 30) then all values of variate t are not given in this table so for
convenient as we have discussed in Unit 2 of this course that when n is
sufficiently large   30  then we know that almost all the distributions are very
closely approximated by normal distribution. Thus in this case t-distribution is
also approximated normal distribution. So the variate
X 
Z ~ N(0, 1)
S/ n
also follows the normal distribution with mean 0 and variance unity. Therefore,
when population variance is unknown and sample size is large then the (1-)
100% confidence interval for population mean may be obtained by using the
same procedure as we have followed in case when σ2 is known by taking S2 in
place of σ2 which is given as
 S S 
 X  z  / 2 n , X  z  / 2 n  … (6)

and corresponding limits are given by

S
X  z / 2 … (7)
n
Following example will explain the application of the above discussion:
Example 2: It is known that the average weight of students of a Study Centre
of IGNOU follows normal distribution. To estimate the average weight, a
sample of 10 students is taken from this Study Centre and measured their
weights (in kg) which are given below:
48, 50, 62, 75, 80, 60, 70, 56, 52, 77
Compute the 95% confidence interval for the average weight of students of
Study Centre of IGNOU.
63
Estimation Solution: Since population variance is unknown, therefore, (1-) 100%
confidence limits for the average weight of students of Study Centre are given
by
S
X  t  n 1,  / 2
n
1 n 1 n 2
where, X  
n i 1
Xi ,S  
n 1 i 1
 Xi  X 

Calculation for X and S :

S. 2

No.
Weight (X)  X  X X  X
1 48 ‒15 225
2 50 ‒13 169
3 62 ‒1 1
4 75 12 144
5 80 17 289
6 60 ‒3 9
7 70 7 49
8 56 ‒7 49
9 52 ‒11 121
10 77 14 196
2
Sum  X  630  X  X   1252

From the above calculation, we have

1 1
X
n
 X   630  63
10
1 2 1
S2 
n 1
  X  X   1252  139.11
9
 S  139.11  11.79
For 95% confidence interval, we have 1    0.95    0.05. Also from
t-table, we have, t  n1,  / 2  t  9,0.025  2.306.

Thus, the 95% confidence limits are

S 11.79
X  t  n 1, 0.025  63  2.306 
n 10
 63  8.60  54.40 and 71.60
Hence, required 95% confidence interval for the average weight of students of
Study Centre of IGNOU is given by
54.4, 71.60
Example 3: The mean life of 100 electric blubs produced by a company is
2550 hours with a standard deviation 54 hours. Find 95% confidence limits for
population mean life of electric blubs produced by the company.
Solution: Here, we are given that
n  100, X  2550, S  54

64
Since population variance is unknown and sample size is large (> 30), Interval Estimation
therefore, we can use (1−α) 100 % confidence limits for population mean for One Population
which are given by
S
X  z / 2
n
where, z/2 is the value of the variate Z having an area of /2 under the right
tail of the probability curve of Z. For 95% confidence limits, we have
1    0.95    0.05 and for  = 0.05, we have z  / 2  z0.025 1.96.
Thus, the 95% confidence limits for mean life of electric bulbs are
S 54
X  1.96  2550  1.96
n 100
 2550  1.96  5.4
 2550  10.58  2539.42 and 2560.58
Now, it is time for you to try the following exercises to make sure that you
have learnt about the confidence interval for population mean in different
cases.
E4) Certain refined oil is packed in tins holding 15 kg each. The filling
machine maintains this but have a standard deviation 0.30 kg. A sample
of 200 tins is taken from the production line. If sample mean is 15.25
kg then find the 95% confidence interval for the average weight of oil
tins.
E5) Sample mean of weights (in kg) of 150 students of IGNOU is found to
be 65 kg with standard deviation 12 kg. Find the 95% confidence limits
in which the average weight of all students of IGNOU expected to lie.
E6) It is known that the average height of cadets of a centre follows normal
distribution. A sample of 6 cadets of the centre was taken and measured
their heights (in inch) which are given below:
70 72 80 82 78 80
From this data, estimate the 95% confidence limits for the average
height of cadets of the particular centre.

7.5 CONFIDENCE INERVAL FOR POPULATION

PROPORTION
In Section 7.4, we have discussed the confidence interval for population mean.
But in many real word situations, in business and other areas, the data are
collected in form of counts or the collected data classified into two categories
or groups according to an attribute or characteristic under study. Generally,
such types of data are considered in terms of proportion of elements /
individuals / units / items possess or not possess a given characteristic or
attribute. For example, the proportion of female in the population, proportion
of diabetes patients in a hospital, proportion of Science books in a library,
proportion of defective articles in a lot, etc.
In such situations, we deal population proportion instead of population mean
and one may want to obtain the confidence interval for population proportion.
65
Estimation For example, a sociologist may want to know the confidence interval for
proportion of female in the population of a state. A doctor may want to know
the confidence interval for proportion of diabetes patients in a hospital, a
product manager may want to know the confidence interval for proportion of
defective articles in a lot, etc.
Generally, population proportion is estimated by sample proportion.
Let X 1 , X 2 , ..., X n be a random sample of size n taken from a population with
population proportion P. Also let X denotes the number of observations or
elements possess a certain attribute (successes) out of n observations of the
sample then sample proportion p can be defined as
X
p  1
n
As we have seen in Section 2.4 of the Unit 2 of this course that mean and
variance of the sampling distribution of sample proportion are
PQ
E(p)  P and Var(p) 
n
where, Q = 1– P.
But sample proportion is generally considered for large sample so if sample
size is sufficiently large, such that np > 5 and nq > 5 then by central limit
theorem, the sampling distribution of sample proportion p is approximately
normally distributed with mean P and variance PQ/n. Therefore, the variate
pP
Z ~ N  0,1
P(1  P)
n
is approximately normally distributed with mean 0 and variance unity. Since
distribution of Z is independent of parameter P so it can be taken as pivotal
quantity, therefore, we introduce two constants z/2 and z(1-/2) = -z/2 such that
P z  / 2  Z  z / 2   1  
where, z/2 is the value of the variate Z having an area of /2 under the right
tail of the probability curve of Z as shown in Fig. 7.6.
Putting the value of Z, we get
Fig. 7.6  
 pP 
P z / 2   z / 2   1   … (8)
 P1  P  
 n 
For large sample, the variance P (1-P) /n can be estimated by p(1-p)/n
therefore, putting p(1-p) / n in place of P (1-P) / n in equation (8), we get
 
 pP 
P z / 2   z / 2   1  
 p1  p  
 n 

Now, for converting the above interval for parameter P, we multiplying each
p(1  p)
term by and then subtracting p from each term in the above
n
inequality, we get
 Interval Estimation
 p1  p  p1  p  
P  p  z α / 2   P  p  z α / 2   1 α for One Population
 n n 
Now, by multiplying each term by (-1) in above inequality, we get
 p1  p p1  p    by multiplying 
Pp  zα / 2  P  p  zα / 2  1 α ( 1)the inequality 
 n n   is reversed 

This can be written as

 p1  p p1  p 
Pp  zα / 2  P  p  zα / 2   1 α
 n n 

Hence, (1-) 100% confidence interval for population proportion is given by

 p1  p  p 1  p  
p  z α / 2 , p  zα / 2  … (9)
 n n 
Therefore, corresponding confidence limits are

p 1  p 
p  z/ 2 … (10)
n
Following example will explain the application of the above discussion:
Example 4: A sample of 200 voters is chosen at random from all voters in a
given city. 60% of them were in favour of a particular candidate. If large
number of voters cast their votes then find 99% and 95% confidence intervals
for the proportion of voters in favour of a particular candidate.
Solution: Here, we are given
n = 200, p  0.60
First we check the condition of normality as
 np  200  0.60  120  5 and nq  200  (1  0.60)  200  0.40  80  5 so
(1-)100% confidence limits for the proportion are given by

p 1  p 
p  z / 2
n
For 99% confidence interval, we have 1    0.99    0.01. For  = 0.01,
we have z0.005  2.58 and for  = 0.05, z0.025  1.96.
Therefore, 99% confidence limits of voters in favour of a particular candidate
are
p 1  p  0.60  0.40
p  z 0.005  0.60  2.58 
n 200
 0.60  2.58  0.03
 0.60  0.08  0.52 and 0.68
Hence, required 99% confidence interval for the proportion of voters in favour
of a particular candidate is given by
[0.52, 0.68]
Similarly, 95% confidence limits are given by
67
Estimation
p 1  p 
p  z 0.025  0.60  1.96  0.03
n
 0.60  0.06  0.54 and 0.66
Hence, 95% confidence interval for the proportion of voters in favour of a
particular candidate is given by
[0.54, 0.66]
It is your time to try the following exercise.

E7) A random sample of 400 apples was taken from a large consignment
and 80 were found to be bad. Obtain the 99% confidence limits for the
proportion of bad apples in the consignment.

7.6 CONFIDENCE INTERVAL FOR POPULATION

VARIANCE
In Sections 7.4 and 7.5, we discussed the confidence interval for population
mean and proportion respectively. But there are many practical situations
where one may be interested to obtain the interval estimate of the population
variance. For example, a manufacturer of steel ball bearings may want to
obtain the interval estimate of the variation of diameter of steel ball bearing, an
economist may wish to know the interval estimate for the variability in income
of the person living in a city, etc.
Similar to confidence interval for the population mean we can determine
confidence interval for population variance into two following cases:
1. When population mean is known and
2. When population mean is unknown.
These two cases are described one by one, in Sub-sections 7.6.1 and 7.6.2
respectively.
7.6.1 Confidence Interval for Population Variance when
Population Mean is Known
Let X1 , X 2 , ..., X n be a random sample of size n taken from normal population
having mean  and variance σ2 where  is known, that is,  has a specified
value. In this case, we know that the variate
n
2
 X i  
2  i 1
2
~ (2n )

follows the chi-square distribution with n degrees of freedom whose
probability density function is given by
n
1 2 / 2 2 2 1
f  2   n/2
e    ; 0  2 
2 n/2
Since distribution of χ 2 is independent of parameter σ2 therefore, χ 2 can be

68
taken as pivotal quantity, therefore, we introduce two constants 2 n,  / 2 Interval Estimation
for One Population
and 2n, 1/ 2 such that

P 2 n, 1 / 2  2  2 n ,  / 2   1   … (11)

where, 2n,  /2 and 2 n, 1 /2 are the value of the χ 2 variate at n df having area of
/2 under the right tail and /2 under the left tail respectively of the
probability curve of χ 2 as shown in Fig. 7.7.
Putting the value of χ2 in equation (11), we get
n
 2 
 2   Xi    
P   n , 1 / 2  i 1 2
 2 n ,  / 2   1  
  
  Fig. 7.7
n
Now, for converting this interval for σ2, we divide each term by   X i    in
i 1
above inequality then we get
 
 2 n , 1  / 2  1 2 n ,  / 2

P n  2 n   1 
 X   2 
 X i    
2

 
i 1
i 
i 1 
Reciprocal each term of the above inequality
 n 2
n
2
 i  X      Xi    
P  i 1 2   2  i 1 2   1   inequality
by reciprocaling, the 
is reversed 
  n , 1 / 2    n ,  / 2  
 
This can be written as
 n 2
n
2
 i  X      Xi    
P  i 1 2   2  i 1 2   1 
   n ,  / 2  n , 1 / 2  
 
Hence, (1-)100% confidence interval for population variance when
population mean is known in normal population is given by
 n 2
n
2
 i  X      Xi    
 i 1 2 , i 1 2  … (12)
   n ,  / 2  n , 1 / 2  
 
and the corresponding (1−α) 100% confidence limits are given by
n n
2 2
  Xi   
i 1
X  
i 1
i
and … (13)
2n ,  / 2 2n , 1 / 2 

Note 3: For different confidence interval and degrees of freedom the values of
2n, / 2 and 2n, 1/ 2 are different. Therefore, for given values of  and n we read
the tabulated value of these from the table of χ2-distribution (χ2-table) (given in
Estimation Appendix at the end of Block 1 of this curse) by the method described in Unit 4
of this course.
For example, if we want to find the 95% confidence interval for 2 then
1    0.95    0.05
From the χ 2 -table, for  = 0.05 and if n =10, we have
2 n,  /2  210, 0.025  20.48 and 2 n, 1 /2  210, 0.975  3.25
Therefore, 95% confidence interval for variance is given by
 n 2
n
2
   X i      Xi    
 i 1 , i1 
 20.48 3.25 
 
Following example will explain the application of the above discussion.
Example 5: Diameter of steel ball bearing produced by a company is known to
be normally distributed. To know the variation in the diameter of steel ball
bearings, the product manager takes a random sample of 10 ball bearings from
the lot having average diameter 5.0 cm and measures diameter (in cm) of each
selected ball bearing. The results are given below:
S. No. 1 2 3 4 5 6 7 8 9 10
Diameter 5.0 5.1 5.0 5.2 4.9 5.0 5.0 5.1 5.1 5.2

Find the 95% confidence interval for variance in the diameter of steel ball
bearings of the lot from which the sample is drawn.
Solution: Here, we are given that
n =10, µ = 5.0
Since population mean is given, therefore, we use (1-)100% confidence
interval for population variance when population mean is known which is
given by
 n 2
n
2 
   X i      Xi    
 i 1 2 , i 1 2 
   n ,  / 2  n , 1 / 2  
 
n
2
Calculation for  X
i 1
i   :

S. No. Diameter (X) X   X  

2

1 5.0 0 0
2 5.1 0.1 0.01
3 5.0 0 0
4 5.2 0.2 0.04
5 4.9 ‒0.1 0.01
6 5.0 0 0
7 5.0 0 0
8 5.1 0.1 0.01
9 5.1 0.1 0.01
10 5.2 0.2 0.04
2
Total  X    0.12

70
From the above calculation, we have Interval Estimation
for One Population
2
  X  
i  0.12

For 95% confidence interval, we have 1    0.95    0.05 then from

2-table, we have
2 n, / 2  210, 0.025  20.48 and 2 n, 1 /2  210,0.975  3.25

Thus, 95% confidence interval for variance in the diameter of steel ball
bearings of the lot is given by
 0.12 0.12 
 20.48 , 3.25 

or 0.0059, 0.0369
7.6.2 Confidence Interval for Population Variance when
Population Mean is Unknown
Let X1 , X 2 , ..., X n be a random sample of size n taken from normal population
with unknown mean  and variance σ2. In this case, the value of sample mean
X is used to estimate µ. As we have seen in Section 4.2 of Unit 4 of this
course that the variate
n
2
 X i  X
 n  1 S2 1 n 2
 2 i 1
2

2
~ (2n 1) where, S2  
n  1 i 1
 Xi  X 

follows the chi-square distribution with (n-1) degrees of freedom whose

probability density function is given by
n 1
1 2 1
f ( 2 )  ( n 1)
n 1
e /2
  2 2 ; 0  2  
2
2
2
Since distribution of 2 is independent of parameter to be estimated, therefore,
χ 2 can be taken as pivotal quantity. So we introduce two constants 2n 1,  / 2 and
2n 1, 1 / 2 such that


P χ 2n1,1α/ 2  χ 2  χ 2n1,α/ 2  1 α  … (14)

where, 2n1, / 2 and 2n1, 1 /2 are the value of the χ2-variate at (n ‒1) df having
area of /2 under the right tail and /2 under the left tail of the probability Fig. 7.8
curve of 2 as shown in Fig. 7.8.
Putting the value of χ2in equation (14), we get

P 2 n1, 1 / 2 
 n  1 S2  2 
2  n 1,  / 2   1  
  
Now, for converting this interval for σ2 we dividing each term in above
inequality by  n  1 S2 then we get
Estimation  2n1, 1 / 2 1 2n1,  / 2 
P 2
 2  1 
  n  1 S   n  1 S2 
By taking reciprocal of each term of above inequality, we get
  n  1 S2  n  1 S2   by reciprocaling, the
P 2 2
  2   1    inequality is reversed 
   n 1, 1 / 2    n 1,  / 2   

This can be written as

  n  1 S2  n  1 S2 
P 2  2  2   1 
   n 1,  / 2   n 1, 1 / 2  

Hence, the (1−) 100% confidence interval for population variance when
population mean is unknown is given by
  n  1 S2  n  1 S2 
 2 , 2  … (15)
   n 1,  / 2   n 1,  1 / 2  
and corresponding confidence limits are
 n  1 S2 and
 n  1 S2 … (16)
2n1,  / 2 2n 1, 1 / 2

where, 2n1, /2 and 2n1, 1/ 2 are the values of χ 2 -variate at (n−1) degrees of
freedom and the values of these can be read from χ 2 -table. For example, if we
want to find the 95% confidence interval for 2 then
1    0.95    0.05
From χ 2 -table for  = 0.05 and n = 10 degrees of freedom, we have
2n1,  / 2  29, 0.025  19.02 and 29,0.975  2.70.

Therefore, 95% confidence interval for variance is given by

 9S2 9S2 
 19.02 , 2.70 
 
Let us do an example based on above discussion.
Example 6: A random sample of 10 workers is taken from a factory. The
wages (in hundreds) per months of these workers are given below:
48, 50, 62, 75, 80, 60, 70, 56, 52, 77
Obtain 95% confidence interval for the variance of wages of all the workers of
the factory.
Solution: Here, the population mean is unknown therefore, we use (1−)
100% confidence interval for population variance when population mean is
unknown which is given by
  n  1 S2  n  1 S2 
 2 , 2 
   n 1,  / 2  n 1, 1 / 2 

72
where, χ2n1, α/ 2 and χ2n1, 1α/ 2 are the values of χ 2 variate at (n−1) degrees of Interval Estimation
for One Population
freedom, whereas
1 2 1
S2 
n 1
  X  X  and X   X
n
Calculation for X and S2 :

S. No. Weight (X)  X  X 2

X  X
1 48 ‒15 225
2 50 ‒13 169
3 62 ‒1 1
4 75 12 144
5 80 17 289
6 60 ‒3 9
7 70 7 49
8 56 ‒7 49
9 52 ‒11 121
10 77 14 196
Sum  X  630  X  X
2
 1252

From the above calculation, we have

1 1 For 95% confidence
X
n
 X   630  63
10
interval
1    0.95    0.05
and α/2 = 0.025 &
1 2 1
S2 
n 1
  X  X    1252  139.11
9
1   / 2  0.975.

From 2-table, we have  2 n 1,  / 2   2 9, 0.025  19.02 and

2n 1,1/ 2  29,0.975  2.70

Thus, 95% confidence interval for the variance of wages of all the workers of
the factory is given by
10  139.11 10  139.11
 19.02 , 2.70  or 73.14, 515.22

Now, you can try following exercises to see how much you have followed.
E8) A study of variation in weights of soldiers was made and it is known
that the mean weight of soldiers follows the normal distribution. A
sample of 12 soldiers is taken from the soldier’s population and sample
variance is found 60 pound 2. Estimate the 95% confidence interval for
the variance of soldier’s weight of the population from which the
sample was drawn.
E9) If X1 = − 5, X2 = 4, X3 = 2, X4 = 6, X5 = −1, X6 = 4, X7 = 0, X8 = 10
and X9 = 7 are the sample observations taken from normal population
N(, σ2), obtain confidence interval for σ2.

73
Estimation
7.7 CONFIDENCE INTERVAL FOR NON-
NORMAL POPULATIONS
So far in this unit, we have kept our discussion on the confidence interval of
the normal population except population proportion. But one may be interested
to find out the confidence interval when the population under study is not
normal. The aim of this section is to give an idea how we can obtain
confidence interval for non-normal populations. For example, one may be
interested to estimate, say, 95% confidence interval of parameter  when
population under study follows exponential distribution().
We know that when the sample size is large then almost all the sampling
distributions of the statistics X, S2 , etc. follow normal distribution. So when
the sample is large we can also obtain the confidence interval as follows:
Let X1 , X 2 , ..., X n be a random sample of size n ( sufficiently large i.e.n  30 )
taken from f(x,) then according to the central limit theorem the sampling
distribution of sample mean X is normal, that is,
X ~ N  E  X  , Var  X  

Then the variate

X  E X
Z ~ N  0,1
Var  X 

is approximately normally distributed with mean 0 and variance unity. Since

distribution of Z is independent of parameter so it can be taken as pivotal
quantity therefore we introduce two constants z/2 and z(1‒/2) = -z/2 such that
P z  / 2  Z  z  / 2   1   … (17)
where, z/2 is the value of the variate Z having an area of /2 under the right
tail of the probability curve of Z.
By putting the value of Z in equation (17), we get
 X  E X  

P z  / 2   z / 2   1  

 Var  
X 

After this, we have to convert this interval for parameter  as discussed in
Section 7.3.
Following example will explain the procedure more clearly.
Example 7: Obtain 95% confidence interval to estimate  when a large sample
is taken from exponential population whose probability density function is
given by
f x, θ   θe θx ; x  0, θ  0
Solution: Let X1 , X 2 , ..., X n be a random sample of size n taken from
exponential population whose probability distribution is given by
f x, θ   θe θx ; x  0, θ  0
We know that for exponential distribution
74
 1 1 Interval Estimation
E X  and Var  X   2 for One Population
 
Since X1 , X 2 , ..., X n are independent and come from same exponential
distribution, therefore,
E  X i   E  X   1/  and Var  X i   Var  X   1/  2 for all i  1, 2, ..., n
Now consider,
1 
E  X   E   X1  X 2  ...  X n    By defination of sample mean
n 
1  E  aX  bY  
  E  X1   E  X 2   ...  E  X n  
n   aE  X   bE  Y  
 
 
11 1 1
    ...  
 
n    

 n  times 
1 1 1
= n  
n  
and
1 
Var  X   Var   X1  X 2  ...  X n  
n  If X and Y are 
 two independent 
1 random variable 
 2  Var  X1   Var  X 2   ...  Var  X n    then Var(aX + bY) 
n = a 2 Var(X) + b 2 Var(Y) 
 

 
1  1 1 1 
 2  2  2  ...  2 
 
n    
 n  times 
1 1
 n 
n2  2 
1
Var  X  
n2
Thus, the variate
X  E X  X  1/ 
Z  ~ N  0,1
Var  X  1/ n2

is approximately normally distributed with mean 0 and variance unity. Since

distribution of Z is independent of parameter so it can be taken as pivotal
quantity, therefore we introduce two constants z/2 and z(1-/2) = -z/2 such that
P z  / 2  Z  z  / 2   1  
where, z/2 is the value of the variate Z having an area of /2 under the right
tail of the probability curve of Z.
For 95% confidence interval, 1    0.95    0.05 and α/2 = 0.025, we
havez/2 = z0.025 = 1.96. So confidence interval for  is
P  1.96  Z  1.96   1  
75
Estimation
Putting the value of Z, we have
 1 
 X 
P  1.96    1.96   0.95
 1 
 n 2 

 1 
  X  1 
 P 1.96    1.96   0.95
 1 1 
  n 

 P 1.96  n  X  1  1.96  0.95

 1.96 1.96 
P   X  1    0.95
 n n
 1.96 1.96 
 P 1   X  1  0.95
 n n 
 1  1.96  1  1.96 
 P  1      X 1    0.95
X
  n   n 
Hence, 95% confidence interval for parameter θ is
 1  1.96 / n 1  1.96 / n 
 , 
 X X 

7.8 SHORTEST CONFIDENCE INTERVAL

It may be noted that for a confidence coefficient, there are many confidence
intervals for a parameter are possible. For example, from normal table (given in
the end of the Block 1 of this course) we can have many sets of a’s and b’s to
give 95% confidence interval for  some of them are given below:
P  1.65  Z  3.0  0.95 , P  1.68  Z  2.70  0.95

P 1.70  Z  2.54  0.95, P  1.96  Z  1.96  0.95 , etc.

Therefore, we need some criterion with the help of which we may choose the
best (best in the sense of minimum length) confidence interval among these set
of confidence intervals.
An obvious criterion (method) of selecting the shortest one out of these is that,
we chose a’s and b’s in such a way that the length of interval is minimum. In
above case, the lengths of these intervals are
L = T2 –T1
So, L1 = 3.0  (  1.65) = 4.65, L2 = 2.70  (  1.68) = 4.38
L3 = 2.54  (  1.70) = 4.24, L4 = 1.96  (  1.96) = 3.92
Hence, the last one has the minimum length. Therefore, it is best confidence
interval for  in all the above intervals on the basis of the minimum length
criterion.
76
 Interval Estimation
7.9 DETERMINATION OF SAMPLE SIZE for One Population

So, far you have become familiar with the main goal of this block. The
discussion of whole block centered on the theme of estimate some population
parameter of interest. To estimate some population parameter we have to draw
a random sample. A natural question which may arise in your mind “how large
should my sample be?” And this is very important question which is
commonly asked. From statistical point of view, the best answer of this
question is “take as large sample as you can afford. That is, if possible ‘sample’
the entire population that means study all units of the population under study
because by taking all units of the population we will have all the information
about the parameter and we will know the exact value of population parameter
which is better than any estimate of that parameter. Generally, this is
impractical to take entire population to be sampled due to economic
constraints, time constraints and other limitations. So the answer, “take as large
sample as you can afford is best if we ignore all costs because as you have
studied in Section 1.4 of Unit 1 of this course that the larger the sample size
smaller the standard error of the statistic that means less the uncertainty.
When the recourses in terms of money and time are limited then the question is
“how to found the minimum sample size which will satisfy some precision
requirements” In such cases, we should require first the answers of the
following three questions about his / her requirement about the survey:
1. How close do you want your sample estimate to be to the unknown
parameter? That means what should be the allowable difference between
the sample estimate and true value of population parameter. This difference
is known as sampling error or margin of error and represented by E.
2. The next question is, what do you want the confidence level to be so that
the difference between the estimate and the parameter is less than or equal
to E? That is, 90%, 95%, 99%, etc.
3. The last question is what is the population variance or population
proportion as may be the case?
When we have the answers of these three questions then we will get an answer
of the minimum required sample size.
In this section, we will describe the procedure of determining of sample size
for estimating population mean and population proportion.
Determination of Minimum Sample Size for Estimating
Population Mean
For determining minimum sample size for estimating population mean we use
confidence interval. Since as you seen in Section 7.4 of this unit that
confidence interval for population mean depends upon the nature of population
and population variance (σ2) is known or unknown so following cases may
arise:
Case I: Population is normal and population variance (σ2) is known
If σ2 is known and population is normal then we know that (1  )100%
confidence interval for population mean is given by
   
P X  z/ 2    X  z/ 2  1 
 n n 

77
Estimation    
Also P   Z / 2  X    z / 2  1  … (18)
 n n 
Since the normal distribution is symmetric so we can concentrate on the
right-hand equality so

X    z/ 2 … (19)
n
This inequality implies that the largest value that the difference X   can

assume is z  / 2 .
n
Also the difference between the estimator (sample mean X) and the population
parameter (population mean µ) is called the sampling error so

E  X    z/2 … (20)
n
Solving this equation for n, we have
z 2 / 22
n … (21)
E2
When population is finite of size N and sampling is to be done without
Nn
replacement then finite population correction is required so equation
N 1
(20) becomes
 Nn
E  z / 2
n N 1
which gives
N z 2 / 2  2
n … (22)
E 2  N  1  z  / 22  2
If the finite population correction is ignored then equation (22) is reduced to
equation (21).
Case II: Population is non-normal and population variance (σ 2) is known
If the population is not assumed to be normal and the population variance σ2 is
known then by central limit theorem we know that sampling distribution of
mean approximate normally distributed as sample size increases. So the above
method can be used determining minimum sample size. Once the required
sample size is obtained, we can check to see it that sample size greater than 30
and if it does we may be confident that our method of solution was appropriate.
Case III: Population is normal or non-normal and population variance
(σ2) is unknown
In this case, we use value of sample variance (S2) to estimate the population
variance but S2 is also calculated from a sample and we have not take a sample
yet. So in this case, determination of sample size is not directly obtained. The
most frequently methods for estimating  2 are as follows:
1. A pilot or preliminary sample may be drawn from the population under
study and the variance computed from this sample may be used as an
estimate of  2 .
2. The variance of previous or similar studies may used to estimate  2 .
3. If the population is assumed to be normal then we may use the fact that the

78
 range is approximately equal to six times of standard deviation i.e. Interval Estimation
  R / 6. This approximate require only knowledge of largest and smallest for One Population
value of the variable under study because range may be defined as
R = largest value – smallest value
Determination of Minimum Sample Size for Estimating
Population Proportion
The method of determination of minimum sample size for estimating
population proportion is similar as that described in estimating population
mean. So the formula for minimum sample size is given by
z 2 / 2 P(1  P)
n … (23)
E2
where, P is the population proportion and E = p – P is the sampling error or
margin of error.
When population is finite of size N and sampling is to be done without
Nn
replacement then finite population correction is required so
N 1
N z 2 / 2 P(1  P)
n 2 … (24)
E  N  1  z  / 22 P(1  P)

Here, sample size is depended upon the population proportion and it is

generally unknown so the most frequently methods for estimating P are as
follows:
1. A pilot or preliminary sample may be drawn from the population under
study and the sample proportion computed from this sample may be used as
an estimate of P.
2. The proportion of previous or similar studies may used to estimate P.
3. Absence of any information, we may use P = 0.5.
Now, it is time to do some examples based on determination of sample size.
Example 8: A hospital administrator wishes to estimate the mean weight of
babies born in her hospital. How large a sample of birth records should be
taken if she wants a 99 percent confidence that the estimate within the range of
0.4 pound? Assume that a reasonable estimate of  is 0.5 pound.
Solution: Here, we are given that
E = margin of error = 0.4, confidence level = 0.99 and σ = 0.5
Also for 99% confidence, 1    0.99    0.01 and α/2 = 0.005, we have
z/2 = z0.005 = 2.58.
Hospital administrator wishes to obtain the minimum sample size for
estimating the mean weight of babies born in her hospital so we have
2 2

n
z 2 / 22

 2.58   0.5  10.40 ~ 11
2 2
E  0.4 
Hence, hospital administrator should obtain a random sample of at least
11babies.
79
Estimation Example 9: The manufacturers of a car want to estimate the proportion of
people who are interested in a certain model. The company wants to know the
population proportion, P, to within 0.05 with 95% confidence. Current
company records indicate that the proportion P may be around 0.20. What is
the minimum required sample size for this survey?
Solution: Here, we are given that
E = margin of error = 0.05, confidence level = 0.95 and P = 0.20
Also for 95% confidence, 1    0.95    0.05 and α/2 = 0.025, we have
z/2 = z0.025 = 1.96.
The manufacturers of a car interested to obtain the minimum sample size for
estimating the population proportion so the required formula is given below
2
z 2 P(1  P) 1.96   0.20  0.80 
n  /2 2  2
 245.86 ~ 246
E  0.05
Hence, the company should require a random sample of at least 246 people.
You will be more cleared about this, when you try the following exercises.
E10) A survey is planned to determine the average annual family medical
expenses of employees of a large company. The management of the
company wishes to be 95% confident that the sample average is correct
to within ± 100 Rs of the true average family expenses. A pilot study
indicates that the standard deviation can be estimated can be estimated
as 400 Rs. How large a sample size is necessary?
E 11) The manager of a bank in a small city would like to determine the
proportion of the depositors per week. The manager wants to be 95%
confident of being correct to within ± 0.10 of the true proportion of
depositors per week. A guess is that the parentage of such depositors is
about 8%, what sample size is needed?
With this we have reached end this unit. Let us summarise what we have
discussed in this unit.

7.10 SUMMARY
In this unit, we have covered the following points:
1. The interval estimation.
2. The method of obtaining the confidence intervals.
3. The method of obtaining confidence interval for population mean of a
normal population when variance is known and unknown
4. The method of obtaining confidence interval for population proportion of a
population.
5. The method of obtaining confidence interval for population variance of a
normal population when population mean is known and unknown.
6. The method of obtaining confidence interval for population parameters of
non-normal populations.
7. The concept of the shortest interval.
8. Determination of sample size.
80
 Interval Estimation
7.11 SOLUTIONS / ANSWERS for One Population

E1) We know that the length of confidence interval [T1,T2] is given by

L = T2 –T1
Therefore, in our case we have
(i) L = 3.0  (  1.65) = 4.65
(ii) L = 2.70  (  1.68) = 4.38
(iii) L = 2.54  (  1.70) = 4.24
(iv) L = 1.96  (  1.96) = 3.92
E2) We know that if we have confidence interval for parameter θ is
P  T1    T2   1  
then
Lower confidence limit (LCL) = T1
Upper confidence limit (UCL) = T2
Confidence coefficient (CC) = 1  
Therefore in our cases, we have
(i) LCL = 0, UCL = 1.5 & CC = 0.90
(ii) LCL = −1, UCL = 2 & CC = 0.95
(iii) LCL = −2, UCL = 2 & CC = 0.98
(iv) LCL = −2.5, UCL = 2.5 & CC = 0.99

E3) Refer Section 7.3.

E4) Here, we are given that
n  200,   0.30, X  15.25
Since population standard deviation or variance is known so we use
(1-) 100% confidence limits for population mean when population
variance is known which are given by

X  z /2
n
where, z/2 is the value of the variate Z having an area of /2 under the
right tail of the probability curve of Z. For 95% confidence interval, we
have 1    0.95    0.05 and for  = 0.05, we have,
z / 2  z0.025  1.96.
Thus, the 95% confidence limits for the average weight of oil tins are
 0.30
X  1.96  15.25  1.96 
n 200
 15.25  0.04  15.21 and 15.29
E5) Here, we are given that
n  150, X  65, S  12
Since population variance is unknown and sample size is large sample,
81
Estimation so we use (1-) 100% confidence limits for population mean when
population variance is unknown which are given by
S
X  z /2
n
For 95% confidence interval, we have 1    0.95    0.05 and for 
= 0.05, we have, z / 2  z0.025  1.96.
Thus, the 95% confidence limits in which the average weight of all
students of a Study centre of IGNOU expected to lie are given by
S 12
X  1.96  65  1.96 
n 150
 65  1.92  63.08 and 66.92
Hence, required confidence limits are
63.08 and 66.92.
E6) Since population variance is unknown and sample size is small (n < 30),
therefore, (1-) 100% confidence limits for average height of credits of
particular centre are given by
S
X  t  n1,  / 2
n
1 1 2
where, X 
n
 X, S 
n 1
  X  X

Calculation for X and S :

S. No X  X  X 2
 X  X
1 70 -7 49
2 72 -5 25
3 80 3 9
4 82 5 25
5 78 1 1
6 80 3 9
 X  462 X  X
2
 118

Form the above calculation, we have

1 1
X   X   462  77
n 6
1 2 1
S2 
n 1
  X  X 
6 1
 118  23.6

For 95% confidence  S  4.86

interval
1    0.95    0.05 From t-table, we have t  n1,  / 2  t 5,0.025  2.571.
and α/2 = 0.025.
Thus, 95% confidence limits are given by
S 4.86
X  t  n 1, 0.025  77  2.571 
n 6
 77  2.571  1.98
 77  5.09  71.91 and 82.09
82
E7) We have Interval Estimation
for One Population
n = 400, X = 80
X 80 1
p    0.20
n 400 5
 np  400  0.20  80  5 and nq  400  (1  0.20)  400  0.80
 320  5 so (1-)100% confidence limits for the proportion are given
by
p 1  p  For 99% confidence
p  z/ 2 interval
n 1    0.99    0.01
For  = 0.01, we have, z/2 = z/2 = 2.58. and α/2 = 0.005.
Therefore, 99% confidence limits are
0.20  0.80
0.20  2.58
400
 0.2  2.58  0.02
 0.2  0.05  0.15 and 0.25
E8) Here, we are given that
n =12, S2 = 60
Since population mean is unknown, therefore, we use (1− ) 100%
confidence interval for population variance when population mean
unknown which is given by
  n  1 S2  n  1 S2 
 2 , 2 
  n 1,  / 2  n 1, 1 / 2 
For 95% confidence
2 2 2 interval
where, χ n1, α/ 2 and χ n1, 1α/ 2 are the values of χ variate at (n−1)
1    0.95    0.05
2
degrees of freedom. From  -table, we have and α/2 = 0.025.

2 n1,  /2  211, 0.025  21.92 and 2n1,1 /2 211,0.975  3.82

Thus, 95% confidence interval for the variance of soldier’s weight is

given by
 11  60 11  60 
 21.92 , 3.82   30.11, 172.77

E9) Since population mean is unknown, therefore, we use (1−) 100%

confidence interval for population variance which is given by
  n  1 S2  n  1 S2 
 2 , 2 
  n 1,  / 2   n 1, 1 / 2  
2 2
where, χn1, α/ 2 and χn1, 1α/ 2 are the values of χ 2 variate at (n−1) degrees
of freedom, whereas
1 2 1
S2 
n 1
  X  X  and X   X
n
83
Estimation Calculation for X and S2 :
2
X  X  X  X  X
-5 -8 64
4 1 1
2 -1 1
6 3 9
-1 -4 16
4 1 1
0 -3 9
10 7 49
7 4 16

 X  27 X  X
2
 166

Therefore,
1 1
X  X   27   3
n 9
2
Also,  n  1 S2    X  X   166 and from 2-table, we have
For 99% confidence
interval 2n1,  / 2  28, 0.005  21.96 and 2n 1, 1 / 2 28,0.995  1.34
1    0.99    0.01
and α/2 = 0.005 Thus, 99% confidence interval for σ2 is
 166 166 
 21.96 , 1.34  or  7.56, 123.88

E10) Here, we are given that

E = margin of error = 100, confidence level = 0.95 and σ = 400
Also for 95% confidence, 1    0.95    0.05 and α/2 = 0.025, we
have z/2 = z0.025 = 1.96.
The management of the company wishes to obtain the minimum sample
size for estimating the average annual family medical expenses of
employees of the company so the required formula is given below
2 2

n
z 2 / 22

1.96  400  61.46 ~ 62
2
E2 100 
Hence, the management of the company should require a random
sample of at least 62 employees.
E 11) Here, we are given that
E = margin of error = 0.1, confidence level = 0.95 and P = 0.08
Also for 95% confidence, 1    0.95    0.05 and α/2 = 0.025, we
have z/2 = z0.025 = 1.96.
The manager wants to obtain the minimum sample size for determining
the proportion so the required formula is given below
2
z 2 P(1  P) 1.96   0.08  0.94 
n  /2 2  2
 115.56 ~ 116
E  0.05
Hence, the manager should require a random sample of at least 116
depositors.

84
UNIT 8 INTERVAL ESTIMATION FOR TWO
POPULATIONS
Structure
8.1 Introduction
Objectives
8.2 Confidence Interval for Difference of Two Population Means
8.3 Confidence Interval for Difference of Two Population Proportions
8.4 Confidence Interval for Ratio of Two Population Variances
8.5 Summary
8.6 Solutions / Answers

8.1 INTRODUCTION
In the previous unit, we have discussed the method of obtaining confidence
interval for population mean, population proportion and population variance
for a population under study. There are so many situations where two
populations exist and one wants to obtain the interval estimate for the
difference or ratio of two parameters as means, proportions, variances, etc. For
example, a company manufacturing two types of blubs and product manager
may be interested to obtain the confidence interval for difference of average
life of two types of bulbs, one may wish to obtain the interval estimate of the
difference of proportions of alcohol drinkers in the two cities, a quality control
engineer wants to obtain the interval estimate for the ratio of variances of the
quality of the product, etc.
Therefore, it becomes necessary to construct the confidence interval for
difference of means, proportions and ratio of variances of two populations. In
this unit, we shall discuss how we construct confidence intervals for difference
or ratio of the above mentioned parameters of two populations.
This unit comprises the following six sections. Section 8.1 introduces the need
of confidence intervals for the difference or ratio of the parameters of two
normal populations. Section 8.2 is devoted to method of obtaining the
confidence interval for difference of two population means when population
variances are known and unknown. Section 8.3 described the method of
obtaining the confidence intervals for difference of two population proportions
with examples, whereas the method of obtaining the confidence interval for
ratio of population variances is explored in Section 8.4. Unit ends by providing
summary of what we have discussed in this unit in Section 8.5 and solution of
exercises in Section 8.6.
Objectives
After studying this unit, you should be able to:
 introduce the confidence intervals in case of two populations;
 describe the method of obtaining the confidence interval for difference of
means of two normal populations when variances are known and unknown;
 describe the method of obtaining the confidence interval for difference of
means of two normal populations when observations are paired;

85
 MST-004
STATISTICAL
Indira Gandhi
National Open University INFERENCE
School of Sciences

Block

3
TESTING OF HYPOTHESIS
UNIT 9
Concepts of Testing of Hypothesis 5
UNIT 10
Large Sample Tests 25
UNIT 11
Small Sample Tests 63
UNIT 12
Chi-Square and F-Tests 91
Curriculum and Course Design Committee
Prof. K. R. Srivathasan Prof. Rahul Roy
Pro-Vice Chancellor Math. and Stat. Unit
IGNOU, New Delhi Indian Statistical Institute, New Delhi

Prof. Parvin Sinclair Dr. Diwakar Shukla

Pro-Vice Chancellor Department of Mathematics and Statistics
IGNOU, New Delhi Dr. Hari Singh Gaur University, Sagar

Prof. Geeta Kaicker Prof. Rakesh Srivastava

Director, School of Sciences Department of Statistics
IGNOU, New Delhi M.S. University of Baroda, Vadodara

Prof. Jagdish Prasad Prof. G. N. Singh

Department of Statistics Department of Applied Mathematics
University of Rajasthan, Jaipur I.S.M., Dhanbad

Prof. R. M. Pandey Dr. Gulshan Lal Taneja

Department of Bio-Statistics Department of Mathematics
All India Institute of Medical Sciences M.D. University, Rohtak
New Delhi
Faculty members of School of Sciences, IGNOU
Statistics Mathematics
Dr. Neha Garg Dr. Deepika
Dr. Nitin Gupta Prof. Poornima Mital
Mr. Rajesh Kaliraman Prof. Sujatha Varma
Dr. Manish Trivedi Dr. S. Venkataraman

Block Preparation Team

Dr. Ramkishan (Editor) Dr. Diwakar Shukla (Units 1 - 4)
Department of Statistics Department of Mathematics and Statistics
D. A. V. (PG) College Dr. Hari Singh Gaur University, Sagar
C.C. S. University, Merrut
Mr. Prabhat Kumar Sangal (Units 1 - 4)
Dr. Parmod Kumar (Language Editor) School of Sciences, IGNOU
School of Humanities, IGNOU

Course Coordinator: Mr. Prabhat Kumar Sangal

Programme Coordinator: Dr. Manish Trivedi

Block Production
Mr. Sunil Kumar, AR (P),School of Sciences, IGNOU
CRC prepared by Mr. Prabhat Kumar Sangal, School of Sciences, IGNOU

Acknowledgement: I gratefully acknowledge my colleague Mr. Rajesh Kaliraman, Statistics

Discipline, School of Sciences for their great support.

July, 2013
© Indira Gandhi National Open University, 2013
ISBN-978-81-266-
All rights reserved. No part of this work may be reproduced in any form, by mimeograph or any other
means, without permission in writing from the Indira Gandhi National Open University.

Further information on the Indira Gandhi National Open University may be obtained from University’s
Office at Maidan Garhi, New Delhi-110068 or visit University’s website http://www.ignou.ac.in

Printed and published on behalf of the Indira Gandhi National Open University, New Delhi by the
Director, School of Sciences.

Printed at:
TESTING OF HYPOTHESIS
In previous block of this course, we have discussed one part of statistical
inference, that is, estimation and we have learnt how we estimate the unknown
population parameter(s) by using point estimation and interval estimation.
In this block we will focus on the second part of statistical inference which is
known as testing of hypothesis.
In this block, you will study the basic concept of testing of hypothesis and
different kinds of well-known tests. At the end of this block you will be aware
of the idea, procedure and applications of the hypothesis testing. This block
contains four units.
Unit 9: Concepts of Testing of Hypothesis
This unit explains the fundamental aspect relating to testing of hypothesis with
examples. It describes basic concepts and methodologies as hypothesis,
critical region, types of errors, level of significance, general procedure of
testing a hypothesis, concept of p-value, relation between confidence interval
and testing of hypothesis.
Unit 10: Large Sample Tests
This unit explores the procedure of testing the hypothesis based on one sample
and two samples when sample size is large and taken from normal or non-
normal population(s). In this unit, you will learn about Z-test for testing of
hypothesis about mean, difference of two means, proportion, difference of two
proportions, variance and equality of two variances.
Unit 11: Small Sample Tests
This unit is devoted to test a hypothesis based on one sample and two samples
when sample sizes are small. In this unit, you will learn about t-test for testing
of hypothesis about mean and difference of two means. This unit also explores
the idea for testing the hypothesis for equality of two means when samples are
dependent and testing of hypothesis about population correlation coefficient.
Unit 12: Chi-square and F Tests
The last unit of this block describes the chi-square test for population variance
and F-test for equality of two population variances.
Notations and Symbols
X1, X2, …, Xn : Random sample of size n
H0 : Null hypothesis
H1 or HA : Alternative hypothesis
 : Critical (rejection) region
 : Non-rejection region
α : Size of critical region or type-I error or level of significance
 : Type-II error
1‒ : Power of the test
z : Critical value of Z-test at α level of significance

t   ,  : Critical value of t-test with ν degrees of freedom at α level

of significance
2,  : Critical value of χ2-test with ν degrees of freedom at α level
of significance
F 1 , 2 ,  : Critical value of F-test with (ν1, ν2) degrees of freedom at α
level of significance
UNIT 9 CONCEPTS OF TESTING OF
HYPOTHESIS
Structure
9.1 Introduction
Objectives
9.2 Hypothesis
Simple and Composite Hypotheses
Null and Alternative Hypotheses
9.3 Critical Region
9.4 Type-I and Type-II Errors
9.5 Level of Significance
9.6 One-Tailed and Two-Tailed Tests
9.7 General Procedure of Testing a Hypothesis
9.8 Concept of p-Value
9.9 Relation between Confidence Interval and Testing of Hypothesis
9.10 Summary
9.11 Solutions /Answers

9.1 INTRODUCTION
In previous block of this course, we have discussed one part of statistical
inference, that is, estimation and we have learnt how we estimate the unknown
population parameter(s) by using point estimation and interval estimation. In
this block, we will focus on the second part of statistical inference which is
known as testing of hypothesis.
In our day-to-day life, we see different commercials advertisements in
television, newspapers, magazines, etc. such as
(i) The refrigerator of certain brand saves up to 20% electric bill,
(ii) The motorcycle of certain brand gives 60 km/liter mileage,
(iii) A detergent of certain brand produces the cleanest wash,
(iv) Ninety nine out of hundred dentists recommend brand A toothpaste for
their patients to save the teeth against cavity, etc.
Now, the question may arise in our mind “can such types of claims be verified
statistically?” Fortunately, in many cases the answer is “yes”.
The technique of testing such type of claims or statements or assumptions is
known as testing of hypothesis. The truth or falsity of a claim or statement is
never known unless we examine the entire population. But practically it is not
possible in mostly situations so we take a random sample from the population
under study and use the information contained in this sample to take the
decision whether a claim is true or false.
This unit is divided into 11 sections. Section 9.1 is introductory in nature. In
Section 9.2, we defined the hypothesis. The concept and role of critical region
in testing of hypothesis is described in Section 9.3. In Section 9.4, we explored
the types of errors in testing of hypothesis whereas level of significance is
explored in Section 9.5. In Section 9.6, we explored the types of tests in testing
5
Testing of Hypothesis of hypothesis. The general procedure of testing a hypothesis is discussed in
Section 9.7. In Section 9.8, the concept of p-value in decision making about the
null hypothesis is discussed whereas the relation between confidence interval
and testing of hypothesis is discussed in Section 9.9. Unit ends by providing
summary of what we have discussed in this unit in Section 9.10 and solution of
exercises in Section 9.11.
Objectives
After reading this unit, you should be able to:
 define a hypothesis;
 formulate the null and alternative hypotheses;
 explain what we mean by type-I and type-II errors;
 explore the concept of critical region and level of significance;
 define one-tailed and two-tailed tests;
 describe the general procedure of testing a hypothesis;
 concept of p-value; and
 test a hypothesis by using confidence interval.
Before coming to the procedure of testing of hypothesis, we will discuss the
basis terms used in this procedure one by one in subsequent sections.

9.2 HYPOTHESIS
As we have discussed in previous section that in our day-to-day life, we see
different commercials advertisements in television, newspapers, magazines,
etc. and if someone may be interested to test such type of claims or statement
then we come across the problem of testing of hypothesis. For example,
(i) a customer of motorcycle wants to test whether the claim of motorcycle
of certain brand gives the average mileage 60 km/liter is true or false,
(ii) the businessman of banana wants to test whether the average weight of a
banana of Kerala is more than 200 gm,
(iii) a doctor wants to test whether new medicine is really more effective for
controlling blood pressure than old medicine,
(iv) an economist wants to test whether the variability in incomes differ in
two populations,
(v) a psychologist wants to test whether the proportion of literates between
two groups of people is same, etc.
In all the cases discussed above, the decision maker is interested in making
inference about the population parameter(s). However, he/she is not interested
in estimating the value of parameter(s) but he/she is interested in testing a
claim or statement or assumption about the value of population parameter(s).
Such claim or statement is postulated in terms of hypothesis.
In statistics, a hypothesis is a statement or a claim or an assumption about the
value of a population parameter (e.g., mean, median, variance, proportion,
etc.).
Similarly, in case of two or more populations a hypothesis is comparative
statement or a claim or an assumption about the values of population
parameters. (e.g., means of two populations are equal, variance of one
population is greater than other, etc.). The plural of hypothesis is hypotheses.
6
In hypothesis testing problems first of all we should being identifying the claim Concepts of Testing of
or statement or assumption or hypothesis to be tested and write it in the words. Hypothesis
Once the claim has been identified then we write it in symbolical form if
possible. As in the above examples,
(i) Customer of motorcycle may write the claim or postulate the hypothesis
“the motorcycle of certain brand gives the average mileage 60 km/liter.”
Here, we are concerning the average mileage of the motorcycle so let µ
represents the average mileage then our hypothesis becomes µ = 60 km /
liter.
(ii) Similarly, the businessman of banana may write the statement or
postulate the hypothesis “the average weight of a banana of Kerala is
greater than 200 gm.” So our hypothesis becomes µ > 200 gm.
(iii) Doctor may write the claim or postulate the hypothesis “ the new
medicine is really more effective for controlling blood pressure than old
medicine.” Here, we are concerning the average effect of the medicines
so let µ1 and µ2 represent the average effect of new and old medicines
respectively on controlling blood pressure then our hypothesis becomes
µ1 > µ2.
(iv) Economist may write the statement or postulate the hypothesis “ the
variability in incomes differ in two populations.” Here, we are concerning
the variability in income so let 12 and 22 represent the variability in
incomes in two populations respectively then our hypothesis becomes
12  22 .
(v) Psychologist may write the statement or postulate the hypothesis “the
proportion of literates between two groups of people is same.” Here, we
are concerning the proportion of literates so let P1 and P2 represent the
proportions of literates of two groups of people respectively then our
hypothesis becomes P1 = P2 or P1 –P2 = 0.
The hypothesis is classified according to its nature and usage as we will discuss
in subsequent subsections.
9.2.1 Simple and Composite Hypotheses
In general sense, if a hypothesis specifies only one value or exact value of the
population parameter then it is known as simple hypothesis. And if a
A hypothesis which
hypothesis specifies not just one value but a range of values that the population completely specifies
parameter may assume is called a composite hypothesis. parameter(s) of a
As in the above examples, the hypothesis postulated in (i) µ = 60 km/liter is theoretical population
(probability distribution)
simple hypothesis because it gives a single value of parameter (µ = 60) is called a simple
whereas the hypothesis postulated in (ii) µ > 200 gm is composite hypothesis hypothesis otherwise
because it does not specify the exact average value of weight of a banana. It called composite
may be 260, 350, 400 gm or any other. hypothesis.

Similarly, (iii) µ1 > µ2 or µ1 −µ2 > 0 and (iv) 12  22 or 12  22  0 are not
simple hypotheses because they specify more than one value as µ1 −µ2 = 4,
µ1 −µ2 = 7, 12  22  2, 12  22  5 , etc. and (v) P1 = P2 or P1 –P2 = 0 is simple
hypothesis because it gives a single value of parameter as P1 –P2 = 0.
9.2.2 Null and Alternative Hypotheses
As we have discussed in last page that in hypothesis testing problems first of
all we identify the claim or statement to be tested and write it in symbolical
7
Testing of Hypothesis form. After that we write the complement or opposite of the claim or statement
in symbolical form. In our example of motorcycle, the claim is µ = 60 km/liter
then its complement is µ ≠ 60 km/liter. In (ii) the claim is µ > 200 gm then its
complement is µ ≤ 200 gm. If the claim is µ < 200 gm then its complement is
µ ≥ 200 gm. The claim and its complement are formed in such a way that they
cover all possibility of the value of population parameter.
Once the claim and its compliment have been established then we decide of
We state the null and these two which is the null hypothesis and which is the alternative hypothesis.
alternative The thump rule is that the statement containing equality is the null hypothesis.
hypotheses in such a
way that they cover
That is, the hypothesis which contains symbols  or  or  is taken as null
all possibility of the hypothesis and the hypothesis which does not contain equality i.e. contains
value of population  or  or  is taken as alternative hypothesis. The null hypothesis is denoted
parameter.
by H0 and alternative hypothesis is denoted by H1 or HA.
In our example of motorcycle, the claim is µ = 60 km/liter and its complement
is µ ≠ 60 km/liter. Since claim µ = 60 km/liter contains equality sign so we take
it as a null hypothesis and complement µ ≠ 60 km/liter as an alternative
hypothesis, that is,
H0 : µ = 60 km/liter and H1: µ ≠ 60 km/liter
In our second example of banana, the claim is µ > 200 gm and its complement
is µ ≤ 200 gm. Since complement µ ≤ 200 gm contains equality sign so we take
complement as a null hypothesis and claim µ > 200 gm as an alternative
hypothesis, that is,
H0 : µ ≤ 200 gm and H1: µ > 200 gm
Formally these hypotheses are defined as
The hypothesis which we wish to test is called as the null hypothesis.
According to Prof. R.A. Fisher,
“A null hypothesis is a hypothesis which is tested for possible rejection under
the assumption that it is true.”
The hypothesis which complements to the null hypothesis is called alternative
hypothesis.
Note 1: Some authors use equality sign in null hypothesis instead of ≥ and ≤
signs.
The alternative hypothesis has two types:
(i) Two-sided (tailed) alternative hypothesis
(ii) One-sided (tailed) alternative hypothesis
If the alternative hypothesis gives the alternate of null hypothesis in both
directions (less than and greater than) of the value of parameter specified in
null hypothesis then it is known as two-sided alternative hypothesis and if it
gives an alternate only in one direction( less than or greater than) only then it is
known as one-sided alternative hypothesis. For example, if our alternative
hypothesis is H1 : θ ≠ 60 then it is a two-sided alternative hypothesis because its
means that the value of parameter θ is greater than or less than 60. Similarly, if
H1 : θ > 60 then it is a right-sided alternative hypothesis because its means that
the value of parameter θ is greater than 60 and if H1: θ < 60 then it is a
left-sided alternative hypothesis because its means that the value of parameter θ
is less than 60.
In testing procedure, we assume that the null hypothesis is true until there is
sufficient evidence to prove that it is false. Generally, the hypothesis is tested
8
with the help of a sample so evidence in testing of hypothesis comes from a Concepts of Testing of
sample. If there is enough sample evidence to suggest that the null hypothesis Hypothesis
is false then we reject the null hypothesis and support the alternative
hypothesis. If the sample fails to provide us sufficient evidence against the null
hypothesis we are not saying that the null hypothesis is true because here, we
take the decision on the basis of a random sample which is a small part of the
population. To say that null hypothesis is true we must study all observations
of the population under study. For example, if someone wants to test that the
person of India has two hands then to prove that this is true we must check all
the persons of India whereas to prove that it is false we require a person he /
she has one hand or no hand. So we can only say that there is not enough
evidence against the null hypothesis.
Note 2: When we assume that null hypothesis is true then we are actually
assuming that the population parameter is equal to the value in the claim. In our
example of motorcycle, we assume that µ = 60 km/liter whether the null
hypothesis is µ = 60 km/liter or µ ≤ 60 km/liter or µ ≥ 60 km/liter.
Now, you can try the following exercises.
E1) A company manufactures car tyres. Company claims that the average life
of its tyres is 50000 miles. To test the claim of the company, formulate
the null and alternative hypotheses.
E2) Write the null and alternative hypotheses in case (iii), (iv) and (v) of our
example given in Section 9.2.
E3) A businessman of orange formulates different hypotheses about the
average weight of the orange which are given below:
(i) H0:  = 100 (ii) H1 :  >100 (iii) H0 :  ≤ 100 (iv) H1:  ≠ 100
(v) H1:  > 150 (vi) H0:  = 130 (vii) H1:   0
Categorize the above cases into simple and composite hypotheses.
After describing the hypothesis and its types our next point in the testing of
hypothesis is critical region which will be described in next section.

9.3 CRITICAL REGION

As we have discussed in Section 9.1 that generally, null hypothesis is tested by
the sample data. Suppose X1, X2,…, Xn be a random sample drawn from a
population having unknown population parameter . The collection of all
possible values of X1, X2,…, Xn is a set called sample space(S) and a particular
value of X1, X2,…, Xn represents a point in that space. A statistic is a function
In order to test a hypothesis, the entire sample space is partitioned into two of sample observations
(not including
disjoint sub-spaces, say,  and S    . If calculated value of the test parameter). Basically, a
statistic lies in ω , then we reject the null hypothesis and if it lies in ω , then we test statistic is a statistic
do not reject the null hypothesis. The region  is called a “rejection region or which is used in
critical region” and the region  is called a “non-rejection region”. decision making about
the null hypothesis.
Therefore, we can say that
“A region in the sample space in which if the calculated value of the test
statistic lies, we reject the null hypothesis then it is called critical region or
rejection region.”
This can be better understood with the help of an example. Suppose, 100
students appear in total 10 papers two of each in English, Physics, Chemistry,
Mathematics and Computer Science of a Programme. Suppose the scores in
Testing of Hypothesis these papers are denoted by X1, X2, …, X10 and maximum marks =100 for each
paper. For obtaining the distinction award in this Programme, a student needs
to have total score equal to or more than 750 which is a rule.
Suppose, we select one student randomly out of 100 students and we want to
test that the selected student is a distinction award holder. So we can take the
null and alternative hypotheses as
H0 : Selected student is a distinction award holder
H1 : Selected student is not a distinction award holder
For taking the decision about the selected student, we define a statistic
10
T10  X
i 1
i as the sum of the scores in all the 10 papers of the student. The

range of T10 is 0  T10  1000. Now, we divide the whole space ( 0-1000) into
two regions as no-distinction awarded region (less than 750) and distinction
awarded region (greater than or equal to 750) as shown in Fig. 9.1. Here, 750 is
the critical value which separates the no-distinction and distinction awarded
regions.

Tn
Fig. 9.1: Non-rejection and critical regions for distinction award
On the basis of scores in all the papers of the selected student, we calculate the
10
value of the statistic T10   Xi . And calculated value may fall in distinction
i 1
award region or not, depending upon the observed value of test statistic.
For making a decision to reject or do not reject H0, we use test statistic
10
T10  X
i 1
i (sum of scores of 10 papers). If calculated value of test statistic T10

lies in no-distinction awarded region (critical region), that is, T10 < 750 then we
reject H0 and if calculated value of test statistic T10 lies in distinction awarded
region (non-rejection region), that is, T10  750 then we do not reject H0. It is a
basic structure of the procedure of testing of hypothesis which needs two
regions like:
(i) Region of rejection of null hypothesis H0
(ii) Region of non-rejection of null hypothesis H0
The point of discussion in this test procedure is “how to fix the cut off value
750”? What is the justification for this value? The distinction award region
may be like T10  800 or at T10  850 or at T10  900. So, there must be a
scientific justification for the cut-off value 750. In a statistical test procedure it
is obtained by using the probability distribution of the test statistic.
The region of rejection is called critical region. It has a pre-fixed area generally
denoted by , corresponding to a cut-off value in a probability distribution of
test statistic.
The rejection (critical) region lies in one-tail or two-tails on the probability
curve of sampling distribution of the test statistic its depends upon the
alternative hypothesis. Therefore, three cases arise: Concepts of Testing of
Hypothesis
Case I: If the alternative hypothesis is right-sided such as H1 : θ > θ0 or
H1 : θ1 > θ2 then the entire critical or rejection region of size α lies on
right tail of the probability curve of sampling distribution of the test
statistic as shown in Fig. 9.2.

Critical value is a
value or values that
separate the region of
rejection from the non-
rejection region.

Fig. 9.2
Case II: If the alternative hypothesis is left-sided such as H1: θ < θ0 or
H1 : θ1 < θ2 then the entire critical or rejection region of size α lies on
left tail of the probability curve of sampling distribution of the test
statistic as shown in Fig. 9.3.

Fig. 9.3
Case III: If the alternative hypothesis is two sided such as H1 : θ ≠ θ0 or
H1 : θ1 ≠ θ2 then critical or rejection regions of size α/2 lies on both
tails of the probability curve of sampling distribution of the test
statistic as shown in Fig. 9.4.

Fig. 9.4
Now, you can try the following exercise.
E4) If H0: θ = 60 and H1: θ ≠ 60 then critical region lies in one-tail or two-
tails.
Testing of Hypothesis
9.4 TYPE-I AND TYPE-II ERRORS
In Section 9.3, we have discussed a rule that if the value of test statistic falls in
rejection (critical) region then we reject the null hypothesis and if it falls in the
non-rejection region then we do not reject the null hypothesis. A test statistic is
calculated on the basis of observed sample observations. But a sample is a
small part of the population about which decision is to be taken. A random
sample may or may not be a good representative of the population.
A faulty sample misleads the inference (or conclusion) relating to the null
hypothesis. For example, an engineer infers that a packet of screws is sub-
standard when actually it is not. It is an error caused due to poor or
inappropriate (faulty) sample. Similarly, a packet of screws may infer good
when actually it is sub-standard. So we can commit two kinds of errors while
testing a hypothesis which are summarised in Table 9.1 which is given below:
Table 9.1: Type of Errors
Decision H0 True H1 True
Reject H0 Type-I Error Correct Decision
Do not reject H0 Correct Decision Type-II Error

Let us take a situation where a patient suffering from high fever reaches to a
doctor. And suppose the doctor formulates the null and alternative hypotheses
as
H0 : The patient is a malaria patient
H1 : The patient is not a malaria patient
Then following cases arise:
Case I: Suppose that the hypothesis H0 is really true, that is, patient actually
a malaria patient and after observation, pathological and clinical
examination, the doctor rejects H0, that is, he / she declares him / her
a non-malaria-patient. It is not a correct decision and he / she
commits an error in decision known as type-I error.
Case II: Suppose that the hypothesis H0 is actually false, that is, patient
actually a non-malaria patient and after observation, the doctor
rejects H0, that is, he / she declares him / her a non-malaria-patient. It
is a correct decision.
Case III: Suppose that the hypothesis H0 is really true, that is, patient actually
a malaria patient and after observation, the doctor does not reject H0,
that is, he / she declares him / her a malaria-patient. It is a correct
decision.
Case IV: Suppose that the hypothesis H0 is actually false, that is, patient
actually a non-malaria patient and after observation, the doctor does
not reject H0, that is, he / she declares him / her a malaria-patient. It
is not a correct decision and he / she commits an error in decision
known as type-II error.
Thus, we formally define type-I and type-II errors as below:
Type-I Error:
The decision relating to rejection of null hypothesis H0 when it is true is called
type-I error. The probability of committing the type-I error is called size of test,
denoted by  and is given by
 = P [Reject H0 when H0 is true] = P [Reject H0 / H0 is true]

12
We reject the null hypothesis if random sample / test statistic falls in rejection Concepts of Testing of
region, therefore, Hypothesis

α = P [ X   / H0 ]
where X = (X1, X2,…,Xn) is a random sample and ω is the rejection region and
1- = 1-P[Reject H0 / H0 is true]
= P[Do not reject H0 / H0 is true] = P[Correct decision]
The (1-) is the probability of correct decision and it correlates to the concept
of 100(1-)% confidence interval used in estimation.
Type-II Error:
The decision relating to non-rejection of null hypothesis H0 when it is false
(i.e. H1 is true) is called type-II error. The probability of committing type-II
error is generally denoted by  and is given by
 = P[Do not reject H0 when H0 is false]
= P[Do not reject H0 when H1 is true]
= P[Do not reject H0 / H1 is true]
= P[ X  ω / H1 ] where,  is the non-rejection region.
and
1- = 1-P[Do not reject H0 / H1 is true]
= P[Reject H0 / H1 is true] = P[Correct decision]
The (1-) is the probability of correct decision and also known as “power of
the test”. Since it indicates the ability or power of the test to recognize
correctly that the null hypothesis is false, therefore, we wish a test that yields a
large power.
We say that a statistical test is ideal if it minimizes the probability of both types
of errors and maximizes the probability of correct decision. But for fix sample
size,  and  are so interrelated that the decrement in one results into the
increment in other. So minimization of both probabilities of type-I and type-II
errors simultaneously for fixed sample size is not possible without increasing
sample size. Also both types of errors will be at zero level (i.e. no error in
decision) if size of the sample is equal to the population size. But it involves
huge cost if population size is large. And it is not possible in all situations such
as testing of blood.
Depending on the problem in hand, we have to choose the type of error which
has to minimize. For this, we have to look at a situation, suppose there is a
decision making problem and there is a rule that if we make type-I error, we
lose10 rupees and if we make type-II error we lose 1000 rupees. In this case,
we try to eliminate the type-II error, since it is more expensive.
In another situation, suppose the Delhi police arrests a person whom they
suspect is a murderer. Now, policemen have to test hypothesis:
H0: Arrested person is innocent (not murderer)
H1: Arrested person is a murderer
The type-I error is
 = P [Reject H0 when it is true]
That is, suspected person who is actually an innocent will be sent to jail when
H0 rejects, although H0 being a true.
Testing of Hypothesis The type-II error is
 = P [Do not reject H0 when H1 is true]
That is, when arrested person truly a murderer but released by the police. Now,
we see that in this case type-I error is more serious than type-II error because a
murderer may be arrested / punished later on but sending jail to an innocent
person is serious.
Consider another situation, suppose we want to test the null hypothesis
H0 : p  0.5 against H1 : p  0.5 on the basis of tossing a coin once, where p is
the probability of getting a head in a single toss (trial). And we reject the null
hypothesis if a head appears and do not reject otherwise. The type-I error, that
is, the probability of Reject H0 when it is true can be calculated easily(as shown
in Example 1) but the computation of type-II error is not possible because there
are infinitely many alternatives for p such as p = 0.6, p = 0.1, etc.
Generally, strong control on α is necessary. It should be kept as low as
possible. In test procedure, we prefix it at very low level like  = 0.05 ( 5%) or
0.01 (1%) .
Now, it is time to do some examples relating to α and .
Example 1: It is desired to test a hypothesis H 0 :p  p0  1/ 2 against the
alternative hypothesis H1 :p  p1  1/ 4 on the basis of tossing a coin once,
where p is the probability of “getting a head” in a single toss (trial) and
agreeing to reject H0 if a head appears and accept H0 otherwise. Find the value
of  and .
Solution: In such type of problems, first of all we search for critical region.
Here, we have critical region  = {head}
Therefore, the probability of type-I error can be obtained as
 = P[Reject H0 when H0 is true]
 P[X   / H 0 ]= P[Head appears / H 0 ]
1  H0 is trueso we take value 
 P  Head appears  1 
p
2 2  of parameter pgiven in H 0 
Also,
 = P[Do not reject H0 when H1 is true]
 P  X   / H1   P [Tail appears / H1 ]

 P  Tail appears   H1 is trueso we take value 

p
1
4 of parameter p given in H1 
1 3
 1  P  Head appears  1  1  
p
4 4 4
Example 2: For testing H0 :  = 1 against H1 :  = 2, the pdf of the variable is
given by
1
 ; 0x
f x ,    
 0; elsewhere

Obtain type-I and type-II errors when critical region is X  0.4. Also obtain
power function of the test.

14
Solution: Here, we have critical (rejection) and non-rejection regions as Concepts of Testing of
Hypothesis
  X : X  0.4 and   X : X  0.4
We have to test the null hypothesis
H 0 : θ  1 against H1 : θ  2
The size of type-I error is given by
  P  X   / H 0   P  X  0.4 /   1
   

   f  x,  dx   
 P X  a   f  x,  dx
 … (1)
0.4 1  a 
1
Now, by using f x ,    ; 0  x  , we get from equation (1)

 1  1
1
    dx    dx   x 0.4  1  0.4  0.6
 0.4   1 0.4
Similarly, the size of type-II error is given by
  P  X   / H1   P  X  0.4 /   2 

 0.4 1  0.4
1 1 0.4 1
    dx    dx   x 0   0.4  0   0.2
 0    2 0 2 2 2

The power function of the test = 1- = 1- 0.2 = 0.8.

Now, you can try the following exercise.
E5) An urn contains either 4 white and 2 black balls or 2 white and 4 black
balls. Two balls are to be drawn from the urn. If less than two white
balls are obtained, it will be decided that this urn contains 2 white and 4
black balls. Calculate the values of  and .

9.5 LEVEL OF SIGNIFICANCE

So far in this unit, we have discussed the hypothesis, types of hypothesis,
critical region and types of errors. In this section, we shall discuss very useful
concept “level of significance”, which play an important role in decision
making while testing a hypothesis.
The probability of type-I error is known as level of significance of a test. It is
also called the size of the test or size of critical region, denoted by α. Generally,
it is pre-fixed as 5% or 1% level (α = 0.05 or 0.01). As we have discussed in
Section 9.3 that if calculated value of the test statistic lies in rejection(critical)
region, then we reject the null hypothesis and if it lies in non-rejection region,
then we do not reject the null hypothesis. Also we note that when H0 is rejected
then automatically the alternative hypothesis H1 is accepted. Now, one point of
our discussion is that how to decide critical value(s) or cut-off value(s) for a
known test statistic.
If distribution of test statistic could be expressed into some well-known
distributions like Z, 2, t, F etc. then our problem will be solved and using the
probability distribution of test statistic, we can find the cut-off value(s) that
provides us critical area equal to 5% (or 1%).
15
Testing of Hypothesis Another viewpoint about the level of significance relates to the trueness of the
conclusion. If H0 do not reject at level, say,  = 0.05 (5% level) then a person
will be confident that “concluding statement about H0” is true with 95%
assurance. But even then it may false with 5% chance. There is no cent-percent
assurance about the trueness of statement made for H0.
As an example, if among 100 scientists, each draws a random sample and use
the same test statistic to test the same hypothesis H0 conducting same
experiment, then 95 of them will reach to the same conclusion about H0. But
still 5 of them may differ (i.e. against the earlier conclusion).
Similar argument can be made for, say,  = 0.01 (=1%). It is like when H0 is
rejected at  = 0.01by a scientist , then out of 100 similar researchers who
work together at the same time for the same problem, but with different
random samples, 99 of them would reach to the same conclusion however, one
may differ.
Now, you can try the following exercise.
E6) If probability of type-I error is 0.05 then what is the level of significance?

9.6 ONE-TAILED AND TWO-TAILED TESTS

We have seen in Section 9.3 that rejection (critical) region lies at one-tail or
two-tails on the probability curve of sampling distribution of the test statistic its
depend upon the form of alternative hypothesis. Similarly, the test of testing
the null hypothesis also depends on the alternative hypothesis.
A test of testing the null hypothesis is said to be two-tailed test if the alternative
hypothesis is two-tailed whereas if the alternative hypothesis is one-tailed then
a test of testing the null hypothesis is said to be one-tailed test.
For example, if our null and alternative hypothesis are
H 0 :   0 and H1 :   0
then the test for testing the null hypothesis is two-tailed test because the
alternative hypothesis is two-tailed that means, the parameter θ can take value
greater than θ0 or less than θ0.
If the null and alternative hypotheses are
H 0 :   0 and H1 :   0
then the test for testing the null hypothesis is right-tailed test because the
alternative hypothesis is right-tailed.
Similarly, if the null and alternative hypotheses are
H 0 :   0 and H1 :   0
then the test for testing the null hypothesis is left-tailed test because the
alternative hypothesis is left-tailed.
The above discussion can be summarised in Table 9.2.
Table 9.2: Null and Alternative Hypotheses and Corresponding One-tailed and
Two-tailed Tests
Null Hypothesis Alterative Hypothesis Types of Critical Region / Test
Two-tailed test having critical
H0 : θ  θ0 H1 : θ  θ 0 regions under both tails.
Right-tailed test having critical
H 0 :   0 H1 : θ  θ 0 region under right tail only.
Left- tailed test having critical
H 0 :   0 H1 : θ  θ 0 region under left tail only.

16
Let us do one example based on type of tests. Concepts of Testing of
Hypothesis
Example 3: A company has replaced its original technology of producing
electric bulbs by CFL technology. The company manager wants to compare the
average life of bulbs manufactured by original technology and new technology
CFL. Write appropriate null and alternate hypotheses. Also say about one tailed
and two tailed tests.
Solution: Suppose the average lives of original and CFL technology bulbs are
denoted by 1 and 2 respectively.
If company manager is interested just to know whether any significant
difference exists in average-life time of two types of bulbs then null and
alternative hypotheses will be:
H0 : µ1 = µ2 [average lives of two types of bulbs are same]
H1 : µ1  µ2 [average lives of two types of bulbs are different]
Since alternative hypothesis is two-tailed therefore corresponding test will be
two-tailed.
If company manager is interested just to know whether average life of CFL is
greater than original technology bulbs then our null and alternative hypotheses
will be
H0 : µ1 ≥ µ2

H1: µ1 < µ2  average life of CFLtechnology bulbs 

 is greater than average life of original technology 
Since alternative hypothesis is left-tailed therefore corresponding test will be
left-tailed test.
Now, you can try the following exercises.
E7) If we have null and alternative hypotheses as
H0:  = 0 and H1:   0
then corresponding test will be
(i) left-tailed test (ii) right-tailed test (iii) both-tailed test
Write the correct option.
E8) The test whether one-tailed or two-tailed depends on
(i) Null hypothesis (H0) (ii) Alternative hypothesis (H1)
(iii) Neither H0 nor H1 (iv) Both H0 and H1

9.7 GENERAL PROCEDURE OF TESTING A

HYPOTHESIS
Testing of hypothesis is a huge demanded statistical tool by many discipline
and professionals. It is a step by step procedure as you will see in next three
units through a large number of examples. The aim of this section is just give
you flavor of that sequence which involves following steps:
Step I: First of all, we have to setup null hypothesis H0 and alternative
hypothesis H1. Suppose, we want to test the hypothetical / claimed /

17
Testing of Hypothesis assumed value θ0 of parameter θ. So we can take the null and
alternative hypotheses as
H 0 :   0 and H1 :   0 for two-tailed test
H 0 :   0 and H1 :   0 
or  for one-tailed test 
H 0 :   0 and H1 :   0 
In case of comparing same parameter of two populations of interest,
say, 1 and 2, then our null and alternative hypotheses would be
H0 : 1  2 and H1 : 1  2 for two-tailed test
H 0 : 1  2 and H1 : 1  2 
or   for one-tailed test 
H 0 : 1  2 and H1 : 1  2 
Step II: After setting the null and alternative hypotheses, we establish a
criteria for rejection or non-rejection of null hypothesis, that is,
decide the level of significance (), at which we want to test our
hypothesis. Generally, it is taken as 5% or 1% (α = 0.05 or 0.01).
Step III: The third step is to choose an appropriate test statistic under H0 for
testing the null hypothesis as given below:
Statistic  Value of the parameter under H 0
Test statistic 
Standard error of statistic
After that, specify the sampling distribution of the test statistic
preferably in the standard form like Z (standard normal), 2, t, F or
any other well-known in literature.
Step IV: Calculate the value of the test statistic described in Step III on the
basis of observed sample observations.
Step V: Obtain the critical (or cut-off) value(s) in the sampling distribution
of the test statistic and construct rejection (critical) region of size .
Generally, critical values for various levels of significance are
putted in the form of a table for various standard sampling
distributions of test statistic such as Z-table, 2-table, t-table, etc.
Step VI: After that, compare the calculated value of test statistic obtained
from Step IV, with the critical value(s) obtained in Step V and
locates the position of the calculated test statistic, that is, it lies in
rejection region or non-rejection region.
Step VII: In testing of hypothesis ultimately we have to reach at a conclusion.
It is done as explained below:
(i) If calculated value of test statistic lies in rejection region at 
level of significance then we reject null hypothesis. It means
that the sample data provide us sufficient evidence against the
null hypothesis and there is a significant difference between
hypothesized value and observed value of the parameter.
(ii) If calculated value of test statistic lies in non-rejection region at
 level of significance then we do not reject null hypothesis. Its
means that the sample data fails to provide us sufficient
evidence against the null hypothesis and the difference between
hypothesized value and observed value of the parameter due to
fluctuation of sample.
18
Note 3: Nowadays the decision about the null hypothesis is taken with the help Concepts of Testing of
of p-value. The concept of p-value is very important, because computer Hypothesis
packages and statistical software such as SPSS, SAS, STATA, MINITAB,
EXCEL, etc. all provide p-value. So, Section 9.8 is devoted to explain the
concept of p-value.
Now, with the help of an example we explain the above procedure.
Example 4: Suppose, it is found that average weight of a potato was 50 gm
and standard deviation was 5.1 gm nearly 5 years ago. We want to test that due
to advancement in agricultural technology, the average weight of a potato has
been increased. To test this, a random sample of 50 potatoes is taken and
calculate the sample mean (X) as 52gm. Describe the procedure to carry out
this test.
Solution: Here, we are given that
Specified value of population mean = 0 = 50 gm,
Population standard deviation = σ = 5.1 gm,
Sample size = n = 50,
Sample mean = X = 52 gm
To carry out the above test, we have to follow up the following steps:
Step I: First of all, we setup null and alternative hypotheses. Here, we want
to test that the average weight of potato is increased. So our claim is
“average weight of potato has increased” i.e. µ > 50 and its
complement is µ ≤ 50. Since complement contains equality sign so
we can take the complement as the null hypothesis and claim as the
alternative hypothesis, that is,
H0 : µ ≤ 50 gm and H1: µ > 50 gm [Here, θ = µ]
Since the alternative hypothesis is right-tailed, so our test is right-
tailed.
Step II: After setting the null and alternative hypotheses, we fix level of
significance α. Suppose, α = 0.01 (= 1 % level).
Step III: Define a test statistic to test the null hypothesis as
Statistic  Value of the parameter under H 0
Test staistic 
Standard error of statistic
X  50
T
σ/ n
Since sample size is large (n = 50 > 30) so by central limit theorem
the sampling distribution of test statistic approximately follows
standard normal distribution (as explained in Unit 1 of this course),
i.e. T ~ N(0,1)
Step IV: Calculate the value of test statistic on the basis of sample
observations as
52  50 2
T   2.78
5.1/ 50 0.72
Step V: Now, we find the critical value. The critical value or cut-off value for
standard normal distribution is given in Table I (Z-table) in the
Appendix at the end of Block 1 of this course. So from this table, the
critical value for right-tailed test at  = 0.01 is zα = 2.33.
19
Testing of Hypothesis Step IV: Now, to take the decision about the null hypothesis, we compare the
calculated value of test statistic with the critical value.
Since calculated value of test statistic (= 2.78) is greater than critical
value (= 2.33), that means calculated value of test statistic lies in
rejection region at 1% level of significance as shown in Fig. 9.5. So
we reject null hypothesis and support the alternative hypothesis. Since
alternative hypothesis is our claim, so we support the claim.
Thus, we conclude that sample does not provide us sufficient
evidence against the claim so we may assume that the average weight
of potato has increased.
Fig. 9.5 Now, you can try the following exercise.
E9) What is the first step in testing of hypothesis?

9.8 CONCEPT OF p-VALUE

In Note 3 of Section 9.6, we promised that p-value will be discussed in Section
9.8. So, it is the time to keep our promise. Nowadays use of p-value is
becoming more and more popular because of the following two reasons:
 most of the statistical software provides p-value rather than critical value.
 p-value provides more information compared to critical value as far as
rejection or do not rejection of H 0 .
The first point listed above needs no explanation. But second point lies in the
heart of p-value and needs to explain more clearly. Moving in this direction, we
note that in scientific applications one is not only interested simply in rejecting
or not rejecting the null hypothesis but he/she is also interested to assess how
strong the data has the evidence to reject H0. For example, as we have seen in
Example 4 based on general procedure of testing a hypothesis where we tested
the null hypothesis
H0 :  ≤ 50 gm against H1:  > 50 gm
To test the null hypothesis, we calculated the value of test statistic as 2.78 and
the critical value (zα) at  = 0.01 was zα = 2.33.
Since calculated value of test statistic (= 2.78) is greater than critical
(tabulated) value (= 2.33), therefore, we reject the null hypothesis at 1% level
of significance.
Now, if we reject the null hypothesis at this level (1%) surely we have to reject
it at higher level because at α = 0.05, zα = 1.645 and at α = 0.10, zα = 1.28.
However, the calculated value of test statistic is much higher than 1.645 and
1.28, therefore, the question arises “could the null hypothesis also be rejected at
values of α smaller than 0.01?” The answer is “yes” and we can compute the
smallest level of significance (α) at which a null hypothesis can be rejected.
This smallest level of significance (α) is known as “p-value”.
The p-value is the smallest value of level of significance(α) at which a null
hypothesis can be rejected using the obtained value of the test statistic and can
be defined as:
The p-value is the probability of obtaining a test statistic equal to or more
extreme (in the direction of sporting H1) than the actual value obtained when
null hypothesis is true.
The p-value also depends on the type of the test. If test is one-tailed then the p- Concepts of Testing of
value is defined as: Hypothesis

For right-tailed test:

p-value = P[Test Statistic (T) ≥ observed value of the test statistic]
For left-tailed test:
p-value = P[Test Statistic (T) ≤ observed value of the test statistic]
If test is two-tailed then the p-value is defined as:
For two-tailed test:
p-value = 2P T  observed value of the test statistic 

Procedure of taking the decision about the null hypothesis on the basis of
p-value:
To take the decision about the null hypothesis based on p-value, the p-value is
compared with level of significance (α) and if p-value is equal or less than 
then we reject the null hypothesis and if the p-value is greater than  we do not
reject the null hypothesis.
The p-value for various tests can be obtained with the help of the tables given
in the Appendix of the Block 1 of this course. But unless we are dealing with
the standard normal distribution, the exact p-value is not obtained with the
tables as mentioned above. But if we test our hypothesis with the help of
computer packages or softwares such as SPSS, SAS, MINITAB, STATA,
EXCEL, etc. then these types of computer packages or softwares present the p-
value as part of the output for each hypothesis testing procedure. Therefore, in
this block we will also describe the procedure to take the decision about the
null hypothesis on the basis of critical value as well as p-value concepts.

9.9 RELATION BETWEEN CONFIDENCE

INTERVAL AND TESTING OF HYPOTHESIS
In Units 7 and 8 of this course, we have learned about confidence intervals
which were used to estimate the unknown population parameters with certain
confidence. In Section 9.7, we have discussed the general procedure of testing
a hypothesis which has been used in making decision about the specified/
assumed/ hypothetical values of population parameters. Both confidence
interval and hypothesis testing have been used for different purposes but have
been based on the same set of concepts. Therefore, there is an extremely close
relationship between confidence interval and hypothesis testing.
In confidence interval, if we construct (1−α) 100% confidence interval for an
unknown parameter then this interval contained all probable values for the
parameter being estimated and relatively improbable values are not contained
by this interval.
So this concept can also be used in hypothesis testing. For this, we contract an
appropriate (1−α) 100% confidence interval for the parameter specified by the
null hypothesis as we have discussed in Units 7 and 8 of this course and if the
value of the parameter specified by the null hypothesis lies in this confident
interval then we do not reject the null hypothesis and if this specified value
does not lie in this confidence interval then we reject the null hypothesis.
Therefore, three cases may arise:
21
Testing of Hypothesis Case I: If we want to test the null hypothesis H0: θ = θ0 against the alternative
hypothesis H1 : θ ≠ θ0 at 5% or 1% level of significance then we
construct two-sided  1  α 100%  95% or 99% confidence interval
for the parameter θ. And we have 95% or 99% (as may be the case)
confidence that this interval will include the parameter value θ0. If the
value of the parameter specified by the null hypothesis i.e. θ0 lies in
this confidence interval then we do not reject the null hypothesis
otherwise we reject the null hypothesis.
Case II: If we want to test the null hypothesis H0 : θ ≤ θ0 against the alternative
hypothesis H1: θ > θ0 then we construct the lower one-sided
confidence bound for parameter θ. If the value of the parameter
specified by the null hypothesis i.e. θ0 is greater than or equal to this
lower bound then we do not reject the null hypothesis otherwise we
reject the null hypothesis.
Case III: If we want to test the null hypothesis H0: θ ≥ θ0 against the
alternative hypothesis H1: θ < θ0 then we construct the upper one-
sided confidence bound for parameter θ. If the value of the parameter
specified by the null hypothesis i.e. θ0 is less than or equal to this
upper bound then we do not reject the null hypothesis otherwise we
reject the null hypothesis.
For example, referring back to Example 4 of this unit, here we want to test the
null hypothesis
H 0 : µ  50gm against H1 : µ  50gm
This was tested with the help of test statistic
X µ
T  Here,θ  μ 
σ/ n
and we reject the null hypothesis at  = 0.01.
This problem could also have been solved by obtaining confidence interval
estimate of population mean which is described in Section 7.4 of Unit 7.
Here, we are given that
n  50, X  50 and σ  5.1
Since alternative hypothesis is right-tailed, therefore, we construct lower one-
sided confidence bound for population mean.
Since population variance is known so lower one-sided (1−α) 100 %
confidence bound for population mean when population variance is known is
given by

X  z
n
Since we test our null hypothesis at  = 0.01 therefore, we contract 99% lower
confidence bound and for  = 0.01, we have z   z0.01  2.33.
Thus, lower one-sided 99% confidence bound for average weight of potatoes is
5.1
52  2.33  52  1.68  50.32
50
Since the value of the parameter specified by the null hypothesis i.e. µ = 50 is
less than lower bound for average weight of potato so we reject the null
22
hypothesis. Concepts of Testing of
Hypothesis
Thus, we can use three approaches (critical value, p-value and confidence
interval) for taking decision about null hypothesis.
With this we end this unit. Let us summarise what we have discussed in this
unit.

9.10 SUMMARY
In this unit, we have covered the following points:
1. Statistical hypothesis, null hypothesis, alternative hypothesis, simple &
composite hypotheses.
2. Type-I and Type-II errors.
3. Critical region.
4. One-tailed and two-tailed tests.
5. General procedure of testing a hypothesis.
6. Level of significance.
7. Concept of p-value.
8. Relation between confidence interval and testing of hypothesis.

9.11 SOLUTIONS / ANSWERS

E1) Here, we wish to test the claim of the company that the average life of
its car tyres is 50000 miles so
Claim: µ = 50000 miles and complement: µ ≠ 50000 miles
Since claim contains equality sign so we take claim as the null
hypothesis and complement as the alternative hypothesis i.e.
H0 :  = 50000 miles
H1 :   50000 miles
E2) (iii) Here, doctor wants to test whether new medicine is really more
effective for controlling blood pressure than old medicine so
Claim: µ1 > µ2 and complement: µ1 ≤ µ2
Since complement contains equality sign so we take complement as the
null hypothesis and claim as the alternative hypothesis i.e.
H0 : µ1 ≤ µ2
H1 : µ1 > µ2
(iv) Here, economist wants to test whether the variability in incomes
differ in two populations so
Claim: 12  22 and complement: 12  22
Since complement contains equality sign so we take complement as the
null hypothesis and claim as the alternative hypothesis i.e.
H0 : 12  22

H1 : 12  22

23
Testing of Hypothesis (v) Here, psychologist wants to test whether the proportion of literates
between two groups of people is same so
Claim: P1 = P2 and complement: P1 ≠ P2
Since claim contains equality sign so we take claim as the null
hypothesis and complement as the alternative hypothesis i.e.
H0 : P1 = P2
H1 : P1 ≠ P2
E3) Here, (i) and (vi) represent the simple hypotheses because these
hypotheses tell us the exact values of parameter average weight of
orange  as  = 100 and  = 130.
The rest (ii), (iii), (iv), (v) and (vii) represent the composite hypotheses
because these hypotheses do not tell us the exact values of parameter .
E4) Since alternative hypothesis H1 : θ ≠ 60 is two tailed so critical region
lies in two-tails.
E5) Let A and B denote the number of white balls and black balls in the urn
respectively. Further, let X be the number of white balls drawn among
the two balls from the urn then we can take the null and alternative
hypotheses as
H0 : A = 4 & B = 2 and H1: A = 2 & B = 4
The critical region is given by
w  X : X  2  X : X  0,1
Thus,
 = P [Reject H0 when H0 is true]
 P  X  w / H0   PX  0 / H0   PX  1/ H0 
4
C0 2 C0 4
C1 2 C1 1  1 4  2 1 8
 6
 6
   
C2 C2 15 15 15 15
9 3
α 
15 5
Similarly,
 = P [Do not reject H0 when H1 is true]
2
C 2 4 C0 1  1 1
 P  X  w / H1   PX  2 / H1   6
 
C2 15 15
E6) Since level of significance is the probability of type-I error so in this
case level of significance is 0.05 or 5%.
E7) Here, the alternative hypothesis is two-tailed therefore, the test will be
two-tailed test.
E8) Whether the test of testing a hypothesis is one-tailed or two-tailed
depends on the alternative hypothesis. So correct option is (ii).
E9) First step in testing of hypothesis is to setup null and alternative
hypotheses.

24
UNIT 10 LARGE SAMPLE TESTS
Structure
10.1 Introduction
Objectives
10.2 Procedure of Testing of Hypothesis for Large Samples
10.3 Testing of Hypothesis for Population Mean Using Z-Test
10.4 Testing of Hypothesis for Difference of Two Population Means Using
Z-Test
10.5 Testing of Hypothesis for Population Proportion Using Z-Test
10.6 Testing of Hypothesis for Difference of Two Population Proportions
Using Z-Test
10.7 Testing of Hypothesis for Population Variance Using Z-Test
10.8 Testing of Hypothesis for Two Population Variances Using Z-Test
10.9 Summary
10.10 Solutions /Answers

10.1 INTRODUCTION
In previous unit, we have defined basic terms used in testing of hypothesis.
After providing you necessary material required for any test, we can move
towards discussing particular tests one by one. But before doing that let us tell
you the strategy we are adopting here.
First we categories the tests under two heads:
 Large sample tests
 Small sample tests
After that, their unit wise distribution is done. In this unit, we will discuss large
sample tests whereas in Units 11 and 12 we will discuss small sample tests.
The tests which are described in these units are known as “parametric tests”.
Sometimes in our studies in the fields of economics, psychology, medical, etc.
we take a sample of objects / units / participants / patients, etc. such as 70, 500,
1000, 10,000, etc. This situation comes under the category of large samples.
As a thumb rule, a sample of size n is treated as a large sample only if it
contains more than 30 units (or observations, n > 30). And we know that, for
large sample (n > 30), one statistical fact is that almost all sampling
distributions of the statistic(s) are closely approximated by the normal
distribution. Therefore, the test statistic, which is a function of sample
observations based on n > 30, could be assumed follow the normal distribution
approximately (or exactly).
But story does not end here. There are some other issues which need to be
taken care off. Some of these issues have been highlighted by making different
cases in each test as you will see when go through Sections 10.3 to 10.8 of this
unit.
This unit is divided into ten sections. Section 10.1 is introductory in nature.
General procedure of testing of hypothesis for large samples is described in
25
Testing of Hypothesis Section 10.2. In Section 10.3, testing of hypothesis for population mean is
discussed whereas in Section 10.4, testing of hypothesis for difference of two
population means with examples is described. Similarly, in Sections 10.5 and
10.6, testing of hypothesis for population proportion and difference of two
population proportions are explained respectively. Testing of hypothesis for
population variance and two population variances are described in Sections
10.7 and 10.8 respectively. Unit ends by providing summary of what we have
discussed in this unit in Section 10.9 and solution of exercises in Section 10.10.

Objectives
After studying this unit, you should be able to:
 judge for a given situation whether we should go for large sample test or
not;
 Applying the Z-test for testing the hypothesis about the population mean
and difference of two population means;
 Applying the Z-test for testing the hypothesis about the population
proportion and difference of two population proportions; and
 Applying the Z-test for testing the hypothesis about the population variance
and two population variances.

10.2 PROCEDURE OF TESTING OF HYPOTHESIS

FOR LARGE SAMPLES
As we have described in previous section that for large sample size (n > 30),
one statistical fact is that almost all sampling distributions of the statistic(s) are
closely approximated by the normal distribution. Therefore, when sample size
is large one can apply the normal distribution based test procedures to test the
hypothesis.
In previous unit, we have given the procedure of testing of hypothesis in
general. Let us now discuss the procedure of testing a hypothesis for large
samples in particular.
Suppose X1, X2, …, Xn is a random sample of size n (> 30) selected from a
population having unknown parameter  and we want to test the hypothesis
about the hypothetical / claimed / assumed value θ0 of parameter θ. For this, a
test procedure is required. We discuss it step by step as follows:
Step I: First of all, we have to setup null hypothesis H0 and alternative
hypothesis H1. Here, we want to test the hypothetical / claimed /
assumed value θ0 of parameter θ. So we can take the null and
alternative hypotheses as
H 0 :   0 and H1 :   0  for two-tailed test 
H 0 :   0 and H 1 :   0 
or  for one-tailed test 
H 0 :   0 and H 1 :   0 
In case of comparing same parameter of two populations of interest,
say, 1 and 2, then our null and alternative hypotheses would be
H 0 : 1  2 and H1 : 1  2 for two-tailed test 
26
 H 0 : 1  2 and H 1 : 1  2  Large Sample Tests
or   for one-tailed test 
H 0 : 1  2 and H 1 : 1  2 

Step II: After setting the null and alternative hypotheses, we have to choose
level of significance. Generally, it is taken as 5% or 1% (α = 0.05 or
0.01). And accordingly rejection and non-rejection regions will be
decided.
Step III: Third step is to determine an appropriate test statistic, say, Z in case
of large samples. Suppose Tn is the sample statistic such as sample
mean, sample proportion, sample variance, etc. for the parameter 
then for testing the null hypothesis, test statistic is given by
 we know that SE of a statistic is 
Tn  E(Tn ) Tn  E(Tn )  theSD of the sampling distribution 
Z 
SE(Tn ) Var(Tn )  of that statistic 
SE(Tn )  SD(Tn )  Var(Tn ) 

where, E(Tn) is the expectation (or mean) of Tn and Var(Tn) is

variance of Tn.
Step IV: As already mentioned for large samples, statistical fact is that almost
all sampling distributions of the statistic(s) are closely approximated
by the normal distribution as the parent population is normal or non-
normal. So, the test statistic Z will assumed to be approximately
normally distributed with mean 0 and variance 1 as
Tn  E(Tn )
Z ~ N(0,1)
Var(Tn )

By putting the values of Tn, E(Tn) and Var(Tn) in above formula we

calculate the value of test statistic Z. Let z be the calculated value of
test statistic Z.
Step V: After that, we obtain the critical (cut-off or tabulated) value(s) in the
sampling distribution of the test statistic Z corresponding to 
assumed in Step II. These critical values are given in Table-I
(Z-table) at the Appendix of Block 1 of this course corresponding to
different level of significance (α). For convenient some useful critical
values at α = 0.01, 0.05 for Z-test are given in Table 10.1 in this
section. After that, we construct rejection (critical) region of size α in
the probability curve of the sampling distribution of test statistic Z.
Step VI: Take the decision about the null hypothesis based on the calculated
and critical values of test statistic obtained in Step IV and Step V.
Since critical value depends upon the nature of the test that it is one-
tailed test or two-tailed test so following cases arise:
In case of one-tailed test:
Case I: When H 0 :   0 and H 1 :   0 (right-tailed test)

In this case, the rejection (critical) region falls under the right tail of
the probability curve of the sampling distribution of test statistic Z.
Fig. 10.1
Suppose z is the critical value at  level of significance so entire
region greater than or equal to z is the rejection region and less than
z is the non-rejection region as shown in Fig. 10.1.
Testing of Hypothesis If z (calculated value ) ≥ z (tabulated value), that means the
calculated value of test statistic Z lies in the rejection region, then we
reject the null hypothesis H0 at  level of significance. Therefore, we
conclude that sample data provides us sufficient evidence against the
null hypothesis and there is a significant difference between
hypothesized or specified value and observed value of the parameter.
If z < z , that means the calculated value of test statistic Z lies in non-
rejection region, then we do not reject the null hypothesis H0 at 
level of significance. Therefore, we conclude that the sample data
fails to provide us sufficient evidence against the null hypothesis and
the difference between hypothesized value and observed value of the
parameter due to fluctuation of sample.
so the population parameter θ may be 0.
Case II: When H 0 :    0 and H 1 :   0 (left-tailed test)

In this case, the rejection (critical) region falls under the left tail of the
probability curve of the sampling distribution of test statistic Z.
Suppose -z is the critical value at  level of significance then entire
region less than or equal to -z is the rejection region and greater
than -z is the non-rejection region as shown in Fig. 10.2.

Fig. 10.2 If z ≤-z, that means the calculated value of test statistic Z lies in the
rejection region, then we reject the null hypothesis H0 at  level of
significance.
If z >-z, that means the calculated value of test statistic Z lies in the
non-rejection region, then we do not reject the null hypothesis H0 at 
level of significance.
In case of two-tailed test: When H 0 :   0 and H 1 :   0

In this case, the rejection region falls under both tails of the
probability curve of sampling distribution of the test statistic Z. Half
the area (α) i.e. α/2 will lies under left tail and other half under the
right tail. Suppose  zα / 2 and zα / 2 are the two critical values at the
left-tailed and right-tailed respectively. Therefore, entire region less
than or equal to z  / 2 and greater than or equal to zα / 2 are the
rejection regions and between  zα / 2 and zα / 2 is the non-rejection
Fig. 10.3
region as shown in Fig. 10.3.

If z  z / 2 or z   z / 2 , that means the calculated value of test

statistic Z lies in the rejection region, then we reject the null
hypothesis H0 at  level of significance.

If zα/2  z  zα/ 2 , that means the calculated value of test statistic Z

lies in the non-rejection region, then we do not reject the null
hypothesis H0 at  level of significance.
Table 10.1 shows some commonly used critical (cut-off or tabulated) values
for one-tailed test and two-tailed test at different level of significance (α) for
Z-test.
 Table 10.1: Critical Values for Z-test Large Sample Tests

Level of Two-Tailed Test One-Tailed Test

Significance (α)
Right-Tailed Test Left- Tailed Test

α = 0.05 (= 5%) ± zα/2 = ± 1.96 zα = 1.645 − zα = − 1.645

α = 0.01 (= 1%) ± zα/2 = ± 2.58 zα = 2.33 − zα = − 2.33

Note 1: As we have discussed in Step IV of this procedure that when sample

size is large then test statistic follows the normal distribution as the parent
population is normal or non-normal so we do not require any assumption of the
form of the parent population for large sample size but when sample size is
small (n < 30) then for applying parametric test we must require the assumption
that the population is normal as we shall in Units 11 and 12. If this assumption
is not fulfilled then we apply the non-parametric tests which will be discussed
in Block 4 of this course.
Decision making procedure about the null hypothesis using the concept of
p-value:
To take the decision about the null hypothesis on the basis of p-value, the p-
value is compared with given level of significance (α) and if p-value is less
than or equal to  then we reject the null hypothesis and if the p-value is
greater than  we do not reject the null hypothesis.
Since test statistic Z follows approximately normal distribution with mean 0
and variance unity, i.e. standard normal distribution and we also know that
standard normal distribution is symmetrical about Z = 0 line therefore, if z
represents the calculated value of Z then p-value can be calculated as follows:
For one-tailed test:
For H1: θ > θ0 (right-tailed test)
p-value = P[Z  z]
For H1: θ < θ0 (left-tailed test)
p-value = P[Z  z]
For two-tailed test:
For H1:   0
p-value = 2P  Z  z 

These p-values for Z-test can be obtained with the help of Table-I (Z-table)
given in the Appendix at the end of Block 1 of this course (which gives the
probability [0  Z  z] for different value of z) as discussed in Unit 14 of
MST-003.
For example, if test is right-tailed and calculated value of test statistic Z is 1.23
then
p-value = P  Z  z   P  Z  1.23  0.5  P  0  Z  1.23

 0.5  0.3907  From Z-table given in Appendix 

of Block 1of this course. 
= 0.1093
29
Testing of Hypothesis Now, you can try the following exercises.
E1) If an investigator observed that the calculated value of test statistic lies in
non-rejection region then he/she will
(i) reject the null hypothesis
(ii) accept the null hypothesis
(iii) not reject the null hypothesis
Write the correct option.
E2) If we have null and alternative hypotheses as
H0:  = 0 and H1:   0
then the rejection (critical) region lies in
(i) left tail
(ii) right tail
(iii) both tails
Write the correct option.
E3) If test is two-tailed and calculated value of test statistic Z is 2.42 then
calculate the p-value for the Z-test.

10.3 TESTING OF HYPOTHESIS FOR

POPULATION MEAN USING Z-TEST
In previous section, we have discussed the general procedure for Z-test. Now
we are discussing the Z-test for testing the hypothesis or claim about the
population mean when sample size is large. Let population under study has
mean  and variance σ2, where  is unknown and σ2 may be known or
unknown. We will consider both cases under this heading. For testing a
hypothesis about population mean we draw a random sample X1, X2,…,Xn of
size n  30 from this population. As we know that for drawing the inference
about the population mean we generally use sample mean and for test statistic,
we require the mean and standard error of sampling distribution of the statistic
(mean). Here, we are considering large sample so we know by central limit
theorem that sample mean is asymptotically normally distributed with mean 
and variance σ2/n whether parent population is normal or non-normal. That
is, if X is the sample mean of the random sample then
σ2
E  X   µ , Var  X   … (1)
n
But we know that standard error = Variance

σ
 SE  X   Var  X   … (2)
n
Now, follow the same procedure as we have discussed in previous section, that
is, first of all we have to setup null and alternative hypotheses. Since here we
want to test the hypothesis about the population mean so we can take the null
and alternative hypotheses as

30
 Here,    and 0  0  Large Sample Tests
H0 :   0 and H1 :   0 for two-tailed test if we compareit with 
general procedure. 
H0 :   0 and H1 :   0 
or  for one-tailed test 
H0 :   0 and H1 :   0 
When we assume that
For testing the null hypothesis, the test statistic Z is given by the null hypothesis is
true then we are actually
X  E X  assuming that the
Z
SE  X  population parameter is
equal to the value in the
X  µ0 null hypothesis. For
Z  Using equations (1) and (2) and  example, we assume that
σ/ n  under H 0 we assume that µ  µ 0 .  µ = 60 whether the null
hypothesis is µ = 60 or
The sampling distribution of the test statistic depends upon σ2 that it is known µ ≤ 60 or µ ≥ 60.
or unknown. Therefore, two cases arise:
Case I: When σ2 is known
In this case, the test statistic follows the normal distribution with
mean 0 and variance unity when the sample size is the large as the
population under study is normal or non-normal. If the sample size is
small then test statistic Z follows the normal distribution only when
population under study is normal. Thus,
X  µ0
Z ~ N  0, 1 
σ/ n
Case II: When σ2 is unknown
In this case, we estimate σ2 by the value of sample variance (S2)
where,
1 n 2
S2  
n  1 i1
 Xi  X 

X  0
Then become test statistic follows the t-distribution with
S/ n
(n−1) df as the sample size is large or small provided the population
under study follows normal as we have discussed in Unit 2 of this
course. But when population under study is not normal and sample
size is large then this test statistic approximately follows normal
distribution with mean 0 and variance unity, that is,
X  0
Z ~ N  0,1
S/ n
After that, we calculate the value of test statistic as may be the case (σ2 is
known or unknown) and compare it with the critical value given in Table 10.1
at prefixed level of significance α. Take the decision about the null hypothesis
as described in the previous section.
From above discussion of testing of hypothesis about population mean, we note
following point:
(i) When σ2 is known then we apply the Z-test as the population under study
is normal or non-normal for the large sample. But when sample size is

31
Testing of Hypothesis small then we apply the Z-test only when population under study is
normal.
(ii) When σ2 is unknown then we apply the t-test only when the population
under study is normal as sample size is large or small. But when the
assumption of normality is not fulfilled and sample size is large then we
can apply the Z-test.
(iii) When sample is small and σ2 is known or unknown and the form of the
population is not known then we apply the non-parametric test as we will
be discussed in Block 4 of this course.
Following examples will help you to understand the procedure more clearly.
Example 1: A light bulb company claims that the 100-watt light bulb it sells
has an average life of 1200 hours with a standard deviation of 100 hours. For
testing the claim 50 new bulbs were selected randomly and allowed to burn
out. The average lifetime of these bulbs was found to be 1180 hours. Is the
company’s claim is true at 5% level of significance?
Solution: Here, we are given that
Specified value of population mean = 0 = 1200 hours,
Population standard deviation = σ = 100 hours,
Sample size = n = 50
Sample mean = X = 1180 hours.
In this example, the population parameter being tested is population mean i.e.
average life of a bulb (µ) and we want to test the company’s claim that average
life of a bulb is 1200 hours. So our claim is  = 1200 and its complement is
 ≠ 1200. Since claim contains the equality sign so we can take the claim as the
null hypothesis and complement as the alternative hypothesis. So
H 0 :    0  1200 average life of a bulb is 1200 hours 
H 1 :   1200 average life of a bulb is not1200 hours 
Also the alternative hypothesis is two-tailed so the test is two-tailed test.
Here, we want to test the hypothesis regarding mean when population SD
(variance) is known and sample size n = 50(> 30) is large. So we will go for
Z-test.
Thus, for testing the null hypothesis the test statistic is given by
X  0
Z
/ n
1180  1200 20
   1.41
100 / 50 14.14
The critical (tabulated) values for two-tailed test at 5% level of significance are
± zα/2 = ± z0.025 = ± 1.96.
Fig. 10.4 Since calculated value of test statistic Z ( = –1.41) is greater than critical value
(= − 1.96) and less than the critical value (= 1.96), that means it lies in non-
rejection region as shown in Fig. 10.4, so we do not reject the null hypothesis.
Since the null hypothesis is the claim so we support the claim at 5% level of
significance.
Decision according to p-value:
The test is two-tailed, therefore,
p-value = 2P Z  z   2P  Z  1.41 Large Sample Tests

 2  0.5  P  0  Z  1.41  2  0.5  0.4207   0.1586

Since p-value (= 0.1586) is greater than α (= 0.05) so we do not reject the null
hypothesis at 5% level of significance.
Decision according to confidence interval:
Here, test is two-tailed, therefore, we contract two-sided confidence interval for
population mean.
Since population standard deviation is known, therefore, we can use
(1−α) 100 % confidence interval for population mean when population
variance is known which is given by
   
X  z  / 2 n , X  z / 2 n 
 

Here,  = 0.05, we have z /2  z0.025  1.96.

Thus, 95% confidence interval for average life of a bulb is given by

 100 100 
1180  1.96 50 ,1180  1.96 50 
 
or 1180  27.71,1180  27.71 
or 1152.29, 1207.71
Since 95% confidence interval for average life of a bulb contains the value of
the parameter specified by the null hypothesis, that is,    0  1200 so we do
not reject the null hypothesis.
Thus, we conclude that sample does not provide us sufficient evidence against
the claim so we may assume that the company’s claim that the average life of a
bulb is 1200 hours is true.
Note 2: Here, we note that the decisions about null hypothesis based on three
approaches (critical value or classical, p-value and confidence interval) are
same. The learners are advised to make the decision about the claim or
statement by using only one of the three approaches in the examination. Here,
we used all these approaches only to give you an idea how they can be used in
a given problem. Those learners who will opt biostatistics specialisation will
see and realize the importance of confidence interval approach in Unit 16 of
MSTE-004.
Example 2: A manufacturer of ball point pens claims that a certain pen
manufactured by him has a mean writing-life at least 460 A-4 size pages. A
purchasing agent selects a sample of 100 pens and put them on the test. The
mean writing-life of the sample found 453 A-4 size pages with standard
deviation 25 A-4 size pages. Should the purchasing agent reject the
manufacturer’s claim at 1% level of significance?
Solution: Here, we are given that
Specified value of population mean = 0 = 460,

33
Testing of Hypothesis Sample size = n = 100,
Sample mean = X = 453,
Sample standard deviation = S = 25
Here, we want to test the manufacturer’s claim that the mean writing-life (µ) of
pen is at least 460 A-4 size pages. So our claim is  ≥ 460 and its complement
is  < 460. Since claim contains the equality sign so we can take the claim as
the null hypothesis and the complement as the alternative hypothesis. So
H 0 :    0  460 and H 1 :   460

Also the alternative hypothesis is left-tailed so the test is left-tailed test.

Here, we want to test the hypothesis regarding population mean when
population SD is unknown. So we should used t-test for if writing-life of pen
follows normal distribution. But it is not the case. Since sample size is n = 100
(n > 30) large so we go for Z-test. The test statistic of Z-test is given by
X  0
Z
S/ n
453  460 7
   2.8
25 / 100 2.5
We have from Table 10.1 the critical (tabulated) value for left-tailed Z test at
1% level of significance is zα = −2.33.
Since calculated value of test statistic Z (= ‒2.8) is less than the critical value
(= −2.33), that means calculated value of test statistic Z lies in rejection region
Fig. 10.5 as shown in Fig. 10.5, so we reject the null hypothesis. Since the null
hypothesis is the claim so we reject the manufacturer’s claim at 1% level of
significance.
Decision according to p-value:
The test is left-tailed, therefore,

p-value = P  Z  z  P Z  2.8  P Z  2.8  Z is symetrical 

about Z  0 line 

 0.5  P 0  Z  2.8  0.5  0.4974  0.0026

Since p-value (= 0.0026) is less than α (= 0.01) so we reject the null hypothesis
at 1% level of significance.
Therefore, we conclude that the sample provide us sufficient evidence against
the claim so the purchasing agent rejects the manufacturer’s claim at 1% level
of significance.
Now, you can try the following exercises.
E4) A sample of 900 bolts has a mean length 3.4 cm. Is the sample regarded
to be taken from a large population of bolts with mean length 3.25 cm
and standard deviation 2.61 cm at 5% level of significance?
E5) A big company uses thousands of CFL lights every year. The brand that
the company has been using in the past has average life of 1200 hours. A
new brand is offered to the company at a price lower than they are paying
 for the old brand. Consequently, a sample of 100 CFL light of new brand Large Sample Tests
is tested which yields an average life of 1220 hours with standard
deviation 90 hours. Should the company accept the new brand at 5% level
of significance?

10.4 TESTING OF HYPOTHESIS FOR

DIFFERENCE OF TWO POPULATION MEANS
USING Z-TEST
In previous section, we have learnt about the testing of hypothesis about the
population mean. But there are so many situations where we want to test the
hypothesis about difference of two population means or two population means.
For example, two manufacturing companies of bulbs are produced same type
of bulbs and one may be interested to test that one is better than the other, an
investigator may want to test the equality of the average incomes of the peoples
living in two cities, etc. Therefore, we require an appropriate test for testing the
hypothesis about the difference of two population means.
Let there be two populations, say, population-I and population-II under study.
Also let 1 ,  2 and 12 ,  22 denote the means and variances of population-I and
population-II respectively where both 1 and 2 are unknown but 12 and 22 may
be known or unknown. We will consider all possible cases here. For testing the
hypothesis about the difference of two population means, we draw a random
sample of large size n1 from population-I and a random sample of large size n2
from population-II. Let X and Y be the means of the samples selected from
population-I and II respectively.
These two populations may or may not be normal but according to the central
limit theorem, the sampling distribution of difference of two large sample
means asymptotically normally distributed with mean (µ1-µ2) and variance
(σ12 / n1  σ 22 / n 2 ) as described in Unit 2 of this course.
Thus,
E  X  Y   E  X   E  Y   1   2 … (3)

and
12 22
Var  X  Y   Var  X   Var  Y   
n1 n 2
But we know that standard error = Variance

12 22
 SE  X  Y   Var  X  Y    … (4)
n1 n 2
Now, follow the same procedure as we have discussed in Section 10.2, that is,
first of all we have to setup null and alternative hypotheses. Here, we want to
test the hypothesis about the difference of two population means so we can take
the null hypothesis as
 Here, 1  1 and 2   2 
H0 : 1  2 (no difference in means) if we compareit with 
 general procedure. 
35
Testing of Hypothesis or H 0 : 1  2  0 (difference in two means is 0)
and the alternative hypothesis as
H1 : 1  2 for two-tailed test 

H0 : 1  2 and H1 : 1  2 
or  for one-tailed test 
H0 : 1  2 and H1 : 1  2 
For testing the null hypothesis, the test statistic Z is given by

Z
 X  Y   E X  Y 
SE  X  Y 

 X  Y    µ1  µ 2 
or Z  using equations (3) and (4) 
σ12 σ 22

n1 n 2
Since under null hypothesis we assume that µ1 = µ2, therefore, we have
XY
Z
σ12 σ 22

n1 n 2

Now, the sampling distribution of the test statistic depends upon 12 and 22
that both are known or unknown. Therefore, four cases arise:
Case I: When 12 & 22 are known and 12 22  2
In this case, the test statistic follows normal distribution with mean
0 and variance unity when the sample sizes are large as both the
populations under study are normal or non-normal. But when
sample sizes are small then test statistic Z follows normal
distribution only when populations under study are normal, that is,
XY
Z ~ N  0,1 
1 1
σ 
n1 n2
Case II: When 12 & 22 are known and 12  22
In this case, the test statistic also follows the normal distribution as
described in case I, that is,
XY
Z ~ N(0, 1)
σ12 σ 22

n1 n 2

Case III: When 12 & 22 are unknown and 12 22  2

In this case, σ2 is estimated by value of pooled sample variance S2p

where,
1
S2p   n1S12  n 2S22 
n1  n 2  2 
and
36
 n n Large Sample Tests
1 1
2 1 2
2
2
S 
1 
(n1  1) i1
 Xi  X  and S22  
(n 2  1) i1
 Yi  Y 

and test statistic follows t-distribution with (n1 + n2 − 2) degrees of

freedom as the sample sizes are large or small provided populations
under study follow normal distribution as described in Unit 2 of this
course. But when the populations are under study are not normal
and sample sizes n1 and n2are large (> 30) then by central limit
theorem, test statistic approximately normally distributed with mean
0 and variance unity, that is,
XY
Z ~ N  0,1 
1 1
Sp 
n1 n 2
Case IV: When 12 & 22 are unknown and 12  22

In this case, 12 & 22 are estimated by the values of the sample
variances S12 &S22 respectively and the exact distribution of test
statistic is difficult to derive. But when sample sizes n1 and n2 are
large (> 30) then central limit theorem, the test statistic
approximately normally distributed with mean 0 and variance unity,
that is,
XY
Z ~ N(0, 1)
S12 S22

n1 n 2
After that, we calculate the value of test statistic and compare it with the
critical value given in Table 10.1 at prefixed level of significance α. Take the
decision about the null hypothesis as described in Section10.2.
From above discussion of testing of hypothesis about population mean, we note
following point:
(i) When 12 & 22 are known then we apply the Z-test as both the population
under study are normal or non-normal for the large sample. But when
sample sizes are small then we apply the Z-test only when populations
under study are normal.
(ii) When 12 & 22 are unknown then we apply the t-test only when the
populations under study are normal as sample sizes are large or small.
But when the assumption of normality is not fulfilled and sample sizes
are large then we can apply the Z-test.
(iii) When samples are small and 12 & 22 are known or unknown and the
form of the population is not known then we apply the non-parametric
test as we will be discussed in Block 4 of this course.
Let us do some examples based on above test.
Example 3: In two samples of women from Punjab and Tamilnadu, the mean
height of 1000 and 2000 women are 67.6 and 68.0 inches respectively. If
population standard deviation of Punjab and Tamilnadu are same and equal to
5.5 inches then, can the mean heights of Punjab and Tamilnadu women be
regarded as same at 1% level of significance?
37
Testing of Hypothesis Solution: We are given
n1 = 1000, n2 = 2000, X  67.6, Y  68.0 and σ1  σ2  σ  5.5
Here, we wish to test that the mean height of Punjab and Tamilnadu women is
same. If 1 and 2 denote the mean heights of Punjab and Tamilnadu women
respectively then our claim is 1 = 2 and its complement is 1 ≠ 2. Since the
claim contains the equality sign so we can take the claim as the null hypothesis
and complement as the alternative hypothesis. Thus,
H 0 : 1   2 and H1 : 1   2
Since the alternative hypothesis is two-tailed so the test is two-tailed test.
Here, we want to test the hypothesis regarding two population means. The
standard deviations of both populations are known and sample sizes are large,
so we should go for Z-test.
So, for testing the null hypothesis, the test statistic Z is given by
XY
Z
σ12 σ 22

n1 n 2
67.6  68.0  0.4
 2 2

  5.5   5.5   1 1 
 1000  2000  5.5   
 1000 2000 
 
 0.4
  1.88
5.5  0.0387
The critical (tabulated) values for two-tailed test at 1% level of significance are
± zα/2 = ± z0.005 = ± 2.58.
Since calculated value of Z ( = −1.88) is greater than the critical value
(= − 2.58) and less than the critical value (= 2.58), that means it lies in non-
rejection region as shown in Fig. 10.6, so we do not reject the null hypothesis
i.e. we fail to reject the claim.
Decision according to p-value:
Fig. 10.6
The test is two-tailed, therefore,
p-value = 2P  Z  z   2P  Z  1.88

 2  0.5  P  0  Z  1.88  2  0.5  0.4699   0.0602

Since p-value (= 0.0602) is greater than  (  0.01) so we do not reject the null
hypothesis at1% level of significance.
Thus, we conclude that the samples do not provide us sufficient evidence
against the claim so we may assume that the average height of women of
Punjab and Tamilnadu is same.
Example 4: A university conducts both face to face and distance mode classes
for a particular course indented both to be identical. A sample of 50 students of
face to face mode yields examination results mean and SD respectively as:
X  80.4, S1  12.8
and other sample of 100 distance-mode students yields mean and SD of their Large Sample Tests
examination results in the same course respectively as:
Y  74.3, S2  20.5
Are both educational methods statistically equal at 5% level?
Solution: Here, we are given that
n1  50, X  80.4, S1  12.8;

n 2  100 , Y  74.3, S2  20.5

We wish to test that both educational methods are statistically equal. If 1 and
2 denote the average marks of face to face and distance mode students
respectively then our claim is 1 = 2 and its complement is 1 ≠ 2. Since the
claim contains the equality sign so we can take the claim as the null hypothesis
and complement as the alternative hypothesis. Thus,
H 0 : 1   2 and H1 : 1   2
Since the alternative hypothesis is two-tailed so the test is two-tailed test.
We want to test the null hypothesis regarding two population means when
standard deviations of both populations are unknown. So we should go for t-
test if population of difference is known to be normal. But it is not the case.
Since sample sizes are large (n1, and n2 > 30) so we go for Z-test.
For testing the null hypothesis, the test statistic Z is given by
XY
Z
S12 S22

n1 n 2

80.4  74.3 6.1

   2.23
12.8 
2
 20.5 
2 3.28  4.20
 Fig. 10.7
50 100
The critical (tabulated) values for two-tailed test at 5% level of significance are
± zα/2 = ± z0.025 = ± 1.96.
Since calculated value of Z ( = 2.23) is greater than the critical values
(= ±1.96 ), that means it lies in rejection region as shown in Fig. 10.7, so we
reject the null hypothesis i.e. we reject the claim at 5% level of significance.
Decision according to p-value:
The test is two-tailed, therefore,

p-value = 2P  Z  z   2P  Z  2.23

 2  0.5  P 0  Z  2.23  2  0.5  0.4871  0.0258

Since p-value (= 0.0258) is less than  (  0.05) so we reject the null hypothesis
at 5% level of significance.
Thus, we conclude that samples provide us sufficient evidence against the
claim so both methods of education, i.e. face-to-face and distance-mode, are
not statistically equal.
Testing of Hypothesis Now, you can try the following exercises.
E6) Two brands of electric bulbs are quoted at the same price. A buyer was
tested a random sample of 200 bulbs of each brand and found the
following information:
Mean Life (hrs.) SD(hrs.)
Brand A 1300 41
Brand B 1280 46
Is there significant difference in the mean duration of their lives of two
brands of electric bulbs at 1% level of significance?
E7) Two research laboratories have identically produced drugs that provide
relief to BP patients. The first drug was tested on a group of 50 BP
patients and produced an average 8.3 hours of relief with a standard
deviation of 1.2 hours. The second drug was tested on 100 patients,
producing an average of 8.0 hours of relief with a standard deviation of
1.5 hours. Does the first drug provide a significant longer period of relief
at a significant level of 5%?

10.5 TESTING OF HYPOTHESIS FOR

POPULATION PROPORTION USING Z-TEST
In Section 10.3, we have discussed the procedure of testing of hypothesis for
population mean when sample size is large. But in many real world situations,
in business and other areas where collected data are in form of counts or the
collected data are classified into two categories or groups according to an
attribute or a characteristic. For example, the peoples living in a colony may be
classified into two groups (male and female) with respect to the characteristic
sex, the patients in a hospital may be classified into two groups as cancer and
non-cancer patients, the lot of articles may be classified as defective and non-
defective, etc. Here, collected data are available in dichotomous or binary
outcomes which is a special case of nominal scale and the data categorized into
two mutually exclusive and exhaustive classes generally known as success and
failure out comes. For example, the characteristic sex can be measured as
success if male and failure if female or vice versa. So in such situations,
proportion is suitable measure to apply.
In such situations, we require a test for testing a hypothesis about population
proportion.
For this purpose, let X1 , X 2 , ..., X n be a random sample of size n taken from a
population with population proportion P. Also let X denotes the number of
observations or elements possess a certain attribute (number of successes) out
of n observations of the sample then sample proportion p can be defined as
X
p 1
n
As we have seen in Section 2.4 of the Unit 2 of this course that mean and
variance of the sampling distribution of sample proportion are
PQ
E(p)  P and Var(p) 
n
where, Q = 1‒ P.
40
Now, two cases arise: Large Sample Tests

Case I: When sample size is not sufficiently large i.e. either of the conditions
np > 5 or nq > 5 does not meet, then we use exact binomial test. But exact
binomial test is beyond the scope of this course.
Case II: When sample size is sufficiently large, such that np > 5 and nq > 5
then by central limit theorem, the sampling distribution of sample proportion p
is approximately normally distributed with mean and variance as
PQ
E(p) = P and Var(p) = … (5)
n
But we know that standard error = Variance

PQ
 SE (p)  … (6)
n
Now, follow the same procedure as we have discussed in Section 10.2, first of
all we setup null and alternative hypotheses. Since here we want to test the
hypothesis about specified value P0 of the population proportion so we can take
the null and alternative hypotheses as
Here,   P and   P0 
H 0 : P  P0 and H1 : P  P0 for two-tailed test   if we compare it with
0

 general procedure. 

H 0 : P  P0 and H1 : P  P0 
or  for one-tailed test 
H 0 : P  P0 and H1 : P  P0 
For testing the null hypothesis, the test statistic Z is given by
p  E p 
Z
SE  p 
p  P0
Z ~ N  0, 1  under H0  using equations(5) and (6) 
P0 Q 0
n
After that, we calculate the value of test statistic and compare it with the
critical value(s) given in Table 10.1 at prefixed level of significance α. Take
the decision about the null hypothesis as described in Section 10.2.
Let us do some examples of testing of hypothesis about population proportion.
Example 5: A machine produces a large number of items out of which 25%
are found to be defective. To check this, company manager takes a random
sample of 100 items and found 35 items defective. Is there an evidence of more
deterioration of quality at 5% level of significance?
Solution: The company manager wants to check that his machine produces
25% defective items. Here, attribute under study is defectiveness. And we
define our success and failure as getting a defective or non defective item.
Let P = Population proportion of defectives items = 0.25(= P0 )
p = Observed proportion of defectives items in the sample = 35/100 = 0.35
Here, we want to test that machine produces more defective items, that is, the
proportion of defective items (P) greater than 0.25. So our claim is P > 0.25
41
Testing of Hypothesis and its complement is P ≤ 0.25. Since complement contains the equality sign so
we can take the complement as the null hypothesis and the claim as the
alternative hypothesis. So
H 0 : P  P0  0.25 and H 1 : P  0.25

Since the alternative hypothesis is right-tailed so the test is right-tailed test.

Before proceeding further, first we have to check whether the condition of
normality meets or not.
 np  100  0.35  35  5
nq  100   1  0.35   100  0.65  65  5

We see that condition of normality meets, so we can go for Z-test.

So, for testing the null hypothesis, the test statistic Z is given by
p  P0
Z
P0Q 0
n
0.35  0.25 0.10
   2.31
0.25  0.75 0.0433
100
Since test is right-tailed so the critical value at 5% level of significance is
zα = z0.05 = 1.645.
Since calculated value of test statistic Z (= 2.31) is greater than the critical
value (= 1.645), that means it lies in the rejection region as shown in Fig. 10.8.
So we reject the null hypothesis and support the alternative hypothesis i.e. we
Fig. 10.8 support the claim at 5% level of significance.
Decision according to p-value:
The test is right-tailed, therefore,

p-value = P  Z  z  P  Z  2.31

 0.5  P 0  Z  2.31  0.5  0.4896

 0.0104
Since p-value (= 0.0104) is less than  (  0.05) so we reject the null
hypothesis at 5% level of significance.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so we may assume that deterioration in quality exists at 5%
level of significance.
Example 6: A die is thrown 9000 times and draw of 2 or 5 is observed 3100
times. Can we regard that die is unbiased at 5% level of significance.
Solution: Let getting a 2 or 5 be our success, and getting a number other than 2
or 5 be a failure then in usual notions, we have
n = 9000, X = number of successes = 3100, p = 3100/9000 = 0.3444
Here, we want to test that the die is unbiased and we know that if die is Large Sample Tests
unbiased then proportion or probability of getting 2 or 5 is
P = Probability of getting a 2 or 5
= Probability of getting 2 + Probability of getting 5
1 1 1
    0.3333
6 6 3
So our claim is P = 0.3333 and its complement is P ≠ 0.3333. Since the claim
contains the equality sign so we can take the claim as the null hypothesis and
complement as the alternative hypothesis. Thus,
H0 : P  P0  0.3333 and H1 :P  0.3333

Since the alternative hypothesis is two-tailed so the test is two-tailed test.

Before proceeding further, first we have to check whether the condition of
normality meets or not.
 np  9000  0.3444  3099.6  5
nq  9000   1  0.3444   9000  0.6556  5900.4  5
We see that condition of normality meets, so we can go for Z-test.
So, for testing the null hypothesis, the test statistic Z is given by
p  P0
Z
P0Q 0
n
0.3444  0.3333 0.0111
   2.22
0.3333  0.6667 0.005
9000
Since test is two-tailed so the critical values at 5% level of significance are
± zα/2 = ± z0.025 = ±1.96.
Since calculated value of Z (= 2.22) is greater than the critical value (= 1.96),
that means it lies in rejection region, so we reject the null hypothesis i.e. we
reject our claim.
Decision according to p-value:
Since test is two-tailed so

p-value = 2P Z  z   2P  Z  2.22

 2 0.5  P  0  Z  2.22  2  0.5  0.4868  0.0264

Since p-value (= 0.0264) is less than  (  0.05), so we reject the null

hypothesis at 5% level of significance.
Thus, we conclude that the sample provides us sufficient evidence against the
claim so die cannot be considered as unbiased.
Now, you can try the following exercises.
E8) In a sample of 100 MSc. Economics first year students of a University, it
was seen that 54 students came from Science background and the rest are
43
Testing of Hypothesis from other background. Can we assume that 50% of the students are from
Science background in MSc. Economics first year students in the
University at 1% level of significance?
E9) Out of 200 patients who are given a particular injection 180 survived.
Test the hypothesis that the survival rate is more than 80% at 5% level of
significance?

10.6 TESTING OF HYPOTHESIS FOR

DIFFERENCE OF TWO POPULATION
PROPORTIONS USING Z-TEST
In Section 10.5, we have discussed the testing of hypothesis about the
population proportion. In some cases, we are interested to test the hypothesis
about difference of two population proportions of an attributes in the two
different populations or groups. For example, one may wish to test whether the
proportions of alcohol drinkers in the two cities are same, one may wish to test
proportion of literates in a group of people is greater than the proportion of
literates in other group of people, etc. Therefore, we require the test for testing
the hypothesis about the difference of two population proportions.
Let there be two populations, say, population-I and population-II under study.
And also let we draw a random sample of size n1 from population-I with
population proportion P1 and a random sample of size n2 from population-II
with population proportion P2. If X1 and X2 are the number of observations
/individuals / items / units possessing the given attribute in the sample of sizes
n1 and n2 respectively then sample proportions can be defined as
X1 X
p1  and p 2  2
n1 n2
As we have seen in Section 2.5 of the Unit 2 of this course that mean and
variance of the sampling distribution of difference of sample proportions are
E(p1-p2) = P1-P2
and variance
P1Q1 P2Q 2
Var(p 1-p2)  
n1 n2

where, Q1 = 1‒ P1 and Q2 = 1‒ P2.

Now, two cases arise:
Case I: When sample sizes are not sufficiently large i.e. any of the conditions
n1p1  5 or n1q1  5 or n 2 p2  5 or n 2q2  5 does not meet, then we use exact
binomial test. But exact binomial test is beyond the scope of this course.
Case II: When sample sizes are sufficiently large, such that n1p1  5, n1q1  5,
n2p2  5 and n2q2  5 then by central limit theorem, the sampling distribution
of sample proportions p 1 and p 2 are approximately normally as

44
  PQ   PQ  Large Sample Tests
p1 ~ N  P1 , 1 1  and p 2 ~ N  P2 , 2 2 
 n1   n2 

Also, by the property of normal distribution described in Unit 13 of MST-003,

the sampling distribution of the difference of sample proportions follows
normal distribution with mean
E(p1-p2) = E(p1)-E(p2) = P1-P2 … (7)
and variance
P1Q1 P2Q 2
Var(p1-p2) = Var(p 1)+Var(p2)  
n1 n2

That is,
 PQ P Q 
p1  p 2 ~ N P1  P2 , 1 1  2 2 
 n1 n2 

Thus, standard error is given by

P1Q1 P2Q 2
SE  p1  p 2   Var  p1  p 2    … (8)
n1 n2

Now, follow the same procedure as we have discussed in Section 10.2, first of
all we have to setup null and alternative hypotheses. Here, we want to test the
hypothesis about the difference of two population proportions so we can take
the null hypothesis as
 Here, 1  P1 and 
H 0 : P1  P2 (no difference in proportions)   2  P2 if we compare 
 it with general 
 procedure. 

or H 0 : P1  P2  0 (difference in two proportions is 0)

and the alternative hypothesis may be
H1 : P1  P2 for two-tailed test 

H 0 : P1  P2 and H1 : P1  P2 
or  for one-tailed test
H 0 : P1  P2 and H1 : P1  P2 

For testing the null hypothesis, the test statistic Z is given by

 p1  p 2   E  p1  p 2 
Z
SE  p1  p 2 

 p1  p2    P1  P2 
or Z  using equations (7) and (8) 
P1Q1 P2Q2

n1 n2
Since under null hypothesis we assume that P1 = P2 = P, therefore, we have

45
Testing of Hypothesis p1  p 2
Z
1 1 
PQ   
 n1 n 2 
where, Q = 1-P.
Generally, P is unknown then it is estimated by the value of pooled proportion
P̂, where
n p  n 2 p 2 X1  X 2 ˆ  1  Pˆ
Pˆ  1 1  and Q
n1  n 2 n1  n 2
After that, we calculate the value of test statistic and compare it with the
critical value(s) given in Table 10.1 at prefixed level of significance α. Take
the decision about the null hypothesis as described in Section 10.2.
Now, it is time for doing some examples for testing of hypothesis about the
difference of two population proportions.
Example 7: In a random sample of 100 persons from town A, 60 are found to
be high consumers of wheat. In another sample of 80 persons from town B, 40
are found to be high consumers of wheat. Do these data reveal a significant
difference between the proportions of high wheat consumers in town A and
town B ( at α = 0.05 )?
Solution: Here, attribute under study is high consuming of wheat. And we
define our success and failure as getting a person of high consumer of wheat
and not high consumer of wheat respectively.
We are given that
n1 = total number of persons in the sample of town A = 100
n2 = total number of persons in the sample of town B = 80
X1 = number of persons of high consumer of wheat in town A = 60
X2 = number of persons of high consumer of wheat in town B = 40
The sample proportion of high wheat consumers in town A is
X1 60
p1    0.60
n1 100
and the sample proportion of wheat consumers in town B is
X 2 40
p2    0.50
n 2 80
Here, we want to test that the proportion of high consumers of wheat in two
towns, say, P1 and P2, is not same. So our claim is P1 ≠ P2 and its complement
is P1 = P2. Since the complement contains the equality sign, so we can take the
complement as the null hypothesis and the claim as the alternative hypothesis.
Thus,
H 0 : P1  P2  P and H 1 : P1  P2

Since the alternative hypothesis is two-tailed so the test is two-tailed test.

Before proceeding further, first we have to check whether the condition of
normality meets or not.

46
 n1p1  100  0.60  60  5, n1q1  100  0.40  40  5 Large Sample Tests

n 2 p2  80  0.50  40  5, n 2q 2  80  0.50  40  5
We see that condition of normality meets, so we can go for Z-test.
The estimate of the combined proportion (P) of high wheat consumers in two
towns is given by
n1p1  n 2p2 X1  X 2 60  40 5
P̂    
n1  n 2 n1  n 2 100  80 9

ˆ  1  Pˆ  1  5  4
Q
9 9
For testing the null hypothesis, the test statistic Z is given by
p1  p 2
Z
ˆ  1  1 
P̂Q
 n1 n 2 
0.60  0.50 0.10
   1.34
5 4 1 1  0.0745
 
9 9  100 80 
The critical values for two-tailed test at 5% level of significance are ± zα/2
= ± z0.025 = ±1.96.
Since calculated value of Z (=1.34) is less than the critical value (= 1.96) and
greater than critical value (= −1.96), that means calculated value of Z lies in
non-rejection region, so we do not reject the null hypothesis and reject the
alternative hypothesis i.e. we reject the claim.
Decision according to p-value:
Since the test is two-tailed, therefore
p-value = 2 P  Z  z   2P  Z  1.34

 2 0.5  P  0  Z  1.34  2  0.5  0.4099   0.1802

Since p-value (= 0.1802) is greater than  (  0.05) so we do not reject the null
hypothesis at 5% level of significance.
Thus, we conclude that the samples provide us the sufficient evidence against
the claim so we may assume that the proportion of high consumers of wheat in
two towns A and B is same.
Example 8: A machine produced 60 defective articles in a batch of 400. After
overhauling it produced 30 defective in a batch of 300. Has the machine
improved due to overhauling? (Take  = 0.01).
Solution: Here, the machine produced articles and attribute under study is
defectiveness. And we define our success and failure as getting a defective or
non defective article. Therefore, we are given that
X1 = number of defective articles produced by the machine before overhauling
= 60
X2 = number of defective articles produced by the machine after overhauling
= 30

47
Testing of Hypothesis and n1  400, n 2  300,
Let p1 = Observed proportion of defective articles in the sample before the
overhauling
X1 60
   0.15
n1 400
and p2 = Observed proportion of defective articles in the sample after the
overhauling
X2 30
   0.10
n 2 300
Here, we want to test that machine improved due to overhauling that means the
proportion of defective articles is less after overhauling. If P1 and P2 denote the
proportion defectives before and after the overhauling the machine so our claim
is P1 > P2 and its complement P1 ≤ P2. Since the complement contains the
equality sign so we can take the complement as the null hypothesis and claim
as the alternative hypothesis. Thus,
H 0 : P1  P2 and H1 : P1  P2
Since the alternative hypothesis is right-tailed so the test is right-tailed test.
Since P is unknown, so the pooled estimate of proportion is given by
X1  X 2 60  30 90 9 ˆ  1  Pˆ  1  9  61 .
P̂     and Q
n1  n 2 400  300 700 70 70 70
Before proceeding further, first we have to check whether the condition of
normality meets or not.
 n1p1  400  0.15  60  5, n1q1  400  0.85  340  5
n 2 p2  300  0.10  30  5, n 2q 2  300  0.90  270  5
We see that condition of normality meets, so we can go for Z-test.
For testing the null hypothesis, the statistic is given by
p1  p 2
Z
ˆ  1  1 
P̂Q
 n1 n 2 
0.15  0.10 0.05
   1.95
9 61  1 1  0.0256
 
70 70  400 300 
The critical value for right-tailed test at 1% level of significance is
zα = z0.01 = 2.33.
Since calculated value of Z (= 1.95) is less than the critical value (= 2.33) that
means calculated value of Z lies in non-rejection region, so we do not reject the
null hypothesis and reject the alternative hypothesis i.e. we reject the claim at
1% level of significance.
Decision according to p-value:
Since the test is right-tailed, therefore,
p-value = P Z  z  P Z  1.95

48
  0.5  P 0  Z  1.95  0.5  0.4744  0.0256 Large Sample Tests

Since p-value (= 0.0256) is greater than  (  0.01) so we do not reject the null
hypothesis at1% level of significance.
Thus, we conclude that the samples provide us sufficient evidence against the
claim so the machine has not been improved after overhauling.
Now, you can try the following exercises.
E10) The proportions of literates between groups of people of two districts A
and B are tested. Out of the 100 persons selected at random from each
district, 50 from district A and 40 from district B are found literates. Test
whether the proportion of literate persons in two districts A and B is
same at 1% level of significance?
E11) In a large population 30% of a random sample of 1200 persons had blue-
eyes and 20% of a random sample of 900 persons had the same blue-
eyes in another population. Test the proportion of blue-eyes persons is
same in two populations at 5% level of significance.

10.7 TESTING OF HYPOTHESIS FOR

POPULATION VARIANCE USING Z-TEST
In Section 10.3, we have discussed testing of hypothesis for population mean
but when analysing quantitative data, it is often important to draw conclusion
about the average as well as the variability of a characteristic of under study.
For example, a company manufactured the electric bulbs and the manager of
the company would probably be interested in determining the average life of
the bulbs and also determining whether or not the variability in the life of bulbs
is within acceptable limits, the product controller of a milk company may be
interested to know variance of the amount of fat in the whole milk processed
by the company is no more than the specified level, etc. So we require a test for
this purpose.
The procedure of testing a hypothesis for population variance or standard
deviation is similar to the testing of population mean.
For testing a hypothesis about the population variance, we draw a random
sample X 1 , X 2 ,..., X n of size n  30 from the population with mean  and
variance σ2 where,  be known or unknown.
We know that by central limit theorem that sample variance is asymptotically
normally distributed with mean σ2 and variance 2σ4/n whether parent
population is normal or non-normal. That is, if S2 is the sample variance of
the random sample then
2σ 4
E(S2 )  σ 2 and Var(S2 ) = … (9)
n

But we know that standard error = Variance

2 2
 SE  S2   Var  S2   σ … (10)
n
The general procedure of this test is explained in the next page.
49
Testing of Hypothesis As we are doing so far in all tests, first Step in hypothesis testing problems is to
setup null and alternative hypotheses. Here, we want to test the hypothesis
specified value 20 of the population variance 2 so we can take our null and
alternative hypotheses as

H 0 : 2  02 and H1 : 2  02 for two-tailed test 

H0 : 2  02 and H1 : 2  02 
or  for one-tailed test 
H0 : 2  02 and H1 : 2  02 

For testing the null hypothesis, the test statistic Z is given by

S2  E  S2 
Z ~ N  0,1 
SE  S2 

S2  σ20  Usingequations(9) and(10) and 

Z 2 2
2  under H0 : σ  σ0 
σ02
n
After that, we calculate the value of test statistic and compare it with the
critical value given in Table 10.1 at prefixed level of significance α. Take the
decision about the null hypothesis as described in Section 10.2.
Note 2: When population under study is normal then for testing the hypothesis
about population variance or population standard deviation we use chi-square
test which will be discussed in Unit 12 of this course. Whereas when the
distribution of the population under study is not known and sample size is large
then we apply Z-test as discussed above.
Now, it is time to do example based on above test.
Example 9: A random sample of size 65 screws is taken from a population of
big box of screws and measured their length (in mm) which gives sample
variance 9.0. Test that the two years old population variance 10.5 is still
maintained at present at 5% level of significance.
Solution: We are given that
n  65, S2  9.0, 20  10.5
Here, we want to test that the two years old screw length population variance
(σ2) is still maintained at 10.5. So our claim is σ2 = 10.5 and its complement is
σ2 ≠ 10.5. Since the claim contains the equality sign so we can take the claim as
the null hypothesis and complement as the alternative hypothesis. Thus,
H 0 : 2  02  10.5 and H1 : 2  10.5
Since the alternative hypothesis is two-tailed so the test is two-tailed test.
Here, the distribution of population under study is not known and sample size
is large (n > 30) so we can go for Z-test.
For testing the null hypothesis, the test statistic Z is given by
S2  σ20
Z ~ N  0,1 
2 2
σ0
n
50
 9.0  10.5  1.5 1.5 Large Sample Tests
    0.81
2 10.5  0.175 1.84
10.5
65
The critical values for two-tailed test at 5% level of significance are
± zα/2 = ± z0.025 = ±1.96.
Since calculated value of Z (= −0.81) is less than critical value (= 1.96) and
greater than the critical value (= −1.96), that means it lies in non-rejection
region, so we do not reject the null hypothesis i.e. we support the claim.
Thus, we conclude that sample fails to provide us sufficient evidence against
the claim so we may assume that two years old screw length population
variance is still maintained at 10.5mm.
Now, you can try the following exercise.
E12) A random sample of size 120 bulbs is taken from a lot which gives the
standard deviation of the life of electric bulbs 7 hours. Test the standard
deviation of the life of bulbs of the lot is 6 hours at 5% level of
significance.

10.8 TESTING OF HYPOTHESIS FOR TWO

POPULATION VARIANCES USING Z-TEST
In previous section, we have discussed testing of hypothesis about the
population variance. But there are so many situations where we want to test the
hypothesis about equality of two population variances or standard deviations.
For example, an economist may want to test whether the variability in incomes
differ in two populations, a quality controller may want to test whether the
quality of the product is changing over time, etc.
Let there be two populations, say, population-I and population-II under study.
Also let 1, 2 and 12 , 22 denote the means and variances of population-I and
population-II respectively where both 12 and 22 are unknown but 1 and  2 may
be known or unknown. For testing the hypothesis about equality of two
population variances or standard deviations, we draw a random sample of large
size n1 from population-I and a random sample of large size n2 from
population-II. Let S12 and S22 be the sample variances of the samples selected
from population-I and population-II respectively.
These two populations may or may not be normal but according to the central
limit theorem, the sampling distribution of difference of two large sample
variances asymptotically normally distributed with mean (σ12  σ 22 ) and
variance (2σ14 / n1  2σ24 / n2 ) .
Thus,
E S12  S22   E S12   E S22   12  22 … (11)

and

214 224
Var S12  S22   Var S12   Var S22   
n1 n2

51
Testing of Hypothesis But we know that standard error = Variance

2 14 2  42
 SE S12  S22   Var  S12  S22    … (12)
n1 n2

Now, follow the same procedure as we have discussed in Section 10.2, that is,
first of all we have to setup null and alternative hypothesis. Here, we want to
test the hypothesis about the two population variances, so we can take our null
and alternative hypotheses as
H 0 : 12  22  2 and H1 : 12  22 for two-tailed test 
H 0 : 12  22 and H1 : 12  22 
or  for one-tailed test 
H 0 : 12  22 and H1 : 12  22 

For testing the null hypothesis, the test statistic Z is given by

 S12  S22   E  S12  S22  ~ N 0,1
Z  
SE  S12  S22 

or
 S12  S22    σ12  σ 22 
Z  using equations (11) and (12) 
2σ14 2σ 42

n1 n2

Since under null hypothesis we assume that σ12  σ22  σ2 , therefore, we have

S12  S22
Z ~ N  0,1 
 1 1 
σ2 2  
 n1 n 2 
Generally, population variances σ12 and σ 22 are unknown, so we estimate them
by their corresponding sample variances S12 and S 22 as

ˆ 12  S12 and ˆ 22  S 22
Thus, the test statistic Z is given by
S12  S22
Z ~ N  0,1
 2S14 2S42 
 n  n 
 1 2 

After that, we calculate the value of test statistic as may be the case and
compare it with the critical value given in Table 10.1 at prefixed level of
significance α. Take the decision about the null hypothesis as described in
Section 10.2.
Note 3: When populations under study are normal then for testing the
hypothesis about equality of population variances we use F- test which will be
discussed in Unit 12 of this course. Whereas when the form of the populations
under study is not known and sample sizes are large then we apply Z-test as
discussed above.
Now, it is time to do an example based on above test.

52
Example 10: A comparative study of variation in weights (in pound) of Army- Large Sample Tests
soldiers and Navy- sailors was made. The sample variance of the weight of 120
soldiers was 60 pound2 and the sample variance of the weight of 160 sailors
was 70 pound2. Test whether the soldiers and sailors have equal variation in
their weights. Use 5% level of significance.

Solution: Given that

n1  120, S12  60, n 2  160 , S22  70

We want to test that the Army-soldiers and Navy-sailors have equal variation
in their weights. If 12 and 22 denote the variances in the weight of Army-
soldiers and Navy-sailors so our claim is 12  22 and its complement is
12  22 . Since the claim contains the equality sign so we can take the claim as
the null hypothesis and complement as the alternative hypothesis. Thus,

H 0 : 12  22  Army-soldiers and Navy-sailors 

 have equal variation in their weights 

and the alternative hypothesis as

H1 : 12   22  Army-soldiers and Navy-sailors 

 have different variation in their weights 
Here, the distributions of populations under study are not known and sample
sizes are large  n1  120  30, n 2  160  30  so we can go for Z-test.

Since population variances are unknown so for testing the null hypothesis, the
test statistic Z is given by
S12  S22
Z
 2S14 2S24 
 n  n 
 1 2 

60  70

2 2
2   60  2   70 

120 160
10 10
   0.91
60.0  61.25 11.01

The critical values for two-tailed test at 5% level of significance are

± zα/2 = ± z0.025 = ±1.96.
Since calculated value of Z (= −0.91) is less than critical value (= 1.96) and
greater than the critical value (= −1.96), that means it lies in non-rejection
region, so we do not reject the null hypothesis i.e. we support the claim.
Thus, we conclude that samples fail to provide us sufficient evidence against
the claim so we may assume that the Army-soldiers and Navy-sailors have
equal variation in their weights.
Now, you can try the following exercise.

53
Testing of Hypothesis E 13) Two sources of raw materials of bulbs are under consideration by a bulb
manufacturing company. Both sources seem to have similar
characteristics but the company is not sure about their respective
uniformity. A sample of 52 lots from source A yields variance 25 and a
sample of 40 lots from source B yields variance of 12. Test whether the
variance of source A significantly differs to the variances of source B at
 = 0.05?
We now end this unit by giving a summary of what we have covered in it.

10.9 SUMMARY
In this unit we have covered the following points:
1. How to judge a given situation whether we should go for large sample test
or not.
2. Applying the Z-test for testing the hypothesis about the population mean
and difference of two population means.
3. Applying the Z-test for testing the hypothesis about the population
proportion and difference of two population proportions.
4. Applying the Z-test for testing the hypothesis about the population variance
and two population variances.

10.10 SOLUTIONS / ANSWERS

E1) Since we have a rule that if the observed value of test statistic lies in
rejection region then we reject the null hypothesis and if calculated
value of test statistic lies in non-rejection region then we do not reject
the null hypothesis. Therefore in our case, we do not reject the null
hypothesis. So (iii) is the correct answer. Remember always on the
basis of the one sample we never accept the null hypothesis.
E2) Since the test is two-tailed therefore rejection region will lie under both
tails.
E3) Since test is two-tailed, therefore,

p-value = 2P  Z  z   2P  Z  2.42

 2 0.5  P  0  Z  2.42  2  0.5  0.4922   0.0156

E4) We are given that

n  900, X  3.4 cm, μ 0  3.25 cm and σ  2.61 cm
Here, we wish to test that the sample comes from a large population of
bolts with mean (µ) 3.25cm. So our claim is µ = 3.25cm and its
complement is µ ≠ 3.25cm. Since the claim contains the equality sign
so we can take the claim as the null hypothesis and the complement as
the alternative hypothesis. Thus,
H0 : µ  µ0  3.25 and H1 : µ  3.25
Since the alternative hypothesis is two-tailed, so the test is two-tailed
test.
54
 Here, we want to test the hypothesis regarding population mean when Large Sample Tests
population SD is unknown, so we should use t-test if the population of
bolts known to be normal. But it is not the case. Since the sample size is
large (n > 30) so we can go for Z-test instead of t-test as an
approximate. So test statistic is given by
X  0
Z
/ n
3.40  3.25 0.15
   1.72
2.61/ 900 0.087
The critical (tabulated) values for two-tailed test at 5% level of
significance are ± zα/2 = ± z0.025 = ± 1.96.
Since calculated value of test statistic Z (= 1.72) is less than the critical
value (= 1.96) and greater than critical value (= −1.96), that means it
lies in non-rejection region, so we do not reject the null hypothesis i.e.
we support the claim at 5% level of significance.
Thus, we conclude that sample does not provide us sufficient evidence
against the claim so we may assume that the sample comes from the
population of bolts with mean 3.25cm.
E5) Here, we given that
µ 0  1200, n  100, X  1220, S  90
Here, the company may accept the new CFL light when average life of
CFL light is greater than 1200 hours. So the company wants to test that
the new brand CFL light has an average life greater than 1200 hours. So
our claim is  > 1200 and its complement is  ≤ 1200. Since
complement contains the equality sign so we can take the complement
as the null hypothesis and the claim as the alternative hypothesis. Thus,
H0 : µ  µ0  1200 and H1 : µ  1200
Since the alternative hypothesis is right-tailed so the test is right-tailed
test.
Here, we want to test the hypothesis regarding population mean when
population SD is unknown, so we should use t-test if the distribution of
life of bulbs known to be normal. But it is not the case. Since the
sample size is large (n > 30) so we can go for Z-test instead of t-test.
Therefore, test statistic is given by
X  0
Z
S/ n
1220  1200 20
   2.22
90/ 100 9
The critical values for right-tailed test at 5% level of significance is
zα = z0.05 = 1.645.
Since calculated value of test statistic Z (= 2.22) is greater than critical
value (= 1.645), that means it lies in rejection region so we reject the
null hypothesis and support the alternative hypothesis i.e. we support
our claim at 5% level of significance.
55
Testing of Hypothesis Thus, we conclude that sample does not provide us sufficient evidence
against the claim so we may assume that the company accepts the new
brand of bulbs.
E6) Given that
n1  200, X  1300, S1  41;

n 2  200, Y  1280, S2  46
Here, we want to test that there is significant difference in the mean
duration of their lives of two brands of electric bulbs. If 1 and 2
denote the mean lives of two brands of electric bulbs respectively then
our claim is 1 ≠ 2 and its complement is 1 = 2. Since the
complement contains the equality sign so we can take the complement
as the null hypothesis and the claim as the alternative hypothesis. Thus,
H 0 : 1   2 and H1 : 1   2
Since the alternative hypothesis is two-tailed so the test is two-tailed
test.
We want to test the null hypothesis regarding equality of two
population means. The standard deviations of both populations are
unknown so we should go for t-test if population of difference is known
to be normal. But it is not the case. Since sample sizes are large (n1, and
n2 > 30) so we go for Z-test.
So for testing the null hypothesis, the test statistic Z is given by
XY
Z
S12 S22

n1 n 2
1300  1280 20 20
    4.59
 41
2
 46
2
8.41  10.58 4.36

200 200
The critical (tabulated) values for two-tailed test at 1% level of
significance are ± zα/2 = ± z0.005 = ± 2.58.
Since calculated value of test statistic Z (= 4.59) is greater than the
critical values (= ± 2.58), that means it lies in rejection region, so we
reject the null hypothesis an support the alternative hypothesis i.e.
support the claim at 1% level of significance.
Thus, we conclude that samples do not provide us sufficient evidence
against the claim so there is significant difference in the mean duration
of their lives of two brands of electric bulbs.
E7) Given that
n1  50, X  8.3, S1  1.2;

n 2  100 , Y  8.0, S2  1.5

Here, we want to test that the first drug provides a significant longer
period of relief than the other. If 1 and 2 denote the mean relief time
due to first and second drugs respectively then our claim is 1 > 2 and
56
 its complement is 1 ≤ 2. Since complement contains the equality sign Large Sample Tests
so we can take the complement as the null hypothesis and the claim as
the alternative hypothesis. Thus,
H 0 : 1   2 and H1 : 1   2
Since the alternative hypothesis is right-tailed so the test is right-tailed
test.
We want to test the null hypothesis regarding equality of two
population means. The standard deviations of both populations are
unknown. So we should go for t-test if population of difference is
known to be normal. But it is not the case. Since sample sizes are large
(n1, and n2 > 30) so we go for Z-test.
So for testing the null hypothesis, the test statistic Z is given by
XY
Z
S12 S22

n1 n 2

8.3  8.0 0.3

 2 2

 1.2   1.5  0.0288  0.0255

50 100
0.3
  1.32
0.2265
The critical (tabulated) value for right-tailed test at 5% level of
significance is zα = z0.05 = 1.645.
Since calculated value of test statistic Z (= 1.32) is less than the critical
value (=1.645), that means it lies in non-rejection region, so we do not
reject the null hypothesis and reject the alternative hypothesis i.e. we
reject the claim at 5% level of significance.
Thus, we conclude that samples provide us sufficient evidence against
the claim so the first drug does not has longer period of relief than the
other.
E8) Here, students are classified as Science background and other. We
define our success and failure as getting student of Science background
and other respectively. We are given that
n = Total number of students is the sample = 100
X = Number of students from Science background = 54
p = Sample proportion of students from Science background
54
  0.54
100
We want to test whether 50% of the students are from Science
background in MSc. If P denotes the proportion of first year Science
background students in the University. So our claim is P = P0 = 0.5 and
its complement is P ≠ 0.5. Since the claim contains the equality sign so
we can take the claim as the null hypothesis and complement as the
alternative hypothesis. Thus,
57
Testing of Hypothesis H0 = P = P0 = 0.5 (= 50%)
H1 : P ≠ 0.5 [Science background differs to 50%]
Since the alternative hypothesis is two-tailed so the test is two-tailed
test.
Before proceeding further, first we have to check whether the condition
of normality meets or not.
 np  100  0.54  54  5
nq  100   1  0.54   100  0.46  46  5

We see that condition of normality meets, so we can go for Z-test.

For testing the null hypothesis, the test statistic Z is given by
p  P0 0.54  0.50
Z   Q 0  1  P0 
P0Q 0 0.5  0.5 /100
n
0.04
  0.80
0.05
The critical (tabulated) values for two-tailed test at 1% level of
significance are ± zα/2 = ± z0.005 = ± 2.58.
Since calculated value of test statistic Z (= 0.80) is less than the critical
value (= 2.58) and greater than critical value (= −2.58), that means it
lies in non-rejection region, so we do not reject the null hypothesis i.e.
we support the claim at 1% level of significance.
Thus, we conclude that sample fails to provide us sufficient evidence
against the claim so we may assume that 50% of the first year students
in MSc. Economics in the University are from Science background.
E9) We define our success and failure as a patient is survived and not
survived. Here, we are given that
n = 200
X = Number of survived patients who are given a particular injection
= 180
p = Sample proportion of survived patients who are given a particular
injection
X 180
   0.9
n 200
80 80
P0  80%   0.80  Q 0  1  P0   0.20
100 100
Here, we want to test that the survival rate of the patients is more than
80%. If P denotes the proportion of survival patients then our claim is P
> 0.80 and its complement is P ≤ 0.80. Since complement contains the
equality sign so we can take the complement as the null hypothesis and
the claim as the alternative hypothesis. Thus,
H 0 : P  P0  0.80 and H1 : P  0.80

58
 Since the alternative hypothesis is right-tailed so the test is right-tailed Large Sample Tests
test.
Before proceeding further, first we have to check whether the condition
of normality meets or not.
 np  200  0.9  180  5

nq  200   1  0.9   200  0.1  20  5

We see that condition of normality meets, so we can go for Z-test.
For testing the null hypothesis, the test statistic Z is given by
p  P0
Z
P0 Q 0
n
0.9  0.8 0.1
   3.53
 0.8  0.2  0.0283
200
The critical (tabulated) value for right-tailed test at 5% level of
significance is zα = z0.05 = 1.645.
Since calculated value of test statistic Z (= 3.53) is greater than the
critical value (=1.645), that means it lies in rejection region, so we
reject the null hypothesis and support the alternative hypothesis i.e.
support our claim at 5% level of significance.
Thus, we conclude that sample fails to provide us sufficient evidence
against the claim so we may assume that the survival rate is greater than
80% in the population using that injection.
E10) Let X1 and X2 stand for number of literates in districts A and B
respectively. Therefore, we are given that
X1 50
n1  100, X1  50  p1    0.50
n1 100
X 40
n 2  100, X 2  40  p 2  2   0.40
n 2 100
Here, we want to test whether the proportion of literate persons in two
districts A and B is same. If P1 and P2 denote the proportions of literate
persons in two districts A and B respectively then our claim is P1 = P2
and its complement P1 ≠ P2. Since the claim contains the equality sign
so we can take the claim as the null hypothesis and complement as the
alternative hypothesis. Thus,
H 0 : P1  P2  P and H1 : P1  P2

Since the alternative hypothesis is two-tailed so the test is two tailed

test.
The estimate of the combined proportion (P) of literates in districts A
and B is given by
n1p1  n 2 p2 X1  X 2 50  40
P̂     0.45
n1  n 2 n1  n 2 100  100
ˆ  1  Pˆ  1  0.45  0.55
Q
59
Testing of Hypothesis Before proceeding further, first we have to check whether the condition
of normality meets or not.
 n1p1  100  0.50  50  5, n1q1  100  0.50  50  5
n 2 p2  100  0.40  40  5, n 2q 2  100  0.60  60  5
We see that condition of normality meets, so we can go for Z-test.
For testing the null hypothesis, the test statistic Z is given by
p1  p 2
Z
ˆ  1  1 
P̂Q
 n1 n 2 
0.50  0.40 0.10
   1.42
1 1  0.0704
0.45  0.55   
 100 100 
The critical (tabulated) values for two-tailed test at 1% level of
significance are ± zα/2 = ± z0.005 = ± 2.58.
Since calculated value of test statistic Z (= 1.42) is less than the critical
value (= 2.58) and greater than critical value (= −2.58), that means it
lies in non-rejection region, so we do not reject the null hypothesis i.e.
we support our claim at 1% level of significance.
Thus, we conclude that samples fail to provide us sufficient evidence
against the claim so we may assume that the proportion of literates in
districts A and B is equal.
E11) Here, we are given that
n1  1200, p1  30%  0.30

n 2  900, p2  20%  0.20

Here, we want to test that the proportion of blue-eye persons in both the
population is same. If P1 and P2 denote the proportions of blue-eye
persons in two populations respectively then our claim is P1 = P2 and its
complement P1 ≠ P2. Since the claim contains the equality sign so we
can take the claim as the null hypothesis and complement as the
alternative hypothesis. Thus,
H 0 : P1  P2  P and H1 : P1  P2

Since the alternative hypothesis is two-tailed so the test is two tailed

test.
The estimate of the combined proportion (P) of literates in districts A
and B is given by

n1p1  n 2 p2 1200  0.30  900  0.20

P̂    0.257
n1  n 2 1200  900
ˆ  1  Pˆ  1  0.257  0.743
Q

Before proceeding further, first we have to check whether the condition

of normality meets or not.
60
  n1p1  1200  0.30  360  5, n1q1  1200  0.70  840  5 Large Sample Tests

n 2p2  900  0.20  180  5, n 2q 2  900  0.80  720  5

We see that condition of normality meets, so we can go for Z-test.
For testing the null hypothesis, the test statistic Z is given by
p1  p 2
Z
ˆ  1  1 
P̂Q
 n1 n 2 
0.30  0.20 0.10
   5.26
 1 1  0.019
0.257  0.743   
 1200 900 
The critical (tabulated) values for two-tailed test at 5% level of
significance are ± zα/2 = ± z0.005 = ± 1.96.
Since calculated value of test statistic Z (= 5.26) is greater than critical
values (= ±1.96), that means it lies in rejection region, so we reject the
null hypothesis i.e. we reject our claim at 5% level of significance.
Thus, we conclude that samples provide us sufficient evidence against
the claim so the proportion of blue-eyed person in both the population
is not same.
E12) Here, we are given that
n  120, S  7, 0  6
Here, we want to test that standard deviation (σ) of the life of bulbs of
the lot is 6 hours. So our claim is σ = 6 and its complement is σ ≠ 6.
Since the claim contains the equality sign so we can take the claim as
the null hypothesis and complement as the alternative hypothesis. Thus,
H 0 :   0  6 and H1 :   6
Since the alternative hypothesis is two-tailed so the test is two-tailed
test.
Here, the distribution of population under study is not known and
sample size is large (n > 30) so we can go for Z-test.
For testing the null hypothesis, the test statistic Z is given by
S2  σ02
Z ~ N  0,1 
2
σ02
n
2 2


 7  6 
13

13
 2.80
2 2 36  0.129 4.64
6
120
The critical values for two-tailed test at 5% level of significance are
± zα/2 = ± z0.025 = ±1.96.
Since calculated value of Z (= 2.8) is greater than critical values
(= ±1.96), that means it lies in rejection region, so we reject the null
hypothesis i.e. we reject our claim at 5% level of significance.

61
Testing of Hypothesis Thus, we conclude that sample provides us sufficient evidence against
the claim so standard deviation of the life of bulbs of the lot is not 6.0
hours.
E13) Here, we are given that
n1  52, S12  25
n2  40, S22  12
Here, we want to test that variance of source A significantly differs to
the variances of source B. If 12 and 22 denote the variances in the raw
materials of sources A and B respectively so our claim is 12  22 and
its complement is 12   22 . Since complement contains the equality
sign so we can take the complement as the null hypothesis and the
claim as the alternative hypothesis. Thus,
H0 : 12  22 and H1 : 12  22
Since the alternative hypothesis is two-tailed so the test is two-tailed
test.
Here, the distributions of populations under study are not known and
sample sizes are large (n1  52  30, n 2  40  30) so we can go for Z-
test.
Since population variances are unknown so for testing the null
hypothesis, the test statistic Z is given by
S12  S22
Z
 2S14 2S24 
 n  n 
 1 2 

25  12 13
   2.36
2
225 212
2 5.5

52 40
The critical values for two-tailed test at 5% level of significance are
± zα/2 = ± z0.025 = ±1.96.
Since calculated value of Z (= 2.36) is greater than critical values (= ±1.96),
that means it lies in rejection region, so we reject the null hypothesis and
support the alternative hypothesis i.e. we support our claim at 5% level of
significance.
Thus, we conclude that samples fail to provide us sufficient evidence against
the claim so variance of source A significantly differs to the variance of source
B.

62
UNIT 11 SMALL SAMPLE TESTS
Structure
11.1 Introduction
Objectives
11.2 General Procedure of t-Test for Testing a Hypothesis
11.3 Testing of hypothesis for Population Mean Using t-Test
11.4 Testing of Hypothesis for Difference of Two Population Means Using
t-Test
11.5 Paired t-Test
11.6 Testing of Hypothesis for Population Correlation Coefficient Using
t-Test
11.7 Summary
11.8 Solutions /Answers

11.1 INTRODUCTION
In previous unit, we have discussed the testing of hypothesis for large samples
in details. Recall that throughout the unit, we were making an assumption that
“if sample size is sufficiently large then test statistic follows approximately
standard normal distribution”. Also recall two points highlighted in this course,
i.e.
 Cost of our study increases as sample size increases.
 Sometime nature of the units in the population under study is such that they
destroyed under investigation.
If there are limited recourses in terms of money then first point listed above
force us not to go for large sample size when items /units under study are very
costly such as airplane, computer, etc. Second point listed above give an alarm
for not to go for large sample if population units are destroyed under
investigation.
So, we need an alternative technique which is used to test the hypothesis based
on small sample(s). Small sample tests do this job for us. But in return they
demand one basic assumption that population under study should be normal as
you will see when you go through the unit. t,  2 and F -tests are some
commonly used small sample tests.
In this unit, we will discuss t-test in details which is based on the t-distribution
described in Unit 3 of this course. And  2 and F-tests will be discussed in next
unit which are based on χ2 and F-distributions described in Unit 3 and Unit 4 of
this course respectively.
This unit is divided into eight sections. Section 11.1 is described the need of
small sample tests. The general procedure of t-test for testing a hypothesis is
described in Section 11.2. In Section 11.3, we discuss testing of hypothesis for
population mean using t-test. Testing of hypothesis for difference of two
population means when samples are independent is described in Section 11.4
whereas in Section 11.5, the paired t-test for difference of two population
means when samples are dependent(paired) is discussed. In Section 11.6
testing of hypothesis for population correlation coefficient is explained. Unit

63
Testing of Hypothesis ends by providing summary of what we have discussed in this unit in Section
11.7 and solution of exercises in Section 11.8.
Before moving further a humble suggestion to you that please revise what you
have learned in previous two units. The concepts discussed there will help you
a lot to better understand the concepts discussed in this unit.
Objectives
After studying this unit, you should be able to:
 realize the importance of small sample tests;
 know the procedure of t-test for testing a hypothesis;
 describe testing of hypothesis for population mean for using t-test;
 explain the testing of hypothesis for difference of two population means
when samples are independent using t-test;
 describe the procedure for paired t-test for testing of hypothesis for
difference of two population means when samples are dependent or paired;
and
 explain the testing of hypothesis for population correlation coefficient using
t-test.

11.2 GENERAL PROCEDURE OF t-TEST FOR

TESTING A HYPOTHESIS
The general procedure of t-test for testing a hypothesis is similar as Z-test
already explained in Unit 10. Let us give you similar details here.
For this purpose, let X1, X2,…, Xn be a random sample of small size n (< 30)
selected from a normal population (recall the demand of small sample tests
pointed out in previous Section 11.1) having parameter of interest, say, 
which is actually unknown but its hypothetical value, say, 0 estimated from
some previous study or some other way is to be tested. t-test involves following
steps for testing this hypothetical value:
Step I: First of all, we setup null and alternative hypotheses. Here, we want
to test the hypothetical value θ0 of parameter θ so we can take the null
and alternative hypotheses as
H 0 :   0 and H 1 :   0 for two-tailed test 
H 0 :   0 and H 1 :   0 
or  for one-tailed test 
H 0 :   0 and H 1 :   0 

In case of comparing same parameter of two populations of interest,

say, 1, and 2 then our null and alternative hypotheses would be

H 0 : 1   2 and H1 : 1  2 for two-tailed test

H 0 : 1  2 and H1 : 1  2 
or  for one-tailed test 
H 0 : 1  2 and H1 : 1  2 

Step II: After setting the null and alternative hypotheses our next step is to
decide a criteria for rejection or non-rejection of null hypothesis i.e.
64
 decide the level of significance , at which we want to test our null Small Sample Tests
hypothesis. We generally take  = 5 % or 1%.
Step III: The third step is to determine an appropriate test statistic, say, t for
testing the null hypothesis. Suppose Tn is the sample statistic (may be
sample mean, sample correlation coefficient, etc. depending upon )
for the parameter  then test-statistic t is given by
Tn  E(Tn )
t
SE  Tn 

Step IV: As we know, t-test is based on t-distribution and t-distribution is

described with the help of its degrees of freedom, therefore, test
statistic t follows t-distribution with specified degrees of freedom as
the case may be.
By putting the values of Tn, E(Tn) and SE(T n) in above formula, we
calculate the value of test statistic t. Let tcal be the calculated value of
test statistic t after putting these values.
Step V: After that, we obtain the critical (cut-off or tabulated) value(s) in the
sampling distribution of the test statistic t corresponding to  assumed
in Step II. The critical values for t-test are given in Table-II (t-table)
of the Appendix at the end of Block 1 of this course corresponding to
different level of significance (α). After that, we construct rejection
(critical) region of size  in the probability curve of the sampling
distribution of test statistic t.
Step VI: Take the decision about the null hypothesis based on calculated and
critical value(s) of test statistic obtained in Step IV and Step V
respectively. Since critical value depends upon the nature of the test
that it is one-tailed test or two-tailed test so following cases arise:
In case of one-tailed test:
Case I: When H 0 :   0 and H 1 :   0 (right-tailed test)

In this case, the rejection (critical) region falls under the right tail of
the probability curve of the sampling distribution of test statistic t.
Suppose t(ν), is the critical value at  level of significance then entire
region greater than or equal to t(ν), is the rejection region and less
than t(ν), is the non-rejection region as shown in Fig. 11.1.
If tcal ≥ t(ν,), that means calculated value of test statistic t lies in the Fig. 11.1
rejection (critical) region, then we reject the null hypothesis H0 at 
level of significance. Therefore, we conclude that sample data
provides us sufficient evidence against the null hypothesis and there is
a significant difference between hypothesized value and observed
value of the parameter.
If tcal < t(ν,), that means calculated value of test statistic t lies in non-
rejection region, then we do not reject the null hypothesis H0 at 
level of significance. Therefore, we conclude that the sample data
fails to provide us sufficient evidence against the null hypothesis and
the difference between hypothesized value and observed value of the
parameter due to fluctuation of sample.
Testing of Hypothesis Case II: When H 0 :    0 and H 1 :   0 (left-tailed test)
In this case, the rejection (critical) region falls under the left tail of the
probability curve of the sampling distribution of test statistic t.
Suppose - t(ν), is the critical value at  level of significance then
entire region less than or equal to - t(ν), is the rejection region and
greater than - t(ν), is the non-rejection region as shown in Fig. 11.2.
If tcal ≤ − t(ν),, that means calculated value of test statistic t lies in the
rejection (critical) region, then we reject the null hypothesis H0 at 
level of significance.
If tcal >- t(ν),, that means calculated value of test statistic t lies in the
Fig. 11.2 non-rejection region, then we do not reject the null hypothesis H0 at 
level of significance.
In case of two-tailed test:
That is, when H 0 :   0 and H 1 :   0
In this case, the rejection region falls under both tails of the
probability curve of sampling distribution of the test statistic t. Half
the area (α) i.e. α/2 will lies under left tail and other half under the
right tail. Suppose  t (  ),  / 2 and t (  ),  / 2 are the two critical values at
the left- tailed and right-tailed respectively. Therefore, entire region
less than or equal to  t (  ),  / 2 and greater than or equal to t (  ),  / 2 are
Fig. 11.3
the rejection regions and between  t (  ),  / 2 and t (  ),  / 2 is the non-
rejection region as shown in Fig. 11.3.
If tcal ≥ t(ν),/2, or tcal ≤ -t(ν),/2, that means calculated value of test
statistic t lies in the rejection(critical) region, then we reject the null
hypothesis H0 at  level of significance.
And if -t(ν),/2 < tcal < t(ν),/2, that means calculated value of test
statistic t lies in the non-rejection region, then we do not reject the
null hypothesis H0 at  level of significance.
Procedure of taking the decision about the null hypothesis on the basis of
p-value:
To take the decision about the null hypothesis on the basis of p-value, the p-
value is compared with given level of significance (α). And if p-value is less
than or equal to α then we reject the null hypothesis and if p-value is greater
than α then we do not reject the null hypothesis at α level of significance.
Since the distribution of test statistic t follows t-distribution with ν df and we
also know that t-distribution is symmetrical about t = 0 line therefore, if tcal
represents calculated value of test statistic t then p-value can be defined as:
For one-tailed test:
For H1: θ > θ0 (right-tailed test)
p-value = P[t  tcal]
For H1: θ < θ0 (left-tailed test)
p-value = P  t  t cal 
For two-tailed test: For H1:   0
 p-value = 2P  t  t cal  Small Sample Tests

These p-values for t-test can be obtained with the help of Table-II (t-table)
given in the Appendix at the end of Block 1 of this course. But this table gives
the t-values corresponding to the standard values of α such as 0.10, 0.05,
0.025, 0.01 and 0.005 only, therefore, the exact p-values are not obtained with
the help of this table and we can approximate the p-value for this test.
For example, if test is right-tailed and calculated (observed) value of test
statistic t is 2.94 with 9 df then p-value is obtained as:
Since calculated value of test statistic t is based on the 9 df therefore, we use
row for 9 df in the t-table and move across this row to find the values in which
calculated t-value falls. Since calculated t-value falls between 2.821 and 3.250,
which are corresponding to the values of one-tailed area α = 0.01 and 0.005
respectively, therefore, p-value will lie between 0.005 and 0.01, that is,
0.005  p-value  0.01
If in the above example, the test is two-tailed then the two values 0.01 and
0.005 would be doubled for p-value, that is,
0.005  2  0.01  p-value  0.02  2  0.01
Note 1: With the help of computer packages and softwares such as SPSS, SAS,
MINITAB, EXCEL, etc. we can find the exact p-values for t-test.
Now, you can try the following exercise.

E1) If test is two-tailed and calculated value of test statistic t is 2.42 with 15
df then find the p-value for t-test.

11.3 TESTING OF HYPOTHESIS FOR

POPULATION MEAN USING t-TEST
In Section 10.3 of the previous unit, we have discussed Z-test for testing the
hypothesis about population mean when population variance σ2 is known and
unknown.
Recall form these, we have already pointed out that one basic difference
between Z-test and t-test is that Z-test is used when population SD is known
whether sample size is large or small and t-test is used when population SD is
unknown whether sample size is small or large. But in case of large sample size
Z-test is an appropriate of t-test as we did in previous unit. But in practice
standard deviation of population is not known and sample size is small so in
this situation, we use t-test provided population under study is normal.
Assumptions
Virtually every test has some assumptions which must be met prior to the
application of the test. This t-test needs following assumptions to work:
(i) The characteristic under study follows normal distribution. In other
words, populations from which random sample is drawn should be
normal with respect to the characteristic of interest.
(ii) Sample observations are random and independent.
(iii) Population variance σ2 is unknown.
67
Testing of Hypothesis For describing this test, let X1, X2,…, Xn be a random sample of small size
n (< 30) selected from a normal population with mean µ and unknown
variance σ2.
Now, follow the same procedure as we have discussed in previous section,
that is, first of all we setup the null and alternative hypotheses. Here, we
want to test the claim about the specified value 0 of population mean µ so
we can take the null and alternative hypotheses as
 Here,    and 0  0 
if we compareit with 
H0 :   0 and H1 :   0  for two-tailed test  general procedure. 

H 0 :   0 and H1 :   0 
or   for one-tailed test 
H 0 :   0 and H1 :   0 
For testing the null hypothesis, the test statistic t is given by
X  0
t ~ t  n 1 under H 0
S/ n

1 n 1 n 2
where, X  
n i1
X i is the sample mean and S 2
 
n  1 i1
 X i  X  is the

sample variance.
For computational simplicity, we may use the following formulae for X, S2 :
2
1 1    d 
n  1 
X  a   d and S 
2 2
 d  
n n 
 
where, d = (X − a ), ‘a’ being the assumed arbitrary value.
Here, the test statistic t follows t-distribution with (n − 1) degrees of freedom as
we discussed in Unit 3 of this course.
After substituting values of X, S and n , we get calculated value of test
statistic t. Then we look for critical (or cut-off or tabulated) value(s) of test
statistic t from the t-table. On comparing calculated value and critical value(s),
we take the decision about the null hypothesis as discussed in previous section.
Let us do some examples of testing of hypothesis about population mean using
t-test.
Example 1: A manufacturer claims that a special type of projector bulb has an
average life 160 hours. To check this claim an investigator takes a sample of 20
such bulbs, puts on the test, and obtains an average life 167 hours with standard
deviation 16 hours. Assuming that the life time of such bulbs follows normal
distribution, does the investigator accept the manufacturer’s claim at 5% level
of significance?
Solution: Here, we are given that
0  160, n  20, X  167 and S  16
Here, we want to test the manufacturer claims that a special type of projector
bulb has an average life (µ) 160 hours. So claim is µ = 160 and its complement
is µ ≠ 160. Since the claim contains the equality sign so we can take the claim
as the null hypothesis and complement as the alternative hypothesis. Thus,
68
 H 0 :    0  160 and H 1 :   160 Small Sample Tests

Since the alternative hypothesis is two-tailed so the test is two-tailed test.

Here, we want to test the hypothesis regarding population mean when
population SD is unknown. Also sample size is small n = 20(n < 30) and
population under study is normal, so we can go for t-test for testing the
hypothesis about population mean.
For testing the null hypothesis, the test statistic t is given by
X  µ0
t
S/ n
167  160 7
   1.96
16 / 20 3.58
The critical value of the test statistic t for various df and different level of
significance  are given in Table II of the Appendix at the end of the Block
1of this course.
The critical (tabulated) values of test statistic for two-tailed test corresponding
(n-1) = 19 df at 5% level of significance are  t ( n 1), α / 2   t (19 ), 0.025   2.093.

Since calculated value of test statistic t (= 1.96) is greater than the critical
value (= − 2.093) and is less than critical value (= 2.093), that means calculated
value of test statistic lies in non-rejection region as shown in Fig. 11.4. So we
do not reject the null hypothesis i.e. we support the manufacture’s claim at 5%
level of significance.
Fig. 11.4
Decision according to p-value:
Since calculated value of test statistic t is based on 19 df therefore, we use row
for 19 df in the t-table and move across this row to find the values in which
calculated t-value falls. Since calculated t-value falls between 1.729 and 2.093
corresponding to one-tailed area α = 0.05 and 0.025 respectively therefore p-
value lies between 0.025 and 0.05, that is,
0.025  p-value  0.05
Since test is two-tailed so
2  0.025  0.05  p-value  0.10  0.05  2
Since p-value is greater than  (  0.05) so we do not reject the null
hypothesis at 5% level of significance.
Thus, we conclude that sample fails to provide us sufficient evidence against
the null hypothesis so we may assume that the manufacture’s claim is true so
the investigator may accept the manufacturer’s claim at 5% level of
significance.
Example 2: The mean share price of companies of Pharma sector is Rs.70. The
share prices of all companies were changed time to time. After a month, a
sample of 10 Pharma companies was taken and their share prices were noted as
below:
70, 76, 75, 69, 70, 72, 68, 65, 75, 72
Assuming that the distribution of share prices follows normal distribution, test
whether mean share price is still the same at 1% level of significance?
Testing of Hypothesis Solution: Here, we wish to test that the mean share price (µ) of companies of
Pharma sector is still Rs.70 besides all changes. So our claim is µ = 70 and its
complement is µ ≠ 70. Since the claim contains the equality sign so we can
take the claim as the null hypothesis and complement as the alternative
hypothesis. Thus,
H 0 :    0  70 mean share price of companies is still Rs. 70 
H1 :    0  70  mean share price of companies is not still Rs. 70
Since the alternative hypothesis is two-tailed so the test is two-tailed test.
Here, we want to test the hypothesis regarding population mean when
population SD is unknown. Also sample size is small n = 10(n < 30) and
population under study is normal, so we can go for t-test for testing the
hypothesis about population mean.
For testing the null hypothesis, the test statistic t is given by
X  µ0
t ~ t ( n 1) … (1)
S/ n
Calculation for X and S:
S. No. Sample value Deviation d2
(X) d = (X-a), a = 70
1 70 0 0
2 76 6 36
3 75 5 25
4 69 -1 1
5 70 0 0
6 72 2 4
7 68 -2 4
8 65 -5 25
9 75 5 25
10 72 2 4
Total 12 124

The assumed value of a is 70.

From the above calculation, we have
1 1
Xa
n  d  70   12  71.2
10

2
S 
1  2
d 
 d 
2


n 1  n 
 

1   12 2  1  144 
  124    124   12.18
10  1  10  9  10 

S 12.18  3.49

Putting the values in equation (1), we have
71.2  70 1.2
t   1.09
3.49 / 10 1.10
70
The critical (tabulated) values of test statistic for two-tailed test corresponding Small Sample Tests
(n-1) = 9 df at 1% level of significance are  t ( n 1), α / 2   t ( 9 ), 0.005   3.250.

Since calculated value of test statistic t (= 1.09) is less than the critical value
(= 3.250) and greater than the critical value (= ‒3.250), that means calculated
value of t lies in non-rejection region as shown in Fig. 11.5. So we do not
reject the null hypothesis i.e. we support the claim at 1% level of significance.
Decision according to p-value:
Since calculated value of test statistic t is based on 9 df therefore, we use row
for 9 df in the t-table and move across this row to find the values in which Fig. 11.5
calculated t-value falls. Since all values in this row are greater than calculated
t-value 1.09 and the smallest value is 1.383 corresponding to one-tailed area
α = 0.10 therefore p-value is greater than 0.10, that is,
p-value  0.10
Since test is two-tailed so
p-value  2  0.10  0.20
Since p-value (= 0.20) is greater than (  0.01) so we do not reject the null
hypothesis at 1% level of significance.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so may assume that the mean share price is still Rs. 70.
Now, you can try the following exercises.
E2) A tyre manufacturer claims that the average life of a particular category
of his tyre is 18000 km when used under normal driving conditions. A
random sample of 16 tyres was tested. The mean and SD of life of the
tyres in the sample were 20000 km and 6000 km respectively.
Assuming that the life of the tyres is normally distributed, test the claim
of the manufacturer at 1% level of significance using appropriate test.
E3) It is known that the average weight of cadets of a centre follows normal
distribution. Weights of 10 randomly selected cadets from the same
centre are as given below:
48, 50, 62, 75, 80, 60, 70, 56, 52, 77
Can we say that average weight of all cadets of the centre from which
the above sample was taken is equal to 60 kg at 5% level of
significance?

11.4 TESTING OF HYPOTHESIS FOR

DIFFERENCE OF TWO POPULATION MEANS
USING t-TEST
In Section 10.4 of the previous unit, we have discussed Z-test for testing the
hypothesis about difference of two population means under different possibility
of population variances 12 and 22 . Recall from there, we have pointed out that
one basic difference between Z-test and t-test is that, Z-test is used when
standard deviations of both populations are known and t-test is used when
standard deviations of both populations are unknown. But in practice standard
Testing of Hypothesis deviations of both populations are not known, so in real life problems t-test is
more suitable compared to Z-test.
Assumptions
This test works under following assumptions:
(i) The characteristic under study follows normal distribution in both the
populations. In other words, both populations from which random
samples are drawn should be normal with respect to the characteristic of
interest.
(ii) Samples and their observations both are independent to each other.
(iii) Population variances 12 and  22 are both unknown but equal.
For describing this test, let there be two normal populations N 1, 12 and  
N 2 , 22  under study. And we have to draw two independent random samples,
say, X1,X2 ,...,Xn1 and Y1, Y2 ,..., Yn2 of sizes n1 and n2 from these normal
   
populations N 1, 12 and N 2 , 22 respectively. Let X and Y be the means of
first and second sample respectively. Further, suppose the variances of both the
populations are unknown but are equal, i.e., 12  22  2 (say). In this case, σ2
is estimated by value of pooled sample variance S 2p where,
1
S2p   n1  1 S12   n 2  1 S22 
n1  n 2  2

and
n n
1 1
2 1 2
2
S12  
(n1  1) i 1
 X i  X  , S 2
2  
(n 2  1) i1
 Yi  Y 

This can also be written as

1  n1 2
n2
2
  Xi  X     Yi  Y  
2
S 
p
n1  n 2  2  i 1 i 1 
For computational simplicity, use the following formulae for X, Y and S 2p :

1 1
Xa
n1  d1 , Y  b 
n2 d 2 and
2 2
1     d1      d 2   
n1  n 2  2    1     d2 
2 2 2
S 
p  d  
n1 n2
    
where, d1= (X − a ) and d2 = (Y − b), ‘a’ and ‘b’ are the assumed arbitrary
values.
Now, follow the same procedure as we have discussed in Section 11.2, that is,
first of all we have to setup null and alternative hypotheses. Here, we want to
test the hypothesis about the difference of two population means so we can take
the null hypothesis as
 Here, 1  1 and 2   2 
H 0 : 1  2 (no difference in means)  if we compare it with 
 general procedure. 

or H 0 : 1  2  0 (difference in two means is 0)

72
and the alternative hypothesis as Small Sample Tests

H1 : 1  2 for two-tailed test 

H0 : 1  2 and H1 : 1  2 
or  for one-tailed test 
H0 : 1  2 and H1 : 1  2 
For testing the null hypothesis, the test statistic t is given by
XY
t ~ t (n 1 n 2 2) under H0
1 1
Sp 
n1 n 2

After substituting values of X, Y, S p , n1 and n 2 , we get calculated value of test

statistic t. Then we look for critical (or cut-off or tabulated) value(s) of test
statistic t from the t-table. On comparing calculated value and critical value(s),
we take the decision about the null hypothesis as discussed in Section 11.2.
Let us do some examples to become more user friendly with the test explained
above.
Example 3: In a random sample of 10 pigs fed by diet A, the gain in weights
(in pounds) in a certain period were
12, 8, 14, 16, 13, 12, 8, 14, 10, 9
In another random sample of 10 pigs fed by diet B, the gain in weights (in
pounds) in the same period were
14, 13, 12, 15, 16, 14, 18, 17, 21, 15
Assuming that gain in the weights due to both foods follows normal
distributions with equal variances, test whether diets A and B differ
significantly regarding their effect on increase in weight at 5% level of
significance.
Solution: Here, we can test that diets A and B differ significantly regarding
their effect on increase in weight of pigs. If 1 and 2 denote the average gain
in weights due to diet A and diet B respectively then our claim is 1 ≠ 2 and
its complement is 1 = 2. Since complement contains the equality sign so we
can take the complement as the null hypothesis and the claim as the alternative
hypothesis. Thus,
H 0 : 1   2 and H 1 : 1   2

Since the alternative hypothesis is two-tailed so the test is two-tailed test.

Since it is given that the increase in the weight due to both foods follows
normal distributions and population variances are equal and unknown. And
other assumptions of t-test for testing a hypothesis about difference of two
population means also meet. So we can go for this test.
For testing the null hypothesis, the test statistic t is given by
XY
t ~ t (n 1n 22) under H0 … (2)
1 1
Sp 
n1 n 2

73
Testing of Hypothesis Calculation for X, Y and S p :
Diet A Diet B
X d1 = (X−a) d1 2 Y d2 = (Y−b) d2 2
a =12 b = 16
12 0 0 14 −2 4
8 −4 16 13 −3 9
14 2 4 12 −4 16
16 4 16 15 −1 1
13 1 1 16 0 0
12 0 0 14 −2 4
8 −4 16 18 2 4
14 2 4 17 1 1
10 −2 4 21 5 25
9 −3 9 15 −1 1
Total −4 70 −5 65

Here, a =12, b = 16 are assumed values.

From above calculations, we have
1  4 
Xa
n1  d1  12 
10
 11.6,

1  5 
Y  b   d 2  16   15.5
n2 10
2 2
1     d1      d2   
n1  n 2  2    1     d2 
2 2 2
S 
p  d  
n1 n2
    


1 
70 
 42   65   52 
   
10  10  2  10   10 

1
  68.4  62.5   7.27
18
 Sp  7.27  2.70

Putting the values in equation (2), we have

11.6  15.5
t
1 1
2.70 
10 10
 3.90 3.90
   3.21
2 .70  0 .45 1.215
The critical values of test statistic t for two-tailed test corresponding
(n1 + n2 -2) = 18 df at 5% level of significance are
 t ( n 1  n 2  2 ), α / 2   t (18 ), 0.025   2.101.

Since calculated value of test statistic t (= −3.21) is less than critical values
(± 2.101) that means calculated value of test statistic t lies in rejection region,
so we reject the null hypothesis and support the alternative hypothesis i.e.
support our claim at 5% level of significance.
74
Thus, we conclude that samples do not provide us sufficient evidence against Small Sample Tests
the claim so diets A and B differ significantly in terms of gain in weights of
pigs.
Example 4: The means of two random samples of sizes 10 and 8 drawn from
two normal populations are 210.40 and 208.92 respectively. The sum of
squares of the deviations from their means is 26.94 and 24.50 respectively.
Assuming that the populations are normal with equal variances, can samples be
considered to have been drawn from normal populations having equal mean.
Solution: In usual notations, we are given that
n1  10, n 2  8, X  210.40, Y  208.92,
2 2
 X  X   26.94,  Y  Y   24.50

Therefore,
1
S 2p     X  X  2    Y  Y 2 
n1  n 2  2  

1
 26.94  24.50  1  51.44  3.215
10  8  2 16
 Sp  3.215  1.79
We wish to test that both the samples are drawn from normal populations
having the same means. If 1 and 2 denote the means of both normal
populations respectively then our claim is 1= 2 and its complement is
1 ≠ 2. Since the claim contains the equality sign so we can take the claim as
the null hypothesis and complement as the alternative hypothesis. Thus,
H 0 : 1   2 [mean of both populations is equal]
H1 : 1   2 [mean of both populations is not equal]
Since the alternative hypothesis is two-tailed so the test is two-tailed test.
Since it is given that two populations are normal with equal and unknown
variances and other assumptions of t-test for testing a hypothesis about
difference of two population means also meet. So we can go for this test.
For testing the null hypothesis, the test statistic t is given by
XY
t ~ t (n 1 n 2 2) under H0
1 1
Sp 
n1 n 2
210.40  208.92 1.48 1.48
    1.76
1 1 1.79  0.47 0.84
1.79 
10 8
The critical values of test statistic t for two-tailed test corresponding
(n1 + n2 -2) = 16 df at 5% level of significance are
 t ( n1  n 2  2 ), α / 2   t (16 ), 0.025   2.12.

Since calculated value of test statistic t (= 1.76) is less than the critical value
(= 2.12) and greater than the critical value (= ‒2.12), that means calculated
value of test statistic t lies in non-rejection region so we do not reject the null
hypothesis i.e. we support the claim.
75
Testing of Hypothesis Thus, we conclude that samples fail to provide us sufficient evidence against
the claim so we may assume that both samples are taken from normal
populations having equal means.
Now, you can try the following exercises.
E4) Two different types of drugs A and B were tried on some patients for
increasing their weights. Six persons were given drug A and other 7
persons were given drug B. The gain in weights (in ponds) is given
below:

Drug A 5 8 7 10 9 6 −

Drug B 9 10 15 12 14 8 12

Assuming that increment in the weights due to both drugs follows

normal distributions with equal variances, do the both drugs differ
significantly with regard to their mean weights increment at 5% level of
significance?
E5) To test the effect of fertilizer on wheat production, 26 plots of land with
equal areas were chosen. Half of these plots were treated with fertilizer
and the other half were untreated. Other conditions were the same. The
mean yield of wheat on the untreated plots was 4.6 quintals with a
standard deviation of 0.5 quintals, while the mean yield of the treated
plots was 5.0 quintals with standard deviations of 0.3 quintals.
Assuming that yields of wheat with and without fertilizer follow normal
distributions with equal variances, can we conclude that there is
significant improvement in wheat production due to effect of fertilizer
at 1% level of significance?

11.5 PAIRED t-TEST

In the previous section, we have discussed t-test for equality of two population
means in case of independent samples. However, there are so many situations
where two samples are not independent and observations are recorded on the
same individuals or items. Generally, such types of observations are recorded
to assess the effectiveness of a particular training, diet, treatment, medicine,
etc. In such situations, the observations are recorded “before and after” the
insertion of training, treatment, etc. as the case may be. For example, if we
wish to test a new diet on, say, 15 individuals then the weight of the individuals
recorded before diet and after the diet will form two different samples in which
observations will be paired as per each individual. Similarly, in the test of
blood-sugar in human body, fasting sugar level before meal and sugar level
after meal, both are recorded for a patient as paired observations, etc. The
parametric test designed for this type of situation is known as paired t-test.
Now, come to the working principle of this test. This test first of all converts
the two populations into a single population by taking the difference of paired
observations.
Now, instead of two populations, we are left with one population, the
population of differences. And the problem of testing equality of two
population mean reduces to test the hypothesis that mean of the population of
differences is equal to zero.

76
Assumptions Small Sample Tests

This test works under following assumptions:

(i) The population of differences follows normal distribution.
(ii) Samples are not independent.
(iii) Size of both the samples is equal.
(iv) Population variances are unknown but not necessarily equal.
Let (X1, Y1), (X2, Y2), …,(Xn, Yn) be a paired random sample of size n and the
difference between paired observations Xi & Yi be denoted by Di, that is,
D i  X i  Yi for all i  1, 2,..., n
Hence, we can assume that D1, D2, …, Dn be a random sample from normal
population of differences with mean D and unknown variance  2D . This is
same as the case of testing of hypothesis for population mean when population
variance is unknown which is described in Section 11.3 of this unit.
Here, we want to test that there is an effect of a diet, training, treatment,
medicine, etc. So we can take the null hypothesis as
H0: µ1 = µ2 or H0 : µD  µ1  µ2  0
and the alternative hypothesis
H1 : 1   2 or H1 :  D  0 for two-tailed test 
H0: 1  2 and H1 : 1  2 
or  for one-tailed test 
H0: 1  2 and H1 : 1  2 
For testing the null hypothesis, the test statistic t is given by
D
t ~ t  n 1  under H 0
SD / n
2
  n  
1 n 1 n 2
 D 
1  n 2  i1  
n i1
2
where, D   Di and SD    Di  D   n  1  
n  1 i1 i 1
D 
n 
 
 
After substituting values of D, S D and n we get calculated value of test
statistic t. Then we look for critical (or cut-off or tabulated) value(s) of test
statistic t from the t-table. On comparing calculated value and critical value(s),
we take the decision about the null hypothesis as discussed in Section 11.2.
Let us do some examples to become more user friendly with paired t-test.
Example 5: A group of 12 children was tested to find out how many digits
they would repeat from memory after hearing them once. They were given
practice session for this test. Next week they were retested. The results
obtained were as follows:
Child Number 1 2 3 4 5 6 7 8 9 10 11 12
Recall Before 6 4 5 7 6 4 3 7 8 4 6 5
Recall After 6 6 4 7 6 5 5 9 9 7 8 7

77
Testing of Hypothesis Assuming that the memories of the children before and after the practice
session follow normal distributions, is the memory practice session improve the
performance of children?
Solution: Here, we want to test that memory practice session improve the
performance of children. If 1 and 2 denote the mean digit repetition before
and after the practice so our claim is 1 < 2 and its complement is 1 ≥ 2.
Since complement contains the equality sign so we can take the complement as
the null hypothesis and the claim as the alternative hypothesis. Thus,
H 0 : 1   2 and H1 : 1   2

Since the alternative hypothesis is left-tailed so the test is left-tailed test.

It is a situation of before and after. Also, it is given that the memories of the
children before and after the practice session follow normal distributions. So,
population of differences will also be normal. Also all the assumptions of
paired t-test meet so we can go for paired t-test.
For testing the null hypothesis, the test statistic t is given by
D
t ~ t  n 1  under H 0 … (3)
SD / n

where, D and S D are mean and standard deviation of the population of

differences.
Calculation for D and S D :
Child Digit recall D = (X−Y) D2
Number Before (X) After (Y)
1 6 6 0 0
2 4 6 −2 4
3 5 4 1 1
4 7 7 0 0
5 6 6 0 0
6 4 5 −1 1
7 3 5 −2 4
8 7 9 −2 4
9 8 9 −1 1
10 4 7 −3 9
11 6 8 −2 4
12 5 7 −2 4
   14
D   32
D 2

From above calculations, we have

1 1
D
n  D   14    1.17
12
2
1    D  
n  1  
2 2
S 
D D  
n
 

1  14 2  1
  32    15.67  1.42
11  12  11

 SD  1.42  1.19

78
Substituting these values in equation (3), we have Small Sample Tests

 1.17 1.17
t   3.44
1.19 / 12 0.34
The critical value of test statistic t for left-tailed test corresponding (n-1) = 11
df at 5% level of significance is  t ( n 1), α   t (11), 0.05   1.796.

Since calculated value of test statistic t (= −3.44) is less than the critical value
(=−1.796), that means calculated value of t lies in rejection region, so we reject
the null hypothesis and support the alternative hypothesis i.e. support the claim
at 5% level of significance.
Thus, we conclude that samples fail to provide us sufficient evidence against
the claim so we may assume that memory practice session improves the
performance of children.
Example 6: Ten students were given a test in Statistics and after one month’s
coaching they were again given a test of the similar nature and the increase in
their marks in the second test over the first are shown below:
Roll No. 1 2 3 4 5 6 7 8 9 10
Increase in Marks 6 −2 8 −4 10 2 5 −4 6 0

Assuming that increment in marks follows normal distribution. Do the data

indicate that students have gained knowledge from the coaching at 1% level of
significance?
Solution: Here, we want to test that students have gained knowledge from the
coaching. If D denotes the average increment in the marks due to one month’s
coaching then our claim is D < 0 but here we are given increment
Di = (Yi – Xi) instead of Di = (Xi – Yi) so we take our claim is D > 0 and its
complement is D ≤ 0. Since complement contains the equality sign so we can
take the complement as the null hypothesis and the claim as the alternative
hypothesis. Thus,
H 0 :  D  0 and H1 :  D  0

Since the alternative hypothesis is right-tailed so the test is right-tailed test.

It is given that increment in marks after one month coaching follows normal
distribution and population variance is unknown. Also participants are same in
both situations before and after the coaching. And all the assumption of paired
t-test meet so we can go for paired t-test.
For testing H0, the test statistic is given by
D
t ~ t  n 1  under H 0 … (4)
SD / n
Calculation for D and S D :
Roll
1 2 3 4 5 6 7 8 9 10 Total
No.
D 6 −2 8 −4 10 2 5 −4 6 0  D  27
2 2
D 36 4 64 16 100 4 25 16 36 0 D  301

From above calculations, we have

1 1
D
n  D   27  2.7
10
79
Testing of Hypothesis 2
1    D  
n  1  
2 2
S 
D D  
n
 

1   27 2  1
 301    228.1  25.34
10  1  10  9

 SD  25.34  5.03
Substituting these values in equation (4), we have
2.7 2.7
t   1.70
5.03/ 10 1.59
The critical value of test statistic t for right-tailed test corresponding (n-1) = 9
df at 1% level of significance is t (n 1), α  t (9), 0.01  2.821.
Since calculated value of test statistic t (= 1.70) is less than the critical value
(= 2.821), that means calculated value of test statistic t lies in non-rejection
region, so we do not reject null hypothesis and reject the alternative hypothesis
i.e. we reject our claim at 1% level of significance.
Thus, we conclude that sample provides us sufficient evidence against the
claim so students are not gained knowledge from the coaching.
Now, you can try the following exercises.
E6) To verify whether the programme “Post Graduate Diploma in Applied
Statistics (PGDAST)” improved performance of the graduate students
in Statistics, a similar test was given to10 participants both before and
after the programme. The original marks out of 100 (before course)
recorded in an alphabetical order of the participants are 42, 46, 50, 36,
44, 60, 62, 43, 70 and 53. After the course the marks in the same order
are 45, 46, 60, 42, 60, 72, 63, 43, 80 and 65. Assuming that marks of
the students before and after the course follow normal distribution. Test
whether the programme PGDAST has improved the performance of the
graduate students in Statistics at 5% level of significance?
E7) A drug is given to 8 patients and the increments in their blood pressure
are recorded to be 4, 0, 7, −2, 0, −3, 2, 0. Assume that increment in their
blood pressure follows normal distribution. Is it reasonable to believe
that the drug has no effect on the change of blood pressure at 5% level
of significance?

11.6 TESTING OF HYPOTHESIS FOR

POPULATION CORRELATION COEFFICIENT
USING t-TEST
In Unit 6 of MST-002, we have discussed the concept of correlation. Where,
we studied that if two variables are related in such a way that change in the
value of one variable affects the value of another variable then the variables are
said to be correlated or there is a correlation between these two variables.
Correlation can be positive, which means the variables move together in the
same direction, or negative, which means they move in opposite directions.
And correlation coefficient is used to measure the intensity or degree of linear
relationship between two variables. The value of correlation coefficient varies

80
between −1 and +1, where −1 representing a perfect negative correlation, 0 Small Sample Tests
representing no correlation, and +1 representing a perfect positive correlation.
Sometime, the sample data indicate for non-zero correlation but in population
they are uncorrelated ( = 0).
For example, price of tomato in Delhi (X) and in London (Y) are not correlated
in population ( = 0). But paired sample data of 20 days of prices of tomato at
both places may show correlation coefficient (r) ≠ 0. In general, in sample data
r ≠ 0 does not ensure in population  ≠ 0 holds.
In this section, we will know how we test the hypothesis that population
correlation coefficient is zero.
Assumptions
This test works under following assumptions:
(i) The characteristic under study follows normal distribution in both the
populations. In other words, both populations from which random
samples are drawn should be normal with respect to the characteristic of
interest.
(ii) Samples observations are random.
Let us consider a random sample (X1, Y1), (X2, Y2), …, (Xn, Yn) of size n taken
from a bivariate normal population. Let  and r be the correlation coefficients
of population and sample data respectively.
Here, we wish to test the hypothesis about population correlation coefficient
(), that is, linear correlation between two variables X and Y in the population,
so we can take the null hypothesis as
H0 :   0 and H1 :   0 for two-tailed test   Here,    and 0  0 if 
we compare it with general 
procedure given inSection11.2.

H 0 :   0 and H1 :   0 
or   for one-tailed test 
H 0 :   0 and H1 :   0 
For testing the null hypothesis, the test statistic t is given by
r n2
t ~ t  n 2 
1 r2
which follows t-distribution with n – 2 degrees of freedom.
After substituting values of r and n, we find out calculated value of test statistic
t. Then we look for critical (or cut-off or tabulated) value(s) of test statistic t
from the t-table. On comparing calculated value and critical value(s), we take
the decision about the null hypothesis as discussed in Section 11.2.
Let us do some examples of testing of hypothesis that population correlation
coefficient is zero.
Example 7: A random sample of 18 pairs of observations from a normal
population gave a correlation coefficient of 0.7. Test whether the population
correlation coefficient is zero at 5% level of significance.
Solution: Given that
n = 18, r = 0.7
Here, we wish to test that population correlation coefficient () is zero so our
claim is  = 0 and its complement is  ≠ 0. Since the claim contains the
81
Testing of Hypothesis equality sign so we can take the claim as the null hypothesis and complement
as the alternative hypothesis. Thus,
H 0 :   0 and H1 :   0

Since the alternative hypothesis is two-tailed so the test is two-tailed test.

Here, we want to test the hypothesis regarding population correlation
coefficient is zero and the populations under study follow normal distributions,
so we can go for t-test.
For testing the null hypothesis, the test statistic t is given by
r n2
t
1  r2
0.7 18  2 0.7  4 2.8
    3.94
1   0.7 
2
0.51 0.71

The critical value of test statistic t for two-tailed test corresponding (n-2) = 16
df at 5% level of significance are  t ( n  2 ), α / 2   t (16 ), 0.025   2.120.

Since calculated value of test statistic t (= 3.94) is greater than the critical
values (= ± 2.120), that means calculated value of test statistic t lies in rejection
region, so we reject the null hypothesis. i.e. we reject the claim at 5% level of
significance.
Thus, we conclude that sample provides us sufficient evidence against the
claim so there exists a relationship between two variables.
Example 8: A random sample of 15 married couples was taken from a
population consisting of married couples between the ages of 30 and 40. The
correlation coefficient between the IQs of husbands and wives was found to be
0.68. Assuming that the IQs of husbands and wives follow normal distributions
then test that IQs of husbands and wives in the population are positively
correlated at1% level of significance.
Solution: Given that
n = 15, r = 0.68
Here, we wish to test that IQs of husbands and wives in the population are
positively correlated. If  denote the correlation coefficient between IQs of
husbands and wives in the population then the claim is  > 0 and its
complement is  ≤ 0. Since complement contains the equality sign so we can
take the complement as the null hypothesis and the claim as the alternative
hypothesis. Thus,
H 0 :   0 and H1 :   0
Since the alternative hypothesis is right-tailed so the test is right-tailed test.
Here, we want to test the hypothesis regarding population correlation
coefficient is zero and the populations under study follow normal distributions,
so we can go for t-test.
For testing the null hypothesis, the test statistic t is given by
r n2
t
1  r2

82
 0.68 15  2 0.68  3.61 Small Sample Tests
   3.36
1   0.68 
2 0.73

The critical value of test statistic t for right-tailed test corresponding (n-2) = 13
df at 1% level of significance is t ( n  2 ), α  t (13 ), 0.01  2.650.

Since calculated value of test statistic t (= 3.36) is greater than the critical value
(= 2.650), that means calculated value of test statistic t lies in rejection region,
so we reject the null hypothesis and support the alternative hypothesis i.e. we
support our claim at 1% level of significance.
Thus, we conclude that sample fail to provide us sufficient evidence against the
claim so we may assume that the correlation between IQs of husbands and
wives in the population is positive.
In the same way, you can try the following exercise.
E8) Twenty families were selected randomly from a colony to determine
that correlation exists between family income and the amount of money
spent per family member on food each month. The sample correlation
coefficient was computed as r = 0.40. Assuming that the family income
and the amount of money spent per family member on food each month
follow normal distributions then test that there is a positive linear
relationship between the family income and the amounts of money
spent per family member on food each month in colony at 1% level of
significance.
We now end this unit by giving a summary of what we have covered in it.

11.7 SUMMARY
In this unit, we have discussed the following points:
1. Need of small sample tests.
2. Procedure of testing a hypothesis for t-test.
3. Testing of hypothesis for population mean using t-test.
4. Testing of hypothesis for difference of two population means when samples
are independent using t-test.
5. The procedure of paired t-test for testing of hypothesis for difference of two
population means when samples are dependent or paired.
6. Testing of hypothesis for population correlation coefficient using t-test.

11.8 SOLUTIONS /ANSWERS

E1) Since calculated value of test statistic t is based on 15 df therefore, we
use row for 15 df in the t-table and move across this row to find the
values in which calculated t-value lies. Since calculated t-value falls
between 2.131 and 2.602, which are corresponding to the values of
one-tailed area α = 0.025 and 0.01 respectively, therefore p-value will
lie between 0.01 and 0.025, that is, 0.01  p-value  0.025
Since test is two-tailed, therefore, the values are doubled, so
0.02  2  0.01  p-value  2  0.025  0.05
E2) Here, we are given that
83
Testing of Hypothesis n  16,  0  18000, X  20000, S  6000
Here, we want to test that manufacturer’s claim is true that the average
life () of tyres is 18000 km. So claim is µ = 18000 and its complement
is µ ≠ 18000. Since the claim contains the equality sign so we can take
the claim as the null hypothesis and complement as the alternative
hypothesis. Thus,
H0 :  = µ0 = 18000 [average life of tyres is 18000 km]
H1 :  ≠ 18000 [average life of tyres is not 18000 km]
Here, population SD is unknown and population under study is given to
be normal. So we can go for t-test.
For testing the null hypothesis, the test statistic t is given by
X  µ 0 20000  18000 2000
t    1.33
S/ n 6000 / 16 1500
The critical value of test statistic t for two-tailed test corresponding
(n-1) = 15 df at 1% level of significance are ± t(15) ,0.005 = ± 2.947.
Since calculated value of test statistic t (= 1.33) is less than the critical
(tabulated) value (= 2.947) and greater that critical value (= − 2.947),
that means calculated value of test statistic lies in non-rejection region,
so we do not reject the null hypothesis i.e. we support the
manufacture’s claim at 1% level of significance.
Thus, we conclude that sample fails to provide sufficient evidence
against the claim so we may assume that manufacturer’s claim is true.
E3) Here, we want to test that the average weight () of all cadets of the
centre is 60 kg. So our claim is µ = 60 and its complement is µ ≠ 60.
Since the claim contains the equality sign so we can take the claim as
the null hypothesis and complement as the alternative hypothesis. Thus,
H 0 :    0  60  average weight of all cadets is 60 kg 
H1 :   60 average weight of all cadets is not 60 kg 
Since the alternative hypothesis is two-tailed so the test is two-tailed
test.
Here, population SD is unknown and population under study is given to
be normal. So we can go for t-test.
For testing the null hypothesis, the test statistic t is given by
X  µ0
t ~ t ( n 1) under H 0 … (5)
S/ n
Before moving further, first we have to calculate value of X and S.
Calculation for X and S:
Sample value (X) X  X  2
X  X
48 −15 225
50 −13 169
62 −1 1
75 12 144
80 17 289
60 −3 9
70 7 49

84
 56 −7 49 Small Sample Tests
52 −11 121
77 14 196
 X  630  X  X
2
1252

From the above calculation, we have

1 1
X
n
 X  10  630  63
1 2
S2 
n 1
  X  X

1
  1252  139.11
10  1
 S  139.11  11.79
Putting the values in equation (5), we have
63  60
t
11.79 / 10
3
  0.80
3.73
The critical values of test statistic t for two-tailed test corresponding
(n-1) = 9 df at 5% level of significance are ± t(9) ,0.025 = ± 2.262.
Since calculated value of test statistic t (= 0.80) is less than the critical
value (= 2.262) and greater than the critical value (= −2.262), that
means calculated value of test statistic t lies in non-rejection region so
we do not reject H0 i.e. we support the claim at 5% level of
significance.
Thus, we conclude that sample fails to provide sufficient evidence
against the claim so we may assume that the average weight of all the
cadets of given centre is 60 kg.
E4) Here, we want to test that there is no difference between drugs A and B
with regard to their mean weight increment. If 1 and 2 denote the
mean weight increment due to drug A and drug B respectively then our
claim is µ1 = 2 and its complement is µ1 ≠ 2. Since the claim contains
the equality sign so we can take the claim as the null hypothesis and
complement as the alternative hypothesis. Thus,
H 0 : 1   2 effect of both drugs is same 
H 1 : 1   2 effect of both drugs is not same 
Since the alternative hypothesis is two-tailed so the test is two-tailed
test.
Since it is given that increments in the weight due to both drugs follow
normal distributions with equal and unknown variances and other
assumptions of t-test for testing a hypothesis about difference of two
population means also meet. So we can go for this test.
For testing the null hypothesis, the test statistic t is given by

85
Testing of Hypothesis XY
t ~ t ( n 1 n 22) under H0 … (6)
1 1
Sp 
n1 n 2

Assume, a = 8, b = 12 and use short-cut method to find X, Y and Sp .

Calculation for X, Y and S p :

Drug A Drug B

X d1 = (X-a) d1 2 Y d2 = (Y-b) d2 2
a=8 b = 12
5 ‒3 9 9 ‒3 9
8 0 0 10 ‒2 4
7 ‒1 1 15 3 9
10 2 4 12 0 0
9 1 1 14 2 4
6 ‒2 4 8 ‒4 16
12 0 0

d 1  3 d 2
1  19 d 2  4  d22  42
From above calculation, we have
1 1
n1  1
Xa d  8   3   7.5,
6
1 1
Y  b   d 2  12   4   11.43
n2 7
2 2
1     d1      d2   
n1  n 2  2    1     d2 
2 2 2
S 
p  d  
n1   n2  

2 2
1    3    4   
  19    42  
6  7  2  6   7 
   
1
 17.5  39.71  5.20
11
 Sp  5.20  2.28
Putting these values in equation (6), we have
7.5  11.43 3.93 3.93
t    3.07
1 1 2.28  0.56 1.28
2.28 
6 7
The critical values of test statistic t for two-tailed test corresponding
(n1 + n2 -2) = 11 df at 5% level of significance are ± t(11) ,0.025 = ± 2.201.
Since calculated value of test statistic t (= −3.07) is less than the critical
values (= ± 2.201) that means calculated value of test statistic t lies in
rejection region, so we reject the null hypothesis i.e. we reject the claim
at 5% level of significance.

86
 Thus, we conclude that samples provide us sufficient evidence against Small Sample Tests
the claim so drugs A and B differ significantly. Any one of them is
better than other.
E5) Here, we are given that
n1  13, X  4.6, S1  0.5,

n 2  13, Y  5.0, S2  0.3

Therefore, the pooled variance S 2p can be calculated as

1
S2p   n1  1 S12   n 2  1 S22 
n1  n 2  2

1 12   0.5  2  12   0.3 2 


13  13  2  
1
  3.00  1.08  0.17
24
 Sp  0.17  0.41

We want to test that the there is significant improvement in wheat

production due to fertilizer. If 1 and 2 denote the mean wheat
productions without and with the fertilizer respectively then our claim
is 1 < 2 and its complement is 1 ≥ 2. Since complement contains the
equality sign so we can take the complement as the null hypothesis and
the claim as the alternative hypothesis. Thus,
H0 : 1  2 and H1 : 1  2
Since the alternative hypothesis is left-tailed so the test is left-tailed
test.
Since it is given that yield of wheat with and without fertilizer follow
normal distributions with equal and unknown variances and other
assumptions of t-test for testing a hypothesis about difference of two
population means also meet. So we can go for this test.
For testing the null hypothesis, the test statistic t is given by
XY
t ~ t (n 1n 2 2) under H0
1 1
Sp 
n1 n 2
4.6  5.0  0.4  0.4
     2.5
1 1 0.41  0.39 0.16
0.41 
13 13
The critical values of test statistic t for left-tailed test corresponding
(n1 + n2 -2) = 24 df at 1% level of significance is − t(24), 0.01 = −2.492.
Since calculated value of test statistic t (= −2.5) is less than the critical
value (= −2.492), that means calculated value of test statistic t lies in
rejection region, so we reject the null hypothesis and support the
alternative hypothesis i.e. we support our claim at 1% level of
significance.

87
Testing of Hypothesis Thus, we conclude that samples fail to provide us sufficient evidence
against the claim so we may assume that there is significant
improvement in wheat production due to fertilizer.
E6) Here, we want to test whether the programme PGDAST has improved
the performance of the graduate students in Statistics. If 1 and 2
denote the average marks before and after the programmed so our claim
is 1 < 2 and its complement is 1 ≥ 2. Since complement contains the
equality sign so we can take the complement as the null hypothesis and
the claim as the alternative hypothesis. Thus,
H0 : 1  2 and H1 : 1  2
Since the alternative hypothesis is left-tailed so the test is left-tailed
test.
It is a situation of before and after. Also, the marks of the students
before and after the programme PGDAST follow normal distributions.
So, population of differences will also be normal. Also all the
assumptions of paired t-test meet. So we can go for paired t-test.
For testing the null hypothesis, the test statistic t is given by
D
t ~ t  n 1  under H 0 … (7)
SD / n
Calculation for D and S D :

Participant Marks D = (X−Y) D2

Before (X) After (Y)

1 42 45 −3 9
2 46 46 0 0
3 50 60 −10 100
4 36 42 −6 36
5 44 60 −16 256
6 60 72 −12 144
7 62 63 −1 1
8 43 43 0 0
9 70 80 −10 100
10 53 65 −12 144
2
 D  70 D  790

From above calculation, we have

1 1
n
D D   70    7
10
2
1    D 
n 1 
2 2
S 
D  D  
n 



1
790 
 702  1
    300  33 . 33
9 10  9

 SD  33.33  5.77

88
 Putting the values in equation (7), we have Small Sample Tests

7.0
t  3.83
5.77 / 10
The critical value of test statistic t for left-tailed test corresponding
(n-1) = 9 df at 5% level of significance is − t(9), 0.05 = −1.833.
Since calculated value of test statistic t (= −3.83) is less than the critical
(tabulated) value (= −1.833), that means calculated value of test statistic
t lies in rejection region, so we reject the null hypothesis and support
the alternative hypothesis i.e. we support our claim at 5% level of
significance.
Thus, we conclude that samples fail to provide us sufficient evidence
against the claim so we may assume that the participants have
significant improvement after the programme “Post Graduate Diploma
in Applied Statistics (PGDAST)”.
E7) Here, we want to test that the drug has no effect on change in blood
pressure. If D denotes the average increment in the blood pressure
before drug then our claim is D = 0 and its complement is D ≠ 0.
Since the claim contains the equality sign so we can take the claim as
the null hypothesis and complement as the alternative hypothesis. Thus,
H 0 :  D  1 2  0  the drug has no effect 

H1: µ D  0  the drug has an effect 

It is given that increment in the blood pressure follows normal
distribution. Also patients are same in both situations before and after
the drug. And all the assumption of paired t-test meet. So we can go for
paired t-test.
For testing H0, the test statistic is given by
D
t ~ t  n 1  under H 0 … (8)
SD / n
Calculation for D and S D :
Patient
1 2 3 4 5 6 7 8 Total
Number
D 4 0 7 −2 0 −3 2 0 D  8
 D  D 3 −1 6 −3 −1 −4 1 −1
2 2
 D  D 9 1 36 9 1 16 1 1  D  D  74

From above calculation, we have

1 1
D
n  D  8 1
8
1 2 1
S2D 
n 1
  D  D    74  10.57
7
 SD  10.57  3.25
Putting the values in test statistic, we have

89
Testing of Hypothesis 1 1
t   0.87
3.25 / 8 1.15
The critical value of test statistic t for two-tailed test corresponding
(n-1) = 7 df at 5% level of significance are ± t(7), 0.025 = ± 2.365.
Since calculated value of test statistic t (= 0.87) is less than the critical
value (= 2.365) and greater than the critical value (=−2.365) that means
calculated value of test statistic t lies in non-rejection region, so we do
not reject the null hypothesis i.e. we support the claim at 5% level of
significance.
Thus, we conclude that samples fail to provide us sufficient evidence
against the claim so we may assume that the drug has no effect on the
change of blood pressure of patients.
E8) We are given that
n = 20, r = 0.40
and we wish to test that there is a positive linear relationship between
the family income and the amounts of money spent per family member
on food each month in colony. If  denote the correlation coefficient
between the family income and the amounts of money spent per family
member then the claim is  > 0 and its complement is  ≤ 0. Since
complement contains the equality sign so we can take the complement
as the null hypothesis and the claim as the alternative hypothesis. Thus,
H0 :   0 and H0 :   0
Since the alternative hypothesis is right-tailed so the test is right-tailed
test.
For testing the null hypothesis, the test statistic t is given by
r n2
t
1  r2
0.40 20  2 0.40  4.24
   1.84
1   0.40 
2 0.92

The critical value of test statistic t for right-tailed test corresponding

(n-2) = 18 df at 1% level of significance is t ( n  2 ), α  t (1 8 ), 0.01  2.552.

Since calculated value of test statistic t (= 1.84) is less than the critical
value (= 2.552), that means calculated value of test statistic t lies in non-
rejection region, so we do not reject the null hypothesis and reject the
alternative hypothesis i.e. we reject our claim at 1% level of significance.
Thus, we conclude that sample provide us sufficient evidence against the
claim so there is no positive linear correlation between the family income
and the amounts of money spent per family member on food each month
in colony.

90
UNIT 12 CHI-SQUARE AND F-TESTS
Structure
12.1 Introduction
Objectives
12.2 Testing of Hypothesis for Population Variance Using χ2-Test
12.3 Testing of Hypothesis for Two Population Variances Using F-Test
12.4 Summary
12.5 Solutions / Answers

12.1 INTRODUCTION
Recall from the previous unit when we test the hypothesis about the difference
of means of two populations, t-test needs an assumption of equality of
variances of two populations under study. Other than this there are situations
where we want to test the hypothesis about the variances of two populations.
For example, an economist may want to test whether the variability in incomes
differ in two population. In such situations, we use F-test when the populations
under study follow the normal distributions.

Similarly, there are many other situations where we need to test the hypothesis
about the hypothetical or specified value of the variance of the population
under study. For example, the manager of the electric bulbs company would
probably be interested whether or not the variability in the life of bulbs is
within acceptable limits, the product controller of a milk company may be
interested in the variance of the amount of fat in the whole milk processed by
the company is no more than the specified level. In such situations, we use
χ2-test when the population under study follows the normal distribution.
This unit is divided into five sections. Section 12.1 is described the need of χ2
and F-tests. χ2-test for testing the hypothesis about the population variance is
discussed in Section 12.2. And F-test for equality of variances of two
populations is discussed in Section 12.3. Unit ends by providing summary of
what we have discussed in this unit in Section 12.4 and solution of exercises in
Section 12.5.
Objectives
After studying this unit, you should be able to:

 describe the testing of hypothesis for population variance; and

 explain the testing of hypothesis for two population variances.

12.2 TESTING OF HYPOTHESIS FOR

POPULATION VARIANCE USING χ2-TEST
In Section 11.3 of previous unit, we have discussed testing of hypothesis for
population mean when the characteristic under study follows the normal
distribution but when analysing quantitative data, it is often important to draw
conclusion about the average as well as the variability of a characteristic under
study. For example, if a company manufactured the electric bulbs then the
91
Testing of Hypothesis manager of the company would probably be interested in determining the
average life of the bulbs and also determining whether or not the variability in
the life of bulbs is within acceptable limits, the product controller of a milk
company may be interested in the variance of the amount of fat in the whole
milk processed by the company is no more than the specified level, etc.
The procedure of testing a hypothesis for population variance or standard
deviation is similar to the testing of population mean. The basic difference is
that here we use chi-square test instead of t-test because here the sampling
distribution of test statistic follows the chi-square distribution.
Assumptions
This test works under the following assumptions:
(i) The characteristic under study follows normal distribution. In other
words, populations from which random sample is drawn should be
normal with respect to the characteristic of interest.
(ii) Sample observations are random and independent.
Let X1 , X 2 ,..., X n be a random sample of size n drawn from a normal
population with mean  and variance σ2, where  and σ2 are unknown.
The general procedure of this test is explained below in detail:
As we are doing so far in all tests, first step in hypothesis testing problems is to
setup null and alternative hypotheses. Here, we want to test the claim about the
hypothesized or specified value σ02 of population variance σ 2 so we can take
our null and alternative hypotheses as
H 0 : 2  02 and H1 : 2  02 for two-tailed test 
H0 : 2  02 and H1 : 2  02 
or  for one-tailed test
H0 : 2  02 and H1 : 2  02 
For testing the null hypothesis, the test statistic χ2 is given by
2

χ 2

X i X

(n  1)S 2
~ χ 2  n 1  under H 0 ... (1)
σ 02 σ 02
1 2
where, S2 
n  1
X X
Here, the test statistic χ2 follows chi square distribution with (n − 1) degrees of
freedom as we have discussed in Unit 3 of this course.
After substituting values of n, S and 02 , we get calculated value of test
statistic. Let 2cal be the calculated value of test statistic χ2.
Obtain the critical value(s) or cut-off value(s) in the sampling distribution of
the test statistic χ2 and construct rejection (critical) region of size . The critical
value of the test statistic χ2 for various df and different level of significance 
are given in Table III of the Appendix given at the end of the Block 1of this
course.
After doing all the calculation discussed above, we have to take the decision
about rejection or non-rejection of the null hypothesis. The procedure of taking
the decision about the null hypothesis is explained is the next page:
92
For one-tailed test: Chi-square and F-Tests

Case I: When H0 :2  20 and H1 : 2  02 (right-tailed test)

In this case, the rejection (critical) region falls under the right tail of
the probability curve of the sampling distribution of test statistic χ2.
Suppose  2( ),  is the critical value at  level of significance where,
ν = n ‒1, so entire region greater than or equal to  2( ),  is the rejection
region and less than  (2 ),  is the non-rejection region as shown in Fig.
12.1.
If 2cal  (2 ),  , that means calculated value of test statistic lies in the
rejection (critical) region, then we reject the null hypothesis H0 at 
level of significance. Therefore, we conclude that sample data
provides us sufficient evidence against the null hypothesis and there
is a significant difference between hypothesized value and observed
value of the population variance σ2.
If 2cal  (2 ),  , that means calculated value of test statistic lies in non-
rejection region, then we do not reject the null hypothesis H0 at  Fig. 12.1
level of significance. Therefore, we conclude that the sample data
fails to provide us sufficient evidence against the null hypothesis and
the difference between hypothesized value and observed value of the
population variance σ2 due to fluctuation of sample.
Case II: When H0 :2  02 and H1 : 2  02 (left-tailed test)
In this case, the rejection (critical) region falls under the left tail of the
probability curve of the sampling distribution of test statistic χ2.
Suppose  2( ), (1 ) is the critical value at  level of significance then
entire region less than or equal to  2( ), (1 ) is the rejection (critical)
region and greater than  (2 ),(1 ) is the non-rejection region as shown
in Fig. 12.2.
If  2cal   (2 ),(1 ) , that means calculated value of test statistic lies in
rejection (critical) region, then we reject the null hypothesis H0 at 
Fig. 12.2
level of significance.
For  2cal   (2 ), (1 ) , that means calculated value of test statistic lies in
the non-rejection region, then we do not reject the null hypothesis H0
at  level of significance.
For two-tailed test:
When H0 :2  20 and H1 : 2  02
In this case, the rejection region falls under both tails of the
probability curve of sampling distribution of the test statistic χ2 and
half the area (α) i.e. α/2 of rejection (critical) region lies at left tail
and other half on the right tail. Suppose  2( ),(1 / 2) and 2( ),  / 2 are the
two critical values at the left-tailed and right-tailed respectively on
pre-fixed -level of significance. Therefore, entire region less than or
equal to  2( ),(1 / 2) and greater than or equal to (2 ),  / 2 are the rejection
Testing of Hypothesis (critical) regions and between  2( ),(1 / 2) and  2( ),  / 2 is the non-
rejection region as shown in Fig. 12.3.
If  2cal   (2 ),  / 2 or 2cal  (2 ),(1 / 2) , that means calculated value of test
statistic χ2 lies in rejection (critical) region, then we reject the null
hypothesis H0 at  level of significance.
If  (2 ),(1 / 2)   cal
2
  (2 ),  / 2 , that means calculated value of test
statistic χ2 lies in non-rejection region, then we do not reject the null
Fig. 12.3
hypothesis H0 at  level of significance.
Procedure of taking the decision about the null hypothesis on the basis of
p-value:
You have done it in so many test, it has become a routine thing for you. You
know that to take the decision about the null hypothesis on the basis of p-value,
the p-value is compared with level of significance (α) and if p-value is less than
or equal to α then we reject the null hypothesis and if the p-value is greater than
α we do not reject the null hypothesis.
For χ2-test, p-value is defined as:
For one-tailed test:
For H1 : 2  02 (right-tailed test)
p-value = P  2  cal
2


For H1 : 2  02 (left-tailed test)

p-value = P  2  cal
2


For two-tailed test: H1 : 2  20

For two-tailed test the p-value is approximated as
p-value = 2P  2  cal
2

The p-value for χ2-test can be obtained with the help of the Table-III
(χ2-table) given in the Appendix at the end of Block 1 of this course. Similar to
t-test, this table gives the χ2 values corresponding to the standard values of α
such as 0.995, 0.99, 0.10, 0.05, 0.025, 0.01, etc only. therefore, the exact
p-value is not obtained with the help of this table and we can approximate the
p-value for this test.
For example, if test is right-tailed and calculated (observed) value of test
statistic χ2 is 25.10 with 12 df then p-value is calculated as:
Since test statistic is based on the 12 df therefore, we use row for 12 df in the
χ2-table and move across this row to find the values in which calculated
χ2-value falls. Since calculated χ2-value falls between 23.24 and 26.22,
corresponding to one-tailed area α = 0.025 and 0.01 respectively, therefore
p-value lies between 0.01 and 0.025, that is,
0.01  p-value  0.025
If in the above example the test is two-tailed then the two values 0.01 and
0.005 would be doubled for p-value, that is,
2  0.01  0.02  p-value  0.05  2  0.025
Note 1: With the help of computer packages and softwares such as SPSS, SAS, Chi-square and F-Tests
MINITAB, EXCEL, etc. we can find the exact p-value for χ2-test.
Let us do some examples to become more user friendly with the test explained
above.
Example 1: The variance of a certain dimension article produced by a machine
is 7.2 over a long period. A random sample of 20 articles gave a variance 8. Is
it justifiable to conclude that variability has increased at 5% level of
significance assuming that the measurement of dimension article is normally
distributed?
Solution: Here, we are given that
Sample size = n = 20
Sample variance = S2 = 8
Specified value of population variance under test = σ20  7.2
Here, we want to test that variability of dimension article produced by a
machine has increased. Since variability is measured in terms of variance (  2 )
so our claim is  2  7.2 and its complement is  2  7.2. Since complement
contains the equality sign so we can take the complement as the null hypothesis
and the claim as the alternative hypothesis. Thus,
H0 : 2  20  7.2
H 1 :  2  7.2  variability of dimension article has increased 
Since the alternative hypothesis is right-tailed so the test is right-tailed test.
Here, we want to test the hypothesis about the population variance and sample
size is small n = 20(< 30). Also we are given that measurement of dimension
article follows normal distribution so we can go for χ2 test for population
variance.
So, test statistic is given by
 n  1  S2
χ2  2
~ χ 2 n 1  under H0
σ
19  8
  21.11
7.2
The critical (tabulated) value of test statistic χ2 for right-tailed test
corresponding (n-1) = 19 df at 5% level of significance is
 2(n 1),    (19),0.05
2
= 30.14.
Since calculated value of test statistic (= 21.11) is less than the critical
(tabulated) value (= 30.14), that means calculated value of test statistic lies in
non-rejection region as shown in Fig. 12.4, so we do not reject the null
hypothesis and reject the alternative hypothesis i.e. we reject our claim at 5%
level of significance.
Fig. 12.4
Decision according to p-value:
Since test statistic is based on 19 df therefore, we use row for 19 df in the χ2-
table and move across this row to find the values in which calculated χ2-value
falls. Since calculated χ2-value falls between 11.65 and 27.20, corresponding to
one-tailed area α = 0.90 and 0.10 respectively therefore p-value lies between
0.10 and 0.90, that is,
0.10  p-value  0.90
Testing of Hypothesis Since p-value is greater than  ( 0.05) so we do not reject the null hypothesis
at 5% level of significance.
Thus, we conclude that sample provides us sufficient evidence against the
claim so we may assume that the variability of dimension article produced by a
machine is not increased.
Example 2: The12 measurements of the same object on an instrument are
given below:
1.6, 1.5, 1.3, 1.5, 1.7, 1.6, 1.5, 1.4, 1.6, 1.3, 1.5, 1.5
If the measurement of the instrument follows normal distribution then carry out
the test at 1% level of significance that variance in the measurement of the
instrument is less than 0.016.
Solution: Here, we are given
Sample size = n = 12
Specified value of population variance under test 20  0.016
Here, we want to test that the variance (σ 2 ) in the measurements of the
instrument is less than 0.016. So our claim is  2  0.016 and its complement
is  2  0.016. Since complement contains the equality sign so we can take
complement as null hypothesis and claim as the alternative hypothesis. Thus,
H0 : 2  02  0.016 and H1 : 2  0.016
Since the alternative hypothesis is left-tailed so the test is left-tailed test.
Here, we want to test the hypothesis about the population variance and sample
size is small n = 12(< 30). Also we are given that the measurement of the
instrument follows normal distribution so we can go for χ2-test for population
variance.
For testing the null hypothesis, the test statistic χ2 is given by
2

χ 2

 X i X
~ χ 2 n 1  under H 0 ... (2)
2
σ 0
2
Calculation for   Xi  X  :
X X  X X  X
2

1.6 0.1 0.01

1.5 0 0
1.3 -0.2 0.04
1.5 0 0
1.7 0.2 0.04
1.6 0.1 0.01
1.5 0 0
1.4 -0.1 0.01
1.6 0.1 0.01
1.3 -0.2 0.04
1.5 0 0
1.5 0 0
  18
X 0
 X  X
2
 0.16

96
From above calculation, we have Chi-square and F-Tests

1 1
n
X X  18  1.5
12
2
Putting the values of  X  X  and σ 02 in equation (2), we have
2

χ 2

 X i X0.16

 10
2
σ 0 0.016
The critical value of test statistic χ2 for left-tailed test corresponding
(n-1) = 11 df at 5% level of significance is (2n 1),(1 )   2(11),0.95 = 4.57.

Since calculated value of test statistic (= 10) is greater than the critical value
(= 4.57), that means calculated value of test statistic lies in non-rejection region
as shown in Fig. 12.5 so we do not reject the null hypothesis and reject the
alternative hypothesis i.e. we reject the claim at 5% level of significance.
Thus, we conclude that sample provide sufficient evidence against the claim so
the variance in the measurement of the instrument is not less than 0.016.
In the same way, you can try the following exercises. Fig. 12.5

E1) An ambulance agency claims that the standard deviation in the length of
serving times is less than 15 minutes. Investigator suspects that this
claim is wrong and takes a random sample of 20 serving times which
has a standard deviation of 17 minutes. Assume that the service time of
the ambulance follows normal distribution. Test at  = 0.01, is there
enough evidence to reject the agency’s claim?
E2) A cigarette manufacturer claims that the variance of nicotine content of
its cigarettes is 0.62. Nicotine content is measured in milligrams and is
normally distributed. A sample of 25 cigarettes has a variance of 0.65.
Test the manufacturer’s claim at 5% level of significance.

12.5 TESTING OF HYPOTHESIS FOR TWO

POPULATION VARIANCES USING
F-TEST
In Section 12.1 we have already mentioned that before applying t-test for
difference of two population means, one of the requirements is to check the Some author uses this
equality of variances of two populations. This assumption can be checked with test as the name
the help of F-test for two population variances. This F-test is also important in “homoscedastic test”.
a number of contexts. For example, an economist may want to test whether the Because two
populations with equal
variability in incomes differ in two populations, a quality controller may want variances are called
to test whether the quality of the product is changing over time, etc. homoscedastic. This
word is derived from
Assumptions two wards, homo
The assumptions for F-test for testing the variances of two populations are: means “the same” and
scedastic means
1. The populations from which the samples are drawn must be normally “variability”.
distributed.
2. The samples must be independent of each other.
Now, we come to the general procedure of this test.
Testing of Hypothesis Let X1 , X2 ,..., X n1 be a random sample of size n1 from a normal population
with mean 1 and variance σ12. Similarly, Y1 ,Y2 ,..., Yn2 be a random sample of
size n2 from another normal population with mean 2 and variance σ22. Here,
we want to test the hypothesis about the two population variances so we can
take our alternative null and hypotheses as
H 0 : 12  22  2 and H1 : 12  22  for two-tailed test 
H 0 : 12  22 and H1 : 12  22 
or  for one-tailed test 
H 0 : 12  22 and H1 : 12  22 
For testing the null hypothesis, the test statistic F is given by
S12
F ~ F( n1 1, n 2 1) under H0 … (3)
S22
1 2 1 2
where, S12 
n1  1
  X  X  and S22 
n2 1
  Y  Y .

For computational simplicity, we can also write

2 2
2 1  2   X  2 1 
2   Y 
S 
1
n1  1 
 X  n  and S2  n  1   Y  n 
1 2 2
   
The test statistic F follows F-distribution with ν1   n1  1  and ν 2   n 2  1 
degrees of freedom as discussed in Unit 4 of this course.
After substituting the values of S12 and S22 , we get calculated value of test
statistic. Let Fcal be calculated value of test statistic F.
Obtain the critical value(s) or cut-off value(s) in the sampling distribution of
the test statistic F and construct rejection (critical) region of size . The critical
values of the test statistic F for various df and different level of significance 
are given in Table IV (F-table) of the Appendix at the end of Block 1of this
course.
After doing all calculations discussed above, we have to take the decision about
rejection or non rejection of the null hypothesis. This is explained below:
In case of one-tailed test:
Case I: When H0 :σ12  σ22 and H1 : 12  22 (right-tailed test)
In this case, the rejection (critical) region falls at the right side of the
probability curve of the sampling distribution of test statistic F.
Fig. 12.6
Suppose F( 1 , 2 ),  is the critical value of test statistic F with
(ν1 = n1 – 1, ν2 = n2 – 1) df at  level of significance so entire region
greater than or equal to F( 1 , 2 ),  is the rejection (critical) region and
less than F( 1 , 2 ),  is the non-rejection region as shown in Fig. 12.6.

If Fcal  F( 1 , 2 ),  , that means calculated value of test statistic lies in

rejection (critical) region, then we reject the null hypothesis H0 at 
level of significance. Therefore, we conclude that samples data
 provide us sufficient evidence against the null hypothesis and there is Chi-square and F-Tests
a significant difference between population variances.
If Fcal  F( 1 , 2 ),  , that means calculated value of test statistic lies in
non-rejection region, then we do not reject the null hypothesis H0 at 
level of significance. Therefore, we conclude that the samples data
fail to provide us sufficient evidence against the null hypothesis and
the difference between population variances due to fluctuation of
sample.
Case II: When H0 :σ12  σ22 and H1 : 12  22 (left-tailed test)
In this case, the rejection (critical) region falls at the left side of the
probability curve of the sampling distribution of test statistic F.
Suppose F( 1 , 2 ), (1) is the critical value at  level of significance then
entire region less than or equal to F( 1 , 2 ),(1) is the rejection(critical)
region and greater than F( 1 , 2 ),(1) is the non-rejection region as shown
in Fig. 12.7.
If Fcal  F( 1 , 2 ),(1 ) , that means calculated value of test statistic lies in
rejection (critical) region, then we reject the null hypothesis H0 at  Fig. 12.7
level of significance.
If Fcal  F( 1 , 2 ),(1 ) , that means calculated value of test statistic lies in
non-rejection region, then we do not reject the null hypothesis H0 at 
level of significance.
Note 2: F-table mentioned above gives only the right tailed critical values with
different degrees of freedom at different level of significance. The left tailed
critical value of F-test can always be obtained by the formula given below (as
described in Unit 4 of this course):
1
F 1 , 2 ,1 
F 2 , 1 , 
In case of two-tailed test:
When H0 :σ12  σ22 and H1 : 12  22
In this case, the rejection (critical) region falls at both sides of the
probability curve of the sampling distribution of test statistic F and
half the area(α) i.e. α/2 of rejection (critical) region lies at left tail and
other half on the right tail.
Suppose F( 1 , 2 ),(1 / 2) and F( 1 , 2 ),  / 2 are the two critical values at the
left-tailed and right-tailed respectively on pre-fixed  level of
significance. Therefore, entire region less than or equal to F( 1 , 2 ), (1 / 2)
and greater than or equal to F( 1 , 2 ),  / 2 are the rejection (critical)
regions and between F( 1 , 2 ),(1 / 2) and F( 1 , 2 ),  / 2 is the non-rejection
Fig. 12.8
region as shown in Fig. 12.8.
If Fcal  F( 1 , 2 ),  / 2 or Fcal  F( 1 , 2 ), (1 / 2) , that means calculated value of
test statistic lies in rejection(critical) region, then we reject the null
hypothesis H0 at  level of significance.
Testing of Hypothesis If F( 1 , 2 ),(1 / 2)  Fcal  F( 1 , 2 ),  / 2 , that means calculated value of test
statistic F lies in non-rejection region, then we do not reject the null
hypothesis H0 at  level of significance.
Procedure of taking the decision about the null hypothesis on the basis of
p-value:
To take the decision about the null hypothesis on the basis of p-value, the p-
value is compared with level of significance (α) and if p-value is less than or
equal to α then we reject the null hypothesis and if the p-value is greater than α
we do not reject the null hypothesis.
For F-test, p-value can be defined as:
For one-tailed test:
For H1 : 12  22 (right-tailed test)
p-value = P  F  Fcal 

For H1 : 12  22 (left-tailed test)

p-value = P  F  Fcal 

For two-tailed test: H1 : 12  22

For two-tailed test the p-value is approximated as
p-value = 2P  F  Fcal 
The p-value for F-test can be obtained with the help of Table-IV (F-table)
given in the Appendix at the end of Block 1 of this course. Similar to t-test or
χ2-test, this table gives F values corresponding to the standard values of α such
as 0.10, 0.05, 0.025 and 0.01 only. Therefore with the help of F-table, we
cannot find the exact p-value. So we can approximate p-value for this test.
For example, if test is right-tailed and calculated (observed) value of test
statistic F is 2.65 with 24 degrees of freedom of numerator and 14 degrees of
freedom of the denominator then we can find the p-value as:
Since test statistic is based on (24, 14) df therefore, we move across all the
values of tabulated F corresponding to (24, 14) df at α such as 0.10, 0.05, 0.025
and 0.01 and find the values in which calculated F-value falls. Since we have F
tabulated with (24, 14) df at α = 0.10, 0.05, 0.025 and 0.01 as
α= 0.10 0.05 0.025 0.01
F(24, 14), α 1.94 2.35 2.79 3.43

Since calculated F-value (= 2.65) falls between 2.35 and 2.79, corresponding
to one-tailed area α = 0.05 and 0.025 respectively therefore p-value lies
between 0.0025 and 0.05, that is,
0.025  p-value  0.05
If in the above example the test is two-tailed then the two values 0.025 and
0.05 would be doubled for p-value, that is,
0.05  p-value  0.10
Note 3: With the help of computer packages and software such as SPSS, SAS,
MINITAB, EXCEL, etc. we can find the exact p-value for F-test.
Let us do some examples based on this test.

100
Example 3: The following data relate to the number of items produced in a Chi-square and F-Tests
shift by two workers A and B for some days:
A 26 37 40 35 30 30 40 26 30 35 45
B 19 22 24 27 24 18 20 19 25

Assuming that the parent populations are normal, can it be inferred that B is
more stable (or consistent) worker compared to A?
Solution: Here, we want to test that worker B is more stable than worker A.
As we know that stability of data is related to variance of the data. Smaller
value of the variance implies data that it is more stable. Therefore, to compare
stability of two workers, it is enough to compare their variances. If 12 and 22
denote the variances of worker A and worker B respectively then our claim is
12   22 and its complement is 12  22 . Since complement contains the
equality sign so we can take the complement as the null hypothesis and the
claim as the alternative hypothesis. Thus,
H0 : 12  22

H1 : 12   22  worker B is more stable than worker A 

Since the alternative hypothesis is right-tailed so the test is right-tailed test.

Here, we want to test the hypothesis about two population variances and
sample sizes n1 = 11(< 30) and n2 = 9 (< 30) are small. Also populations under
study are normal and both samples are independent so we can go for F-test for
two population variances.
For testing the null hypothesis, test statistic is given by
S12
F ~ F n1 1, n 2 1  … (4)
S22

1 n1 2 1 n2 2
2
where, S1  
n1  1 i1
 Xi  X  and S2
2  
n2  1 i1
 Yi  Y 

Calculation for S12 and S22 :

Items X  X  2 Items Y  Y 2

Produced by
X  X  Produced by
Y  Y
A   X  34  B   Y  22 
( Variable X) (Variable Y)

26 −8 64 19 −3 9
37 3 9 22 0 0
40 6 36 24 2 4
35 1 1 27 5 25
30 −4 16 24 2 4
30 −4 16 18 −4 16
40 6 36 20 −2 4
26 −8 64 19 −3 9
30 −4 16 25 3 9
35 1 1
45 11 121
Total = 374 0 380 198 0 80

101
Testing of Hypothesis
Therefore, we have
1 1
X   X   374  34
n1 11
and
1 1
Y   Y   198  22
n2 9
Thus,
1 2 1
S12 
n1  1
  X  X    380  38
10
1 2 1
S22 
n2  1
  Y  Y    80  10
8
Putting the value of S12 and S 22 in equation (4), we have
38
F  3.8
10
The critical (tabulated) value of test statistic F for right-tailed test
corresponding (n1-1, n2 -1) = (10, 8) df at 1% level of significance is
F( n 1, n 1),   F(10,8), 0.01 = 5.81.
1 2

Since calculated value of test statistic (= 3.8) is less than the critical value
Fig. 12.9 (= 5.81), that means calculated value of test statistic lies in rejection region as
shown in Fig. 12.9, so we do not reject the null hypothesis and reject the
alternative hypothesis i.e. we reject the claim at 1% level of significance.
Thus, we conclude that samples provide us sufficient evidence against the
claim so worker B is not more stable (or consistent) worker compared to A.
Example 4: Two random samples drawn from two normal populations gave
the following results:

Sample Size Mean Sum of Squares of Deviation

from the Mean
Sample I 9 59 26
Sample II 11 60 32

Test whether both samples are from the same normal populations?
Solution: Since we have to test whether both the samples are from same
normal population, therefore, we will test two hypotheses separately:
(i) Two population means are equal, i.e. H 0 : µ1  µ 2

(ii) Two population variances are equal, i.e. H0 : σ12  σ22

Since sample sizes are small and populations under study are normal so two
means will be tested using t-test whereas two variances will be tested using
F-test. But t-test is based on the prior assumption that both population
variances are same, therefore, first we apply F-test and later the t- test (when
F- test accepts equality hypothesis).
Given that
2
n1  9, X  59,  X  X   26
 2 Chi-square and F-Tests
n 2  11, Y  60,  Y  Y   32
Therefore,
1 2 1
S12 
n1  1
  X  X 
9 1
 26  3.25

1 2 1
S22 
n2  1
  Y  Y 
11  1
 36  3.60

First we want to test that the variances of both normal populations are equal so
our claim is 12   22 and its complement is 12  22 . Thus, we can take the null
and alternative hypotheses as
H0 : 12  22 and H1 : 12  22
Since the alternative hypothesis is two-tailed so the test is two-tailed test.
For testing this, the test statistic F is given by
S12
F ~ F n1 1,n 2 1 
S22
3.20
  0.88
3.65
The critical (tabulated) value of test statistic F for two-tailed test corresponding
(n1-1, n2 -1) = (8, 11) df at 5% level of significance are F( n1 1, n 2 1),  / 2 
1 1 1
F(8,11),0.025  3.66 and F( n 1, n 1),(1 / 2)     0.24.
1 2
F(n2 1,n1 1),  / 2  F(11,8),0.025 4.24

Since calculated value of test statistic (= 0.88) is less than the critical value
(= 3.66) and greater than the critical value (= 0.24), that means calculated value
of test statistic F lies in non-rejection region so we do not reject the null
hypothesis i.e. we support the claim.
Thus, we conclude that both samples may be taken from normal populations
having equal variances.
Now, we test that the means of two normal populations are equal so our claim
is 1  2 and its complement is 1  2. Thus, we can take the null and
alternative hypotheses as
H0 : 1  2 and H1 : 1  2
The test statistic is given by
XY
t … (5)
1 1
Sp 
n1 n 2
1   X  X 2   Y  Y 2 
where, S2p 
n1  n 2  2  

1
 26  32  1  58  3.22
9  11  2 18

103
Testing of Hypothesis
Sp  3.22  1.79

Putting the values of X,Y,Sp ,n1 and n 2 in equation (5), we have,

59  60 1 1
t     1.23
1 1 1.79  0.45 0.81
1.79 
9 11
The critical values of test statistic t for (n1 + n2 -2) = 18 df at 5% level of
significance for two-tailed test are ± t(18) ,0.025 = ± 2.101.
Since calculated value of test statistic t (= −1.23) is less than the critical value
(= 2.101) and greater than the critical value (= ‒2.101), that means calculated
value of test statistic t lies in non-rejection region so we do not reject the null
hypothesis i.e. we support the claim.
Thus, we conclude that both samples may be taken from normal populations
having equal means.
Hence, overall we conclude that both samples may come from the same normal
populations.
Now, you can try the following exercises.
E3) Two sources of raw materials are under consideration by a bulb
manufacturing company. Both sources seem to have similar
characteristics but the company is not sure about their respective
uniformity. A sample of 12 lots from source A yields a variance of 125
and a sample of 10 lots from source B yields a variance of 112. Is it
likely that the variance of source A significantly differs to the variance
of source B at significance level  = 0.01?
E4) A laptop computer maker uses battery packs of two brands, A and B.
While both brands have the same average battery life between charges
(LBC), the computer maker seems to receive more complaints about
shorter LBC than expected for battery packs of brand A. The computer
maker suspects that this could be caused by higher variance in LBC for
brand A. To check that, ten new battery packs from each brand are
selected, installed on the same models of laptops, and the laptops are
allowed to run until the battery packs are completely discharged. The
following are the observed LBCs in hours:
Brand A 3.2 3.7 3.1 3.3 2.5 2.2 3.2 3.1 3.2 4.3

Brand B 3.4 3.6 3.0 3.2 3.2 3.2 3.0 3.1 3.2 3.2

Assuming that the LBCs of both brands follows normal distribution,

test the LBCs of brand A have a larger variance that those of brand B
at 5% level of significance.
We now end this unit by giving a summary of what we have covered in it.

12.4 SUMMARY
In this unit, we have discussed the following points:
1. Testing of hypothesis for population variance using χ2-test.
2. Testing of hypothesis for two population variances using F-test.
104
 Chi-square and F-Tests
12.5 SOLUTIONS / ANSWERS
E1) Here, we are given that
σ0  15, n  20, S  17
Here, we want to test the agency’s claim that the standard deviation ()
of the length of serving times is less than 15 minutes. So our claim is
 < 15 and its complement is  ≥ 15. Since complement contains the
equality sign so we can take complement as null hypothesis and claim
as the alternative hypothesis. Thus,
H 0 :   0  15

H1 :   15 SD of the length of serving 

 times is less than 15 minutes 
Since the alternative hypothesis is left-tailed so the test is left-tailed
test.
Here, we want to test the hypothesis about the population standard
deviation and sample size is small n = 20(< 30). Also we are given that
the service time of the ambulance follows normal so we can go for χ2
test for population variance.
The test statistic is given by
 n  1  S2
χ2  2
~ χ 2 n 1 
σ
2
19   17.0 
  24.40
 15.0 2
The critical value of test statistic χ2 for left-tailed test corresponding
(n-1) = 19 df at 1% level significance is (2n 1), (1 )  2(19),0.99 = 7.63.
Since calculated value of test statistic (= 24.40) is greater than the
critical value (= 7.63), that means calculated value of test statistic lies in
non-rejection region so we do not reject the null hypothesis and reject
the alternative hypothesis i.e. we reject our claim at 1% level of
significance.
Thus, we conclude that sample provides us sufficient evidence against
the claim so agency’s claim that the standard deviation () of the length
of serving times is less than 15 minutes is not true.
E2) Here, we are given that
σ20  0.62, n  25, S2  0.65
Here, we want to test the cigarette manufacturer’s claims that the
variance (2) of nicotine content of its cigarettes is 0.62 milligrams. So
claim is 2 = 0.62 and its complement is  ≠ 0.62. Since claim contains
the equality sign so we can take claim as null hypothesis and
complement as the alternative hypothesis. Thus,
H 0 : σ 2  σ 02  0.62 and H1 : σ 2  0.62
Since the alternative hypothesis is two-tailed so the test is two-tailed
test.

105
Testing of Hypothesis Here, we want to test the hypothesis about the population variance and
sample size is small n = 25(< 30). Also we are given that the nicotine
content of its cigarettes follows normal distribution so we can go for χ2
test for population variance.
The test statistic is given by
 n  1  S2
χ2  ~ χ 2 n 1 
σ2
24  0.65
  25.16
0.62
The critical (tabulated) values of test statistic χ2 for two-tailed test
corresponding (n-1) = 24 df at 5% level of significance are
 2( n 1),  / 2   (224),0.025  39.36 and  2( n 1), (1 / 2)   (224), 0.975 =12.40.
Since calculated value of test statistic (= 25.16) is less than the critical
value (= 39.36) and greater than the critical value (= 12.40), that means
calculated value of test statistic lies in non-rejection region, so we do
not reject the null hypothesis i.e. we support the claim at 5% level of
significance.
Thus, we conclude that sample fails to provide sufficient evidence
against the claim so we may assume that manufacturer’s claim that the
variance of the nicotine content of the cigarettes is 0.62 milligram is
true.
E3) Here, we are given that
n1  12, S12  125, n 2  10, S22  112
Here, we want to test that variance of source A significantly differs to
the variances of source B. If 12 and 22 denote the variances in the raw
materials of sources A and B respectively so our claim is 12  22 and
its complement is 12  22 . Since complement contains the equality sign
so we can take the complement as the null hypothesis and the claim as
the alternative hypothesis. Thus,
H0 : 12  22 and H1 : 12  22
Since the alternative hypothesis is two-tailed so the test is two-tailed
test.
Here, we want to test the hypothesis about two population variances
and sample sizes n1 = 12(< 30) and n2 = 10 (< 30) are small. Also
populations under study are normal and both samples are independent
so we can go for F-test for two population variances.
For testing this, the test statistic is given by
S12
F ~ F n1 1,n 2 1 
S22
125
  1.11
112
The critical (tabulated) value of test statistic F for two-tailed test
corresponding (n1-1, n2 -1) = (11, 9) df at 5% level of significance are
F( n1 1, n 2 1),  / 2  F(11,9),0.025  3.91 and

106
 1 1 1 Chi-square and F-Tests
F( n1 1, n 2 1),(1 / 2)     0.28.
F(n2 1,n1 1),  / 2  F(9, 11),0.025 3.59
Since calculated value of test statistic (= 1.11) is less than the critical
value (= 3.91) and greater than the critical value (= 0.28), that means
calculated value of test statistic lies in non-rejection region, so we do
not reject the null hypothesis and reject the alternative hypothesis i.e.
we reject the claim at 5% level of significance.
Thus, we conclude that samples provide us sufficient evidence against
the claim so we may assume that the variances of source A and B is
differ.
E4) Here, we want to test that the LBCs of brand A have a larger variance
than those of brand B. If 12 and 22 denote the variances in the LBCs
of brands A and B respectively so our claim is 12  22 and its
complement is 12  22 . Since complement contains the equality sign
so we can take the complement as the null hypothesis and the claim as
the alternative hypothesis. Thus,
H0 : 12  22 and H1 : 12  22
Since the alternative hypothesis is right-tailed so the test is right-tailed
test.
Here, we want to test the hypothesis about two population variances
and sample sizes n1 = 10(< 30) and n2 = 10 (< 30) are small. Also
populations under study are normal and both samples are independent
so we can go for F-test for two population variances.
For testing the null hypothesis, test statistic is given by
S12
F ~ F n1 1, n 2 1  … (6)
S22
2
1 2 1    X 
where, S 
n1  1 
2
1  X  X   n  1   X  n  and
2

1  1 
2
1 2 1    Y  .
S22 
n2  1 
 Y  Y  n 1
 Y 2

n2 

2  
2 2
Calculation for S1 and S2 :
LBCs of Brand A X2 LBCs of Brand A Y2
(X) (Y)
3.7 13.69 3.6 12.96
3.2 10.24 3.2 10.24
3.3 10.89 3.2 10.24
3.1 9.61 3.0 9.00
2.5 6.25 3.0 9.00
2.2 4.84 3.2 10.24
3.1 9.61 3.2 10.24
3.2 10.24 3.1 9.61
4.3 18.49 3.2 10.24
3.2 10.24 3.1 9.61
Total = 3.18 104.1 31.8 101.38
107
Testing of Hypothesis
From the calculation, we have
1 1
X
n1  X   31.8  3.18,
10
1 1
Y
n2  Y   31.8  3.18
10

Thus,
2
1    X  1  31.8 2 
n1  1  
2 2
S   X    104.10 
1
n1  9  10 
 
1
  2.98  0.33
9
2
1    Y    1 101.38   31.8 2 
n2  1  
S22   Y 2
 
n 2  9  10 
 
1
  0.26  0.03
9
Putting the values of S12 and S22 in equation (6), we have
0.33
F  1.1
0.03
The critical (tabulated) value of test statistic F for right-tailed test
corresponding (n1-1, n2 -1) = (9, 9) df at 1% level of significance is
F( n1 1, n2 1),   F(9,9),0.01 = 5.35.

Since calculated value of test statistic (= 1.1) is less than the critical
value (= 5.35), that means calculated value of test statistic lies in non-
rejection region so we do not reject the null hypothesis and reject the
alternative hypothesis i.e. we reject the claim at 1% level of
significance.
Thus, we conclude that samples provide us sufficient evidence against
the claim so variance in LBCs of brand A is not greater than variance in
LBCs of brand B.

108
 MST-004
Indira Gandhi
STATISTICAL
National Open University
School of Sciences
INFERENCE

Block

4
NON-PARAMETRIC TESTS
UNIT 13
One-Sample Tests 5
UNIT 14
Two-Sample Tests 41
UNIT 15
k-Sample Tests 71
UNIT 16
Analysis of Frequencies 87
Appendix 109
Curriculum and Course Design Committee
Prof. K. R. Srivathasan Prof. Rahul Roy
Pro-Vice Chancellor Math. and Stat. Unit
IGNOU, New Delhi Indian Statistical Institute, New Delhi

Prof. Parvin Sinclair Dr. Diwakar Shukla

Pro-Vice Chancellor Department of Mathematics and Statistics
IGNOU, New Delhi Dr. Hari Singh Gaur University, Sagar

Prof. Geeta Kaicker Prof. Rakesh Srivastava

Director, School of Sciences Department of Statistics
IGNOU, New Delhi M.S. University of Baroda, Vadodara

Prof. Jagdish Prasad Prof. G. N. Singh

Department of Statistics Department of Applied Mathematics
University of Rajasthan, Jaipur I.S.M., Dhanbad

Prof. R. M. Pandey Dr. Gulshan Lal Taneja

Department of Bio-Statistics Department of Mathematics
All India Institute of Medical Sciences M.D. University, Rohtak
New Delhi
Faculty members of School of Sciences, IGNOU
Statistics Mathematics
Dr. Neha Garg Dr. Deepika
Dr. Nitin Gupta Prof. Poornima Mital
Mr. Rajesh Kaliraman Prof. Sujatha Varma
Dr. Manish Trivedi Dr. S. Venkataraman

Block Preparation Team

Prof. G. K. Shukla (Editor) Prof. Vyas Dubey (Units 13 & 14)
Decision Science Group Department of Statistics
Indian Institute of Management Pt. Ravishankar Shukla University, Raipur
Lucknow (UP)
Mr. Prabhat Kumar Sangal (Units 13 - 16)
Dr. Parmod Kumar (Language Editor) School of Sciences, IGNOU
School of Humanities, IGNOU

Course Coordinator: Mr. Prabhat Kumar Sangal

Programme Coordinator: Dr. Manish Trivedi

Block Production
Mr. Sunil Kumar, AR (P),School of Sciences, IGNOU
CRC prepared by Mr. Prabhat Kumar Sangal, School of Sciences, IGNOU

Acknowledgement: I gratefully acknowledge my colleague Mr. Rajesh Kaliraman , Statistics

Discipline, School of Sciences for their great support.

July, 2013
© Indira Gandhi National Open University, 2013
ISBN-978-81-266-
All rights reserved. No part of this work may be reproduced in any form, by mimeograph or any other
means, without permission in writing from the Indira Gandhi National Open University.

Further information on the Indira Gandhi National Open University may be obtained from University’s
Office at Maidan Garhi, New Delhi-110068 or visit University’s website http://www.ignou.ac.in

Printed and published on behalf of the Indira Gandhi National Open University, New Delhi by the
Director, School of Sciences.

Printed at:
NON-PARAMETRIC TESTS
In the previous block, we described the procedure for testing various
hypotheses involving population parameter(s) such as mean(s), proportion(s),
variance(s), etc. But in many real life problems particularly in Social and
Behavioural Sciences where the requirement of parametric tests cannot be
satisfied, that is, neither one can know the form of population nor the
observations measured in quantitative form. In such situations, the parametric
tests are not applicable. Thus, statisticians discovered various tests and
methods which are independent of population distribution and also applicable
when the observations are not measured in numerical scale that is in ordinal
scale or nominal scale. These tests are known as “Non-parametric tests” or
“Distribution Free Tests”.
Unit 13: One-Sample Tests
This unit explains the need of non-parametric tests and application of non-
parametric tests in various fields with their advantages and disadvantages over
parametric tests. In this unit, some of the frequently used non-parametric tests
for one sample such as sign test, Wilcoxon signed-rank test, run test and
Kolmogorov-Smirnov goodness of fit test are discussed.
Unit 14: Two-Sample Tests
This unit explores two samples tests such as paired sign test, Wilcoxon
matched-pair signed-rank test, Mann-Whitney U test and Kolmogorov-
Smirnov test.
Unit 15: k-Sample Tests
This unit provides the brief discussion on the procedures for testing for the
significance of differences among three or more populations. Kruskal -Wallis
test and Friedman test for k( > 2) samples are discussed.
Unit 16: Analysis of Frequencies
Last unit of this block is devoted to describe the procedure of chi-square tests
for categorical data such as goodness of fit test and test for independence of
two attributes.
Notations and Symbols
X1 , X 2 ,..., X n  : Random sample of size n 
H0 : Null hypothesis
H1 or HA : Alternative hypothesis
α : Size of critical region or type-I error or level of significance
 : Median

d : Absolute value of d

S and S : Number of plus and minus signs

T  and T  : Sum of positive and negative ranks

T : Critical value of Wilcoxon signed-rank test at α level of
significance
R : Number of runs
R L and R U : Lower and upper critical values of run test at α level of
significance
F(x) : Cumulative distribution function of the population
S(x) : Empirical or observed or sample cumulative distribution
function
sup d(x) : Supreme over all x, of the absolute value of d(x)
x

U L, and U U, : Lower and upper critical values of Mann-Whitney U test at
α level of significance
 2
 ,  : Critical value of χ2-test with ν degrees of freedom at α level
of significance
UNIT 13 ONE-SAMPLE TESTS
Structure
13.1 Introduction
Objectives
13.2 Advantages, Disadvantages and Types of Non-parametric Tests
13.3 Sign Test
13.4 Wilcoxon Signed-Rank Test
13.5 Run Test
13.6 Kolmogorov-Smirnov Goodness of Fit Test
13.7 Summary
13.8 Solutions /Answers

13.1 INTRODUCTION
In the previous block, we have described the procedure for testing various
hypotheses involving population parameter(s) such as mean(s), proportion(s),
variance(s), etc., together with a large number of examples. What we were
doing there can be summarised in following two points:
 We were making an assumption regarding the form of the distribution
function of the parent population from which the sample has been drawn is
known to us, and
A parametric test is a
 We were testing a statistical hypothesis regarding population parameters test which is based on
under study such as mean(s) or variance(s) or proportion(s), etc. the fundamental
assumption that the
Those types of tests where we deal above two points are known as form of parent
“parametric tests”. population is known
and involved
But in many real life problems particularly in social and behavioural sciences parameter(s).
the requirements of parametric tests do not meet. Therefore, in such situations,
the parametric tests are not applicable. Thus, statisticians discovered various A non-parametric test
is a test that does not
tests and methods which are independent of the form of population distribution make any assumption
and also applicable when the observations are not measured in interval scale regarding the form of
that is observations are measured in ordinal scale or nominal scale. These tests the parent population
are known as “Non-parametric tests” or “Distribution Free Tests”. The from which the sample
is drawn.
name distribution free test is suggested, since the distribution of the test
statistic is free (independent) from the form of population distribution. In
addition to it the non-parametric test is suggested, since these tests are
developed for such hypothesis testing which are not involving the population
parameter(s).
These tests are based on the “order statistics” theory which has the property
that “The distribution of the area under the density function between any two
ordered observations is independent of the form of the density function”.
Therefore, in these tests we use median, range, quartile, etc., in our hypothesis.
Assumptions Associated with Non-parametric Tests:
Although the non-parametric tests are not based on the assumption that the
form of parent population is known but these tests require some basic
assumptions which are listed below:
(i) Sample observations are independent.
5
Non-Parametric Tests (ii) The variable under study is continuous.
(iii) Lower order moments exist.
Obviously, these assumptions are fewer and weaker than those associated with
parametric tests.
This unit is divided into eight sections. Section 13.1 is described the need of
non-parametric tests. The advantages and disadvantages of the non-parametric
tests over parametric tests are described in Section 13.2. In Section 13.3 and
13.4, we discuss sign test and Wilcoxon signed-rank test for one-sample which
are generally used when assumption(s) of t-test is (are) not fulfilled. To test the
randomness of the data we use run test so in Section 13.5 we explore the run
test. In Section 13.6, we describe the Kolmogorov-Smirnov test for goodness
of fit. Unit ends by providing summary of what we have discussed in this unit
in Section 13.7 and solution of exercises in Section 13.8.
Objectives
After completion of this unit, you should be able to:
 describe the need of non-parametric tests;
 describe the applications of non-parametric tests in various fields with their
advantages and disadvantages over parametric tests;
 describe the one-sample non-parametric tests;
 describe an appropriate one-sample non-parametric test for different
situations;
 perform the test which is used in place of t-test for mean when
assumption(s) of the t-test is (are) not fulfilled, that is, sign test and
Wilcoxon signed-rank test;
 perform the test for testing the randomness of the data, that is, run test; and
 perform the Kolmogorov-Smirnov test for goodness of fit.

13.2 ADVANTAGES, DISADVANTAGES AND

TYPES OF NON-PARAMETRIC TESTS
In previous section, we have already said that non-parametric tests do the job
for us when assumptions of parametric tests do not meet. With this advantage
and some other advantages, non parametric test also have some disadvantages.
Both are listed in the following table:

Advantages Disadvantages

The non-parametric tests do When the assumptions of parametric tests are fulfilled
not make any assumption then parametric tests are more powerful than non-
regarding the form of the parametric tests. For example, Wilcoxon test has
parent population from which approximately 95% power compare to t-test. It means that
the sample is drawn. if null hypothesis is wrong then Wilcoxon test requires 5%
more observations in the sample compare to t-test to
achieve same power. As an example, Wilcoxon test
requires 20 observations compare to t-test which requires
19 observations to achieve same power.

Non-parametric tests have They do not use all the information available in terms of

6
 simple calculation compare to sample observations because they generally use ranks One-Sample Tests
parametric tests because they rather than actual values of the sample observations.
generally use ranks rather than
actual values of the sample
observations.

Because of using ranks they In non-parametric tests the required sample size reduces if
are also applicable even if data samples observations are tied as we will see during
is available in ordinal calculation.
measurement scale.

Now, you can answer the following exercise.

E1) Write any three advantages and disadvantages of non-parametric tests
compare to parametric tests?

Types of Non-parametric Tests

After discussing the advantages and disadvantages of the non-parametric tests,
now, you are interesting to know type of non-parametric tests. Broadly, the
non-parametric tests are classified into three categories as:
1. One-sample Tests
2. Two-sample Tests
3. k-sample Tests.

In this unit, we will discuss one-sample tests and other type of tests will be
discussed in subsequent units of this block.
Types of One-sample Tests
Some of the commonly used one-sample tests are listed below:
1. Sign Test
2. Wilcoxon Signed-Rank Test
3. Run Test
4. Kolmogorov-Smirnov Goodness of Fit Test
Let us discuss these tests one by one in subsequent sections.

13.3 SIGN TEST

The sign test is the simplest of the non-parametric tests. It is called the sign test The hypothesis
because this test is based on signs as plus and minus as you will see when we concerning the median
proceed through an example. The sign test is used as an alternative of the t-test cannot be considered as
for testing population median instead of population mean under the non-parametric testing
circumstances when the parent population is not normal. Further it also works problem in strict sense
if data is available in ordinal scale while for t-test we needed data at least in (median is itself a
interval scale. parameter) but it is taken
as a valid non-parametric
Assumptions testing problem since the
This test works under following assumptions: distribution of the test
statistic does not involve
(i) The sample is selected from the population with unknown median. the parent population
(ii) The variable under study is continuous. distribution.

(iii) The variable under study is measured on at least ordinal scale.

7
Non-Parametric Tests Let us discuss general procedure of this test:
Let X1 ,X 2 ,...,X n be a set of n random observations arranged in the order in
which they occur taken from the parent population having unknown median  .
Suppose, we wish to test the hypothesis about the specified value  0 of
population median  . So we can take the null and alternative hypotheses as

H 0 :    0 and H 1 :    0 for two-tailed test 

H 0 :    0 and H1 :    0 
or  for one-tailed test 
H 0 :    0 and H1 :    0 

After setting null and alternative hypotheses, sign test involves following steps:
Step 1: First of all, we convert the given observations into a sequence of plus
and minus signs. For this, we subtract the postulated value of median
( 0 ) from each observation, that is, we obtain the differences X i   0
for all observations and check their signs. Another way, we compare
each observation X i with  0 . If Xi is greater than  0 , that is, Xi   0 ,
then we replace the observation Xi by a plus sign and if Xi is less than
 0 , that is, Xi   0 , then we replace Xi by a minus sign. But when the
observation Xi which is equal to  0 give no information in terms of
plus or minus signs so we exclude all such observations from the
analysis part. Due to such observations our sample size reduces and
let reduced sample size be denoted by n.
Step 2: After that count the number of plus signs and number of minus signs.
Suppose they are denoted by S and S respectively.
Step 3: When null hypothesis is true and the population is dichotomised on
the basis of postulated value of median  0 then we expect that the
number of plus signs (success) and number of minus signs (failure)
approximately equal. And number of plus signs or number of minus
signs follows binomial distribution (n, p = 0.5). For convenient
consider the smaller number of plus or minus signs. If plus sign (S ) is
less than minus sign (S ) then we will take plus sign (S ) as success
and minus sign (S ) as failure. Similarly, if minus sign (S ) is less than
The critical values of this plus sign (S ) then we will assume minus sign (S ) as success and plus
test are not generally in
tabular form and slightly
sign (S ) as failure.
difficult to obtain whereas Step 4: To take the decision about the null hypothesis, we use concept of
p-value is easy to obtained p-value. We have discussed p-value in Unit 9 of this course. For
with the help of cumulative
p-value we determine the probability that test statistic is less than or
binomial table so we use
equal to the calculated value of test statistic i.e. actually observed plus
concept of p-value for take
or minus signs. Since distribution of number of plus or minus signs is
the decision about the null
hypothesis.
binomial (n, p = 0.5) therefore, this probability can be obtained with
the help of Table I given in Appendix at the end of this block which
provide the cumulative binomial probability and compare this
probability with the level of significance (α). Here, test statistic
depends upon the alternative hypothesis so the following cases arise:
8
For one-tailed test: One-Sample Tests

Case I: When H 0 :    0 and H1 :    0 (right-tailed test)

In this case, we expect that number of minus signs (S−) is smaller than
number of plus signs (S+) therefore, the test statistic(S) is the number
of minus signs (S−). The p-value is determined as
p-value = P S  S 
If p-value is less than or equal to α, that is, P S  S     then we
reject the null hypothesis at α level of significance and if the p-value
is greater than α then we do not reject the null hypothesis.
Case II: When H 0 :    0 and H1 :    0 (left-tailed test)
In this case, we expect that number of plus signs (S+) is smaller than
number of minus signs (S−) therefore, the test statistic(S) is the
number of plus signs (S+). The p-value is determined as
p-value = P S  S  Here, we take p-value as
 the probability less than or
If p-value is less than or equal to α, that is, P S  S    then we
equal to observed value of
reject the null hypothesis at α level of significance and if the p-value the test statistic in each
is greater than α then we do not reject the null hypothesis. case because in this test we
consider the smaller sign
For two-tailed test: When H 0 :    0 and H1 :    0 so critical region lies in
lower tail.
For two tailed test, the test statistic(S) is the smaller of number of plus
signs (S+) and minus signs (S−), that is,
S  min S ,S 

and approximate p-value is determined as

p-value = 2P S  S  ; if S is small

p-value = 2P S  S  ; if S  is small

If p-value is less than or equal to α, then we reject the null hypothesis
at α level of significance and if the p-value is greater than α then we
do not reject the null hypothesis.
For large sample  n  20  :
For a large sample size n greater than 20, we use normal
approximation to binomial distribution with mean
1 n
E(S)  np  n   … (1)
2 2
and variance
1 1 n
Var(S)  npq  n    … (2)
2 2 4
Therefore in this case, we use normal test (i.e. Z-test which is
described in Unit 10 of Block 3 of this course). The test statistic of
Z-test is given by
S  E  S
Z
SE  S
9
Non-Parametric Tests n
S
 2 ~ N  0,1 [Using equations (1) and (2)] … (3)
n
4
After that, we calculate the value of test statistic Z and compare it
with the critical value given in Table 10.1 of the Unit 10 of this
course at prefixed level of significance α. Take the decision about the
null hypothesis as described in the Section 10.2 of Unit 10 of this
course.
Note 1: Here, you may shock on the point that we used n > 20 for large sample
instead of n > 30, because binomial distribution with p = 0.5 approximated by
normal distribution less than 30.
Let us do some examples to become more user friendly with this test.
Example 1: The breaking strength (in pounds) of a random sample of 10 ropes
made by a manufacturer is given by
163 165 165 160 171 158 151 162 169 172
Use the sign test to test the manufacturer’s claim that the average breaking
strength of a rope is greater than 160 pounds at 5% level of significance.
Solution: Here, distribution of the population of the breaking strengths of the
ropes is not given. So the assumption of normality for t-test is not fulfilled.
Also sample size is small so we can not use Z-test. So we go for sign test.
Here, we want to test the manufacturer’s claim that the average (median)
breaking strength ( ) of a rope is greater than 160 pounds. So the claim is
  160 and its complement is   160. Since complement contains the
equality sign so we can take the complement as the null hypothesis and the
claim as the alternative hypothesis. Thus,
H 0 :    0  160 and H 1 :   160
Since the alternative hypothesis is right-tailed so the test is right-tailed test.
For applying sign test, we compare each observation with  0 ( 160) and
replacing each observation greater than 160 with a plus sign and each
observation less than 160 with a minus sign and discarding the one observation
which equals to 160, we get
+ + + + − − + + +
By counting, we have
S   number of plus signs  7
S   number of mimus si gns  2
n = total number of plus and minus signs = 9
Since alternative hypothesis is right-tailed so the test statistic (S) is the number
of minus signs (S−), that is,
S = number of minus signs (S−) = 2
Here, n = 9(< 20) so it is a case of small sample. Thus, to take the decision
about the null hypothesis, we determine p-value with the help of Table I given
in Appendix at the end of this block. Here, n = 9, p = 0.5 and r = 2. Thus, we
have
10
 p-value = P S  2 One-Sample Tests

 0.0899
Since p-value = 0.0899 > 0.05     . So we do not reject the null hypothesis
and reject the alternative hypothesis i.e. we reject the manufacturer’s claim at
5% level of significance.
Thus, we conclude that the sample provide us sufficient evidence against the
claim so manufacturer’s claim that the breaking strength of a rope is greater
than 160 pounds is not true.
Example 2: An economist believes that the median starting salary for a
computer programmer in a certain city is Rs 12,000. To verify this claim, 12
computer programmers with similar backgrounds who recently got the jobs
were randomly selected. Their starting salaries (in Rs) were 18000, 15000,
12000, 10000, 13000, 12000, 10000, 16000, 11000, 9000, 10000 and 9000. At
1% level of significance give your conclusion where you reached after using
sign test.
Solution: Here, distribution of the salaries of the computer programmers is not
given. So the assumption of normality for t-test is not fulfilled. Also sample
size is small so we can not use Z-test. So we go for sign test.
Here, we want to test that the median starting salary ( ) for a computer
programmer in a certain city is Rs 12000. So our claim is   12000 and it
complement is   12000. Since claim contains the equality sign so we can
take the claim as the null hypothesis and the complement as the alternative
hypothesis. Thus,
H 0 :    0  12000 and H1 :   12000
Since the alternative hypothesis is two-tailed so the test is two-tailed test.
For applying sign test, we replacing each value greater than  0 (  12000) with a
plus sign and each value less than 12000 with a minus sign and discarding the
two values which are equal to 12000, we get
+ + − + − + − − − −
By counting, we have
S   number of plussigns  4
S   number of minus signs = 6
n = total number of plus and minus signs = 10
Since alternative hypothesis is two-tailed so the test statistic(S) is the minimum
of number of plus signs (S+) and minus signs (S−), that is,
S  min S ,S   min 4, 6  4

Here, n =10 (< 20), so it is a case of small sample. Thus, to take the decision
about the null hypothesis, we determine p-value with the help of Table I given
in Appendix at the end of this block. Here, n = 10, p = 0.5 and r = 4. Thus, we
have
p-value  P S  4   0.3770

11
Non-Parametric Tests Since p-value = 0.37707 > 0.01(α) so we do not reject the null hypothesis i.e.
we support the claim at 1% level of significance.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so we may assume that median salary of the newly appointed
computer programmers is Rs 12000.
Example 3: The following data give the milk production (in thousand kg) in
full cream by 40 different dairies:
17 15 20 29 19 19 22 25 27 9
24 20 17 6 24 14 15 23 24 26
19 23 28 19 16 22 24 17 20 13
19 10 23 18 31 13 20 17 24 14
Use the sign test to test that median ( ) production of milk in dairies is 21.5
thousand kg at 1% level of significance.
Solution: Here, distribution of the milk production of the different dairies is
not given. So the assumption of normality for t-test is not satisfied. So we go
for sign test.
Here, we want to test that median production ( ) of milk in dairies is 21.5
thousand kg. So our claim is   21.5 and it complement is   21.5. Since
claim contains the equality sign so we can take the claim as the null hypothesis
and the complement as the alternative hypothesis. Thus,
H 0 :    0  21.5 and H 1 :   21.5
Since the alternative hypothesis is two-tailed so the test is two-tailed test.
For applying sign test, we replacing each value greater than 21.5 (median
value) with a plus sign and each value less than 21.5 with a minus sign, we get
− − − + − − + + + −
+ − − − + − − + + +
− + + − − + + − − −
− − + − + − − − + −
By counting, we have

S  number of plussigns  16

S   number of minus signs = 24

n = total number of plus and minus signs = 40
Since alternative hypothesis is two-tailed so the test statistic (S) is given by
S  min S ,S   min 16, 24  16

Here, n = 40 (> 20), so it is a case of large sample. In this case, we use Z-test.
The test statistic of Z-test is given by
n
S
Z 2 ~ N  0,1
n
4

12
 40 One-Sample Tests
16 
 2
40
4
16  20
  1.27
10
The critical (tabulated) values for two-tailed test at 1% level of significance are
± zα/2 = ± z0.005 = ± 2.58.
Since calculated value of Z (= −1.27) is greater than the critical value
(= − 2.58) and less than the critical value (= 2.58), that means it lies in
non-rejection region, so we do not reject the null hypothesis i.e. we support the
claim at 1% level of significance.
Decision according to p-value:
Since test is two-tailed, therefore,

p-value = 2P  Z  z   2P  Z  1.27

 2  0.5  P 0  Z  1.27 

 2  0.5  0.3980   0.204

Since p-value (= 0.204) is greater than ( 0.01) so we do not reject the null
hypothesis i.e. we support the claim at1% level of significance.

Thus, we conclude that the sample fails to provide us sufficient evidence

against the claim so we may assume that the median production of milk in
dairies is 21.5 thousand kg.
Now, you can try the following exercises.

E2) Give one difference between t-test and sign test for one-sample.
E3) A Metropolitan Area Road Transport Authority claims that the average
(median) waiting time on a well travelled rout is 20 minutes. A random
sample of 12 passengers showed weighting times of 22, 13, 17, 14, 25,
26, 19, 20, 22, 30, 10 and15 minutes. Test the Metropolitan Area Road
Transport Authority (MARTA) claim’s that average (median) waiting
time on a well traveled rout is 20 minutes by sign test at 1% level of
significance.
E4) The following data shows the weight (in kg) of a random sample of 30
cadets of a centre:
49 46 57 40 45 57 50 65 34 50
58 46 47 42 53 67 40 52 49 66
53 48 68 40 53 49 61 54 48 38

Use sign test to examine whether the median weight of all the cadets of
the centre is 50 kg at 5% level of significance.

13
Non-Parametric Tests
13.4 WILCOXON SIGNED-RANK TEST
In the previous section, we discussed sign test. We recall that the sign test
utilizes only information whether the difference between each observation and
the postulated value of the median is positive or negative, that is, it considers
only sign of differences. This test is fine if the information about the
observation sample is available in ordinal scale only. But if the measurement of
the observation is available interval or ratio scales then choice of sign test is
not recommended. Because this test do not take into account the information
available in terms of the magnitude of the differences. To overcome this
drawback of the sign test Wilcoxon signed-rank test do the job for us, which
takes into account the information of signs as well as of magnitude of
differences. Since Wilcoxon signed-rank test use more information than sign
test so it is more powerful than the sign test. The Wilcoxon signed-rank test is
also used as an alternative of the t-test.
Assumptions
Wilcoxon signed-rank test requires following assumptions to work:
(i) The sample is selected from the population with unknown median.
(ii) The sampled population is symmetric about its median.
(iii) The variable under study is continuous.
(iv) The variable under study is measured on at least interval scale.
Let us discuss general procedure of this test:
Let X1 , X2 ,..., Xn be a random sample of size n from a continuous symmetric
population. Let  be median of the population. Here, we wish to test the
hypothesis about the specified value  0 of population median  . So we can
take the null and alternative hypotheses as
H 0 :    0 and H1 :    0 for two-tailed test 
H 0 :    0 and H1 :    0 
or   for one-tailed test 
H 0 :    0 and H1 :    0 
After setting null and alternative hypotheses, Wilcoxon signed-rank test
involves following steps:
Step 1: We subtract  0 from each observation and obtain the difference di
with their plus and minus sign as
d i  Xi   0 for all observations
But when the observation Xi equal to  0 give no information in
terms of signs as well as magnitude so we exclude all such
observations from the analysis part. Due to such observations our
sample size reduces and let reduced sample size be denoted by n.
Step 2: After that, we find the absolute value of these di’s obtained in Step 1
as d 1 , d 2 ,..., d n .
Step 3: In this step, we are ranked di ’s (obtained in Step 2) with respect to
their magnitudes from smallest to largest, that is, the rank 1 is given
to the smallest of di ’s, rank 2 is given to the second smallest and so
14
 on up to the largest di ’s. If several values are same (tied), we assign One-Sample Tests
each the average of ranks they would have received if there were no For more detailed about
repeated (tied) ranks
repetition. please go through
Step 4: Now assign the signs to the ranks which the original differences have. Section 7.4 of the Unit
2 of MST-002.
Step 5: Finally, we calculate the sum of the positive ranks (T+) and sum of
negative ranks (T−) separately.
Under H0, we expect approximately equal number of positive and
negative ranks. And under the assumption that population under study
is symmetric about its median we expect that sum of the positive
ranks (T+) and sum of negative ranks (T−) are equal.
Step 6: Decision Rule:
To take the decision about the null hypothesis, the test statistic is the
smaller of T+ and T−. And the test statistic is compared with the
critical (tabulated) value for a given level of significance (α) under the
condition that the null hypothesis is true. Table II given in Appendix
at the end of this block provides the critical values of test statistic at α
level of significance for both one-tailed and two-tailed tests. Here, test
statistic depends upon the alternative hypothesis so the following
cases arise:
For one-tailed test:
Case I: When H0 :    0 and H1 :    0 (right-tailed test)
In this case, we expect that sum of negative ranks (T−) is smaller than
sum of positive ranks (T+) therefore, the test statistic (T) is the sum of Here, we consider that if
negative ranks (T−). test statistic less than or
equal to critical value of
If computed value of test statistic (T) is less than or equal to the the test statistic then we
critical value T at α level of significance, that is, T  T then we reject null hypothesis in
each case because in this
reject the null hypothesis at α level of significance, otherwise we do
test we consider the
not reject the null hypothesis. smaller sum so critical
Case II: When H0 :    0 and H1 :    0 (left-tailed test) region lies in lower tail.

In this case, we expect that sum of positive ranks (T+) is smaller than
sum of negative ranks (T−) therefore, the test statistic (T) is the sum of
positive ranks (T+).
If computed value of test statistic (T) is less than or equal to the
critical value T at α level of significance, that is, T  T then we
reject the null hypothesis at α level of significance, otherwise we do
not reject the null hypothesis.
For two-tailed test:
When H0 :    0 and H1 :    0
In this case, the test statistic (T) is the smaller of sum of positive
ranks (T+) and sum of negative ranks (T−), that is,
T  min T ,T 
If computed value of test statistic (T) is less than or equal to the
critical value Tα/2 at α level of significance , that is, T  T / 2 then we
reject the null hypothesis at α level of significance, otherwise we do
not reject the null hypothesis.
15
Non-Parametric Tests For large sample (n > 25):
For a large sample size n greater than 25, the distribution of test
statistic (T) approximated by a normal distribution with mean
n ( n  1)
E T   … (4)
4
and variance
n n  12n  1
Var T   … (5)
24
The proof of mean and variance of T is beyond the scope of this
course.
Therefore in this case, we use normal test (Z-test). The test statistic of
Z-test is given by
T  ET  T  E T 
Z 
SE  T  Var  T 

n  n  1
T
 4 ~ N  0,1  Using equations  … (6)
n  n  1 2n  1  (4) and (5) 
24
After that, we calculate the value of test statistic Z and compare it
with the critical value given in Table 10.1 at prefixed level of
significance α. Take the decision about the null hypothesis as
described in the Section 10.2 of Unit 10 of this course.
Now, it is time to do some examples based on above test.
Example 4: A random sample of 15 children of one month or older shows the
following pulse rates (beats per minute):
119, 120, 125, 122, 118, 117, 126, 114, 115, 126, 121, 120, 124, 127,
126
Assuming that the distribution of pulse rate is symmetric about its median and
continuous, is there evidence to suggest that the median pulse rate of one
month or older children is 120 beats per minute at 5% level of significance?
Solution: Here, distribution of pulse rate is not given. So the assumption of
normality for t-test is not fulfilled although all the assumptions of Wilcoxon
signed-rank test hold. So we go for this test.
Here, we want to test that median pulse rate ( ) of children of one month or
older is 120 beats per minute. So our claim is   120 and it complement is
  120. Since claim contains the equality sign so we can take the claim as the
null hypothesis and the complement as the alternative hypothesis. Thus,
H 0 :    0  120  median pluse rate is equal to 120

H1 :   120  median pluse rate is not equal to 120

Since the alternative hypothesis is two-tailed so the test is two-tailed and the
test statistic is the smaller of sum of positive ranks (T+) and sum of negative
ranks (T−), that is,
T  min T ,T 

16
 Calculation for T: One-Sample Tests
S. Beats per Difference Absolute Value Rank of Signed
No. Minute d  X   of Difference d Rank
(X)  X  120 d
1 119 −1 1 1.5 −1.5
2 120 Tie --- --- ---
3 125 5 5 7.5 7.5
4 122 2 2 3.5 3.5
5 118 −2 2 3.5 −3.5
6 117 −3 3 5 −5
7 126 6 6 10.5 10.5
8 114 −6 6 10.5 −10.5
9 115 −5 5 7.5 −7.5
10 126 6 6 10.5 10.5
11 121 1 1 1.5 1.5
12 120 Tie --- --- ---
13 124 4 4 6 6
14 127 7 7 13 13
15 126 6 6 10.5 10.5

From the above calculations, we have

T+ = 7.5 + 3.5 + 10.5 + 10.5 + 1.5 + 6 + 13 + 10.5 = 63.0
T− = 1.5 + 3.5 + 5 + 10.5 + 7.5 = 28.0
n = number of non-zero di’s =13
Putting the values in test statistic, we have
T  min T ,T  min 63.0,28.0  28.0
The critical (tabulated) value for two-tailed test corresponding n = 13 at 5%
level of significance is 18.
Since calculated value of test statistic T (= 28.0) is greater than the critical
value (= 18) so we do not reject the null hypothesis i.e. we support the claim at
5% level of significance.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so we may assume that median pulse rate of children of one
month or older is 120 beats per minute.
Example 5: The following data show the weight (in kg) of a random sample
of 30 cadets of a college:
49 46 57 37 45 57 50 65 34 50 58
46 47 42 53 67 40 52 49 66 53 48
68 40 53 49 61 54 48 38
Assume that distribution of the weight of the cadets is symmetric about its
median, test to examine whether the median weight of all the cadets of the
college is 50 kg by useing Wilcoxon signed-rank at 5% level of significance.
Solution: Here, distribution of weight of the cadets is not given. So the
assumption of normality for t-test is not fulfilled. We are given to us that
distribution of the weight of the cadets is symmetric about its median and
assumption of continuity holds because the characteristic weight is continuous
in nature. So we go for Wilcoxon signed-rank test.

17
Non-Parametric Tests Here, we want to test that median weight ( ) of all the cadets of a college is 50
kg. So our claim is   50 and it complement is   50. Since claim contains
the equality sign so we can take the claim as the null hypothesis and the
complement as the alternative hypothesis. Thus,
H 0 :    0  50  median weight is equal to 50kg

H1 :   50  median weight is not equal to 50kg

Since the alternative hypothesis is two-tailed so the test is two-tailed test and
the test statistic is the smaller of sum of positive ranks (T+) and sum of negative
ranks (T−), that is,
T  min T ,T 
Calculation for T:
S. Weight Difference Absolute Value Rank of Signed
No. (in kg) d  X   of Difference d Rank
(X)  X  50 d
1 49 −1 1 2 −2
2 46 −4 4 12 −12
3 57 7 7 15.5 15.5
4 37 −13 13 23 −23
5 45 −5 5 14 −14
6 57 7 7 15.5 15.5
7 50 Tie --- --- ---
8 65 15 15 24 24
9 34 −16 16 25.5 −25.5
10 50 Tie --- --- ---
11 58 8 8 17.5 17.5
12 46 −4 4 12 −12
13 47 −3 3 8.5 −8.5
14 42 −8 8 17.5 −17.5
15 53 3 3 8.5 8.5
16 67 17 17 27 27
17 40 −10 10 19.5 −19.5
18 52 2 2 5 5
19 49 −1 1 2 −2
20 66 16 16 25.5 25.5
21 53 3 3 8.5 8.5
22 48 −2 2 5 −5
23 68 18 18 28 28
24 40 −10 10 19.5 −19.5
25 53 3 3 8.5 8.5
26 49 −1 1 2 −2
27 61 11 11 21 21
28 54 4 4 12 12
29 48 −2 2 5 −5
30 38 −12 12 22 −22

From the above calculations, we have

T+ = 15.5 + 15.5 + 24 + 17.5 + 8.5 + 27 + 5 + 25.5 + 8.5 + 28 +8.5 + 21+ 12
= 216.5
18
T− = 2 + 12 + 23 + 14 + 25.5 + 12 + 8.5 + 17.5 + 19.5 + 2 + 5 + 19.5 + 2 + 5 + 22 One-Sample Tests

= 189.5
n = number of non-zero di’s = 28
Putting the values in test statistic, we have
T  min T ,T   min 216.5,189.5  189.5
Also n = 28 (> 25) therefore, it is the case of large sample. So in this case, we
use Z-test. The test statistic of Z-test is given by
T  E T 
Z ~ N  0,1 
SE  T 
n(n  1) 28(28  1)
where, E  T     203 and
4 4
n  n  1   2n  1  28  28  1   2  28  1 
SE  T     43.91
24 24
Putting the values in test statistic Z, we have
189.5  203
Z   0.31
43.91
The critical (tabulated) values for two-tailed test at 5% level of significance are
± zα/2 = ± z0.025 = ± 1.96.
Since calculated value of Z (= −0.31) is greater than the critical value
(= − 1.96) and less than the critical value (= 1.96), that means it lies in non-
rejection region, so we do not reject the null hypothesis i.e. we support the
claim.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so we may assume that the median weight of all the cadets of
the college is 50 kg.
Now, you can try the following exercises in the same manner.
E5) Give one main difference between sign test and Wilcoxon signed-rank
test.
E6) The observations of a random sample of size 10 from a distribution
which is continuous and symmetric about its median are given below:
20.2, 24.1, 21.3, 17.2, 19.8, 16.5, 21.8, 18.7, 17.1, 19.9
Use Wilcoxon test to test the hypothesis that the sample is taken from a
population having median greater than 18 at 5% level of significance.
E7) The breaking strength (in pounds) of a random sample of 29 rods made
by a manufacturer is given as follows:
19 30 28 35 23 37 36 32 40 24
31 33 21 36 30 32 26 40 38 28
30 41 22 43 25 28 24 37 17
On the basis of this sample test the manufacturer’s claim that the
average (median) breaking strength of the rope is 30 pounds by
Wilcoxon singed-rank test at 1% level of significance. Assume that
breaking strength of the rods is symmetric about its median.

19
 Non-Parametric Tests
13.5 RUN TEST
One of the fundamental assumptions of the parametric test is that the observed
data are random and test statistic and the subsequent analysis are based on this
assumption. It is always better to check whether this assumption is true or not.
A run is a succession of
identical letters or A very simple tool for checking this assumption is run test. This section is
symbols that is followed devoted to throw light on the run test. Before discussing the run test first we
and preceded by a have to explain what we mean by a “run”.
different letter or by no
letter at all. A run in observations, is defined as a sequence of letters or symbols of one
kind, immediately preceded and succeeded by letters of other kind or no letters.
For example a sequence of two letters H and T as given below:
HHTHTTTHTHHHTTT
In this sequence, we start with first letter H and go up to other kind of letter,
that is, T. In this way, we get first run of two H’s. Then we start with this T and
go up to other kind of letter, that is, H. Then we get a run of one T and so on
and finally a run of three T’s. In all, we see that there are eight runs as shown
below:
HH
 TH
 TTT
 HT
 HHH
 TTT
1 2 3 4 5 6 7 8

Under run test, randomness of observations is judged by number of runs in the

observed sequence. Too few runs indicate that there is some clustering or trend
and too large runs indicate that there is some kind of repeated or cycles
according to some pattern.
For example, the following sequence of H’s and T’s is obtained when tossing a
coin 20 times:
HHHHHHHHHH
  TTTTTTTTTT

1 2

In this sequence, we have 2 runs-a run of 10 heads followed by a run of 10

tails. Since similar items tend to cluster together, therefore, such a sequence
could not be considered random even though as would theoretically be
expected 10 heads and 10 tails out of 20 tosses. In another example, suppose
the following sequence has been obtained:
H
TH
TH
TH
TH
TH
TH
TH
TH
TH
T
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

In this sequence there are 20 runs-10 runs of one head each and 10 runs of one
tail each. This sequence also could not be considered random because too many
runs indicate a pattern.
On the other hand, if the sequence of H’s and T’s is thoroughly mixed, such as,
H
 TTT
HH TTT H T
H TT
HHH T HH

1 2 3 4 5 6 7 8 9 10 11

where neither the number of runs too small nor too large, this type of sequence
may be considered random. Therefore, in testing for randomness, ordering or
positioning of the items in the sequence is essential, not the frequency of the
items of each type.
Assumptions
Run test make the following assumptions:
(i) Observed data should be such that we can categorise the observations into
two mutually exclusive types.
20
(ii) The variable under study is continuous. One-Sample Tests

Let X1 ,X2 ,...,Xn be a set of n observations arranged in the order in which

they occur. Generally, we are interested to test whether a population or a
sample or a sequence is random or not. So here we consider only two-tailed
case. Thus, we can take the null and alternative hypotheses as
H0 : The observations are random
H1 : The observations are not random [two-tailed test]
Let us discuss the general procedure of this test:
Step 1: First of all, we check the form of the given dada that the given data
are in symbolical form such as sequence of H and T, A and B, etc. or
in the numeric form. If the data in symbolical form then it is ok, but if
data in numeric form then first we convert numeric data in symbolical
form. For this, we calculate median of the given observations by
using either of the following formula (discussed in Unit 1 of the
course MST-002.) given below:
th
 n  1 
Median    observation; if n' is odd
 2 
th th
1  n '   n'  
Median    observation    1 observation  ; if n' is even
2  2  2  
provided observations should be either in ascending or descending
order of magnitude.
After that, we replace the observations which are above the median by
a symbol ‘A’ (say) and the observation which are below the median
by a symbol ‘B’ (say) without altering the observed order. The
observations which are equal to median are discarded form the
analysis and let reduced size of the sample denoted by n.
Step 2: Counts number of times first symbol (A) occurs and denote it by n1.
Step 3: Counts number of times second symbol (B) occurs and denote it by
n 2 where, n1  n2  n.
Step 4: For testing the null hypothesis, the test statistic is the total number of
runs so in this step we count total number of runs in the sequence of
symbols and denote it by R.
Step 5: Obtain critical values of test statistic corresponding n1 ,n 2 at α % level
of significance under the condition that null hypothesis is true. Part I
and Part II of Table III in the Appendix respectively provide lower
(RL) and upper (RU) critical values of the number of runs for a given
combination of n1 and n2 at 5% level of significance.
Note1: Generally, critical values for run test are available at 5% level of
significance so we test our hypotheses for 5% level of significance.
Step 6: Decision Rule:
To take the decision about null hypothesis, the test statistic is
compared with the critical (tabulated) values.

21
Non-Parametric Tests If the observed number of runs(R) is either less than or equal to the
lower critical value (RL) or greater than or equal to the upper critical
value (RU), that is, if R  R Lor R  R U then we reject the null
hypothesis at 5% level of significance.
If R lies between R L and R U , that is, R L  R  R U , then we do not
reject the null hypothesis at 5% level of significance.
For large sample (n > 40):
For a sample size n greater than 40 or either n1 or n 2 exceeds 20, the
statistic R is approximately normally distributed with mean
2n1n 2
E(R)  1 ... (7)
n
and variance
2n1n1  2n1n 2  n 
Var  R   … (8)
n 2  n  1
The proof of mean and variance of R is beyond the scope of this
course.
Therefore in this case, we use normal test (Z-test). The test statistic of
Z-test is given by
R  E R
Z ~ N  0,1
Var  R 

2 n 1n 2
R 1
Z n [Using equations (7) and (8)] ... (9)
2 n 1n 2 2 n 1n 2  n 
n 2 n  1

After that, we calculate the value of test statistic Z and compare it

with the critical value given in Table 10.1 at prefixed level of
significance α. Take the decision about the null hypothesis as
described in the Section 10.2 of Unit 10 of this course.
Now, it is time to do some examples based on above test.
Example 6: A sequence of Heads (H) and Tails (T) in tossing of a coin 16
times is given below:
HTTTHTHTHHTHTTHH
(i) Count the number of runs.
(ii) Test whether the Heads and Tails occur in random order.
Solution: Here, we are given a sequence of Heads of Tails as
HTTTHTHTHHTHTTHH
(i) For counting the number of runs, we start with first letter H and go up to
other kind of letter, that is, T. In this way, we get first run of one H. Then
we start with this T and go up to other kind of letter, that is, H then we
get a run of three T’s and so on in the last there is a run of two H’s. In this
way, the total number of runs is 11 as shown below:
H
 TTT
 HT
HT
 HH
 TH
 TT
 HH
1 2 3 4 5 6 7 8 9 10 11

22
(ii) Here, we want to test the randomness of the sequence so we go for run One-Sample Tests
test.

We want to test that the given sequence of H and T is in random order.

So our claim is “the given sequence of H and T is in random order” and
its complement is “the given sequence of H and T is not in random
order”. So we can take claim as the null hypothesis and complement as
the alternative hypothesis. Thus,
H0 : Given sequence of Heads (H) and Tails (T) is in random order
H1 : Given sequence of Heads (H) and Tails (T) is not in random order
Here by counting, we have
n1 = number of times H occurs = 8
n 2 = number of times T occurs = 8
R = total number of runs in given sequence = 11
For run test, the test statistic is the total number of runs in the given
sequence.
Since n1 = 8 (< 20) and n2 = 8 (< 20), so it is a small sample case.
The lower and upper critical values of runs corresponding to n1 ( 8)and
n 2 (  8) at 5% level of significance are R L  4 and R U  14.
Since observed number of runs (R = 11) lies between critical values of
runs, R L  4 and R U  14, so we do not reject the null hypothesis i.e. we
support the claim at 5% level of significance.
Thus, we conclude that the given sequence of H and T is in random order.
Example 7: A machine produced a number of defective items per day. To
check the randomness of the defective items, a sample of defective items
produced per day over a period of 20 days is taken and found following results:
13 17 14 20 18 17 16 14 19 21
18 20 17 13 14 19 20 15 19 17
Test at 5% level of significance that the machine produced defective items in
random order.
Solution: Here, we want to test the randomness of the defective items
produced per day so we go for run test.
We want to test that machine produced defective items in random order. So our
claim is “the machine produced defective items are in random order” and its
complement is “the machine produced defective items are not in random
order”. So we can take claim as the null hypothesis and complement as the
alternative hypothesis. Thus,
H0 : The machine produced defective items are in random order
H1 : The machine produced defective items are not in random order
For applying run test, we need given observations in a sequence of two
symbols. In order to convert theses observations into a sequence of two
symbols first of all, we have to calculate median of the given observations. For
23
Non-Parametric Tests obtaining median, we have to arrange the given observations in ascending or
descending order of magnitude. So arrange the observations in ascending order
of magnitude, we have
13 13 14 14 14 15 16 17 17 17 17 18 18 19 19 19 20
20 20 21
Number of observations (n)  20 (even) therefore the median can be obtained
as
th th
1  n    n  
Median    observation    1 observation 
2  2  2  
th th
1  20   20  
   observation    1  observation 
2  2   2  

1
 (10) th observation  (11) th observation 
2
1
 17  17   17
2
Now, we replace the given observations by the symbol ‘A’ if it is above the
median value (= 17) and by symbol ‘B’ if it is below the median value and
discard the observation equal to median value. In this way, the sequence
obtained by discarding 4 observations which are equal to median value 17 is
given below:
BBAABBAAAABBAABA
For applying run test, we count the number of runs. We see that there is a first
run of two B’s then there is a run of two A’s and so on in the last there is a run
of single A as shown below:
BBAA
 BBAAAA

 BBAA
 BA

1 2 3 4 5 6 7 8

By counting, we have
n1 = number of symbol A occurs = 9

n 2 = number of symbol B occurs = 7

R = total number of runs in above sequence = 8
For run test, the test statistic is the total number of runs in the given sequence.
Since n1 = 9(< 20) and n2 = 7 (< 20) so it is a small sample case.
The lower and upper critical values of runs corresponding to n1  9 and n 2  7
and at 5% level of significance are R L  4 and R U  14.

Since observed number of runs (R = 9) lies between critical values of runs,

R L  4, and R U  14 so we do not reject the null hypothesis i.e. we support the
claim at 5% level of significance.
Thus, we conclude that the sample fails to provide us sufficient evidence

24
against the claim so we may assume that the machine produced defective items One-Sample Tests
are in random order.
Example 8: A sample of 36 tools produced by a machine shows the following
sequence of good (G) and bad (B) tools:
GGGBGGGGGBBBGGGGGGBBGGGGGGGGGBBGGGGG
Test the machine produced good (G) and bad (B) tools in random order at 5%
level of significance.
Solution: Here, we want to test the randomness of the sequence so we go for
run test.
We want to test that machine produced good (G) and bad (B) tools in random
order. So our claim is “the machine produced good and bad tools are in random
order” and its complement is “the machine produced good and bad tools are not
in random order”. So we can take claim as the null hypothesis and complement
as the alternative hypothesis. Thus,
H0 : The machine produced good and bad tools are in random order
H1 : The machine produced good and bad tools are not in random order
For applying run test, we count the number of runs.
We see that there is a first run of three G’s then there is a run of single B and so
on in the last there is a run of five G’s as shown below:
GGG
 BGGGGG
 BBBGGGGGG
  BBGGGGGGGGG
  BBGGGGG
 
1 2 3 4 5 6 7 8 9

By counting, we have
n1 = number of symbol A occurs = 28
n 2 = number of symbol B occurs = 8
R = total number of runs in above sequence = 9
Since n1 = 28 (>20) therefore, it is the case of large sample so we use Z-test.
The test statistic of Z-test is given by
R  E R
Z ~ N  0,1
SE  R 

2n1n 2 2  28  8
where, E(R)  1  1  13.44 and
n 36

2n1n 2 (2n1n 2  n) 2  28  8  2  28  8  36 
SE(R)  2

n (n  1) 362  36  1

 4.07  2.02

Putting the values in test statistic Z, we have

9  13.44
Z  2.20
2.02
The critical (tabulated) values for two-tailed test at 5% level of significance are
± zα/2 = ± z0.005 = ± 1.96.

25
Non-Parametric Tests Since calculated value of Z (= −2.20) is less than critical values (= ± 1.96), that
means it lies in rejection region so we reject the null hypothesis i.e. we reject
the claim at 5% level of significance.
Thus, we conclude that the sample provides us sufficient evidence against the
claim so the machine produced good and bad tools are not in random order.
Now, you can try the following exercises.
E8) The following observations were taken from a table of random numbers
from a distribution F(x) with median 0:
0.464 0.137 2.455 – 0.323 – 0.068
0.907 –0.513 – 0.525 0.595 0.886
– 0.482 1.678 – 0.057 –1.229 – 0.486
–1.787 –0.261 1.237 1.046 0.962
Test whether the data are random at 5% level of significance?
E9) The cars that enter in a parking are classified either Foreign-made (F) or
Indian-made (I). To check that the car entre in the parking are in
random order, a sample of first 40 cars that entered in the parking is
taken and found the following sequence:
IIFFFIFIFFFFIIFIIIFFIFFIFFFFIIFFFIFFIIII
Test that the cars enter in the parking are in random order at 1% level of
significance.

13.6 KOLMOGOROV-SMIRNOV GOODNESS OF

FIT TEST
This test has its name on the names of its discovers A. N. Kolmogorov and
N.V. Smirnov. It is a simple non-parametric test for testing whether data follow
a specified or assumed distribution or sample has come from a specified or
assumed distribution or there is a significant difference between an observed
distribution and a theoretical distribution. Therefore, it is a measure of
goodness of fit to a theoretical distribution. We also use chi-square test for
goodness of fit (will be described in Unit 16 of this block). The main difference
between chi-square test and Kolmogorov-Smirnov (K-S) test is that the chi-
square test is designed for categorical data whereas K-S test is designed for the
continuous data.
Assumptions
This test works under the following assumptions:
(i) The sample is randomly selected from some unknown distribution.
(ii) The observations are independent.
(iii) The variable under study is continuous.
(iv) The variable under study is measured on at least ordinal scale.
Let X1, X2,..., Xn be a random sample from a population with unknown
continuous distribution function F(x). Generally, we are interested to test
whether data follow a specified distribution F0(x) or a sample has come from a
specified or assumed distribution or not. So here we consider only two-tailed
case. Thus, we can take the null and alternative hypotheses as
26
 H0 : Data follow a specified distribution One-Sample Tests

H1 : Data do not follow a specified distribution

In symbolical form
H0 : F(x)  F0 (x) for all values of x

H1 : F(x)  F0 (x) for at least one value of x  two-tailed test 

We summarise the Kolmogorov-Smirnov goodness of fit test in the following
steps:
Step 1: K-S goodness of fit test is based on the comparison of empirical
(observed or sample) cumulative distribution function with the
theoretical (specified or population) cumulative distribution function.
Therefore first of all, we compute the empirical cumulative
distribution function, say, S(x) from the sample data which is defined
as the proportion of sample observations less than or equal to some
value x, that is,
S(x) = proportion of sample observations less than or equal to some
value x
The number of sample observation less than or equal to x
S(x) 
Total number of observations
Step 2: In this Step, we find the theoretical cumulative distribution function
F0(x) for all possible values of x on the basis of specified distribution.
Step 3: After finding the empirical and theoretical cumulative distribution
functions for all possible values of x, we find the deviation between
the empirical and theoretical cumulative distribution functions for all
x. That is,
S(x)  F0 (x) for all x
Step 4: Since S(x) is the statistical image of the population distribution
function F(x) so if null hypothesis is true then the difference between
S(x) and F(x) should be small for all values of x. Thus, in all
deviations obtained in Step 3, we find the point(s) at which the two
functions show the maximum deviation.
The test statistic is denoted by Dn and it is the greatest vertical
deviation between S(x) and F0(x), that is,
D n  sup S(x)  F0 (x)
x

which is read as “Dn equals the supreme over all x, of the absolute
value of the difference S(x)  F0 (x) ”

Step 5: Obtain critical value of test statistic at α % level of significance under

the condition that null hypothesis is true. Table IV in Appendix
provides the critical values at α level of significance.
Step 6: Decision Rule:
To take the decision about the null hypothesis, the test statistic
(calculated in Step 4) is compared with the critical (tabulated) value
(obtained in Step 5) for a given level of significance (α).
27
Non-Parametric Tests If computed value of Dn is greater than or equal to critical value Dn,
at α level of significance, that is, if Dn  Dn, then we reject the null
hypothesis at α level of significance, otherwise we do not reject H0.
For large sample (n > 40):
For a sample size n greater than 40, the critical value of test statistic at
given α level of significance, is approximated by the formula given in
last row of Table IV in the Appendix. As the critical value for two-
tailed test at 5% level of significance is calculated by the formula
given by
1.36
D n , 0.05 
n
where, n is the sample size.
Let us do some examples to become more user friendly with the calculations
explained above:
Example 9: A coin is tossed 400 times in sets of four. The frequencies of
getting 0 to 4 tails are:
Number of Tails 0 1 2 3 4 Total
Frequency 23 95 164 92 26 400

Examine the fairness of coin using K-S test at 5% level of significance.

Solution: Here, we want to test that the coin is fair. We know that if a coin is
fair then number of tails follows the binomial distribution with parameter n
and p = 1/2. So we can test that the number of tails follows the binomial
distribution. If F0(x) denotes the cumulative distribution function of binomial
distribution with parameter n = 4 and p = 1/2 then our claim is F(x) = F0(x) and
its complement is F(x) ≠ F0(x). Since the claim contains the equality sign so we
can take claim as the null hypothesis and complement as the alternative
hypothesis. Thus
H0 : F(x)  F0 (x) for all values of x
H1 : F(x)  F0 (x) for at least one value of x
For testing the null hypothesis, the test statistic is given by
D n  sup S  x   F0  x 
x
where, S(x) is the empirical distribution function based on the observed
sample.
For obtaining the empirical cumulative distribution function S(x), first we
obtain cumulative frequencies and then by the definition of empirical
distribution function we obtain S(x) as
The number of sample observations  x
S x  
Total number of observations
Therefore, at x = 0, 1, 2, 3, 4, we have
23 118 282
S 0    0.0575, S  1    0.2950, S  2    0.705,
400 400 400
374 400
S 3   0.935, S  4   1
400 400
28
The theoretical distribution can be obtained by the definition of cumulative One-Sample Tests
distribution function as
F0 (x)  P  X  x 
In this case, our theoretical distribution is binomial distribution with parameter
n = 4 and p = 1/2. Therefore,
x 4 x
1 1
i.e. F0 (x)  P  X  x   Cx    
4

2 2
4
1
 4 C x   , x  0,1, 2,3,4
2
With the help of Table I given in Appendix at the end of this block we can find
the values of F0 (x) at different values of x.

Thus, at x = 0, we have
F0 (0)  P  X  0   0.0625
Similarly,
At x  1, F0 (1)  P  X 1   0.3125

At x  2, F0 (2)  P  X  2   0.6875
At x  3, F0 (3)  P  X  3   0.9375
At x  4, F0 (4)  P  X  4   1
Calculation for S  x   F0  x  :

Number Observed Cumulative Observed Theoretical Absolute

of Tails Frequency Observed Distribution Distribution Difference
(x) Frequency Function Function S  x   F0  x 
S(x) = (3)/n
(1) (2) (3) (4) (5) (6)
0 23 23 0.0575 0.0625 0.0050
1 95 118 0.2950 0.3125 0.0175
2 164 282 0.7050 0.6875 0.0175
3 92 374 0.9350 0.9375 0.0025
4 26 400 1 1 0

From the above calculation, we have

D n  sup S  x   F0  x   0.0175
0 x  4

Here, n = 400 > 40, therefore, critical value of test statistic can be obtained by
the formula given by
1.36
Dn,0.05 
n
1.36
  0.068
400

29
Non-Parametric Tests Since calculated value of test statistic Dn (= 0.0175) is less than critical value
Dn,0.05 (= 0.068) so we do not reject the null hypothesis i.e. we support the
claim at 5% level of significance.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so we may assume that number of tails follows the binomial
distribution or the coin is fair or unbiased.
Example 10: The following data were obtained from a table of random
The Kolmogorov-
Smirnov test for
numbers of normal distribution with mean 3 and standard deviation 1:
goodness of fit is
2.1, 1.9, 3.2, 2.8, 1.0, 5.1, 4.2, 3.6, 3.9, 2.7
applied only when all
the parameters of the Test by K-S test that the data have come form the same normal distribution at
fitting distribution are 5% level of significance.
known. If all parameter
(s) of the fitted Solution: Here, we want to test that the data have come from normal
distribution is (are) not distribution with mean 3 and variance 1. If F0(x) is the cumulative distribution
known then we use chi- function of N(3,1) then our claim is F(x) = F0(x) and its complement is
square goodness of fit F(x) ≠ F 0(x). Since the claim contains the equality sign so we can take claim as
test. the null hypothesis and complement as the alternative hypothesis. Thus
H0 : F(x)  F0 (x) for all values of x
H1 : F(x)  F0 (x) for at least one value of x
For testing the null hypothesis, the test statistic is given by
D n  sup S  x   F0  x 
x

where, S(x) is the empirical distribution function based on the observed

sample.
For obtaining the empirical cumulative distribution function S(x), first we
obtain cumulative frequencies and then by the definition of empirical
distribution function we obtain S(x) as
The number of sample observations  x
S x  
Total number of observations

Here, theoretical cumulative distribution function F0 (x) can be obtained with

the help of the method as described in Unit 4 of MST-003. Therefore, by the
definition of distribution function, we have

X   x  
F0 x   PX  x   P    PZ  z
   

where, P  Z  z  can be to obtained with he help of Table I given in Appendix

at the end of the Block 1 of this course.
x  1.0  3
Thus, when x = 1.0, z    2.0 and
 1
F0  1.0   P  Z  z   P  Z  2.0   0.5  P  2.0  Z  0 

 0.5  P  0  Z  2.0 
 0.5  0.4772  0.0228
30
 3.2  3 One-Sample Tests
Similarly, when x = 3.2, z   0.2 and
1
F0  3.2   P  Z  z   P  Z  0.2 

 0.5  P  0  Z  0.2 

 0.5  0.0793  0.5793 and so on.

Calculation for S  x   F0  x  :
X Observed Cumulative Observed X Theoretical Absolute
Frequency Observed Distribution Z Difference
 Distribution
Frequency Function X3 Function S  x   F0  x 

S(x) 1 F0(x)

1.0 1 1 1/10 = 0.1 –2.0 0.0228 0.0772

1.9 1 2 2/10 = 0.2 –1.1 0.1562 0.0438
2.1 1 3 3/10 = 0.3 –0.9 0.1841 0.1159
2.7 1 4 4/10 = 0.4 –0.3 0.3821 0.0179
2.8 1 5 5/10 = 0.5 –0.2 0.4207 0.0793
3.2 1 6 6/10 = 0.6 0.2 0.5793 0.0207
3.6 1 7 7/10 = 0.7 0.6 0.7257 0.0257
3.9 1 8 8/10 = 0.8 0.9 0.8159 0.0159
4.2 1 9 9/10 = 0.9 1.2 0.8849 0.0151
5.1 1 10 10/10 = 1.0 2.1 0.9821 0.0179

From the above calculations, we have

D n  sup S  x   F0  x   0.1159
x

The critical value of test statistic at 5% level of significance is 0.409.

Since calculated value of test statistic Dn (= 0.1159) is less than critical value
(= 0.409) so we do not reject the null hypothesis i.e. we support the claim at
5% level of significance.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so we may assume that the data have come from the normal
distribution with mean 3 and standard deviation 1.
Now, you can try the following exercises.

E10) Write the main difference between chi-square test and Kolmogorov-
Smirnov (K-S) test for goodness of fit.
E11) The following data were obtained from a table of random numbers of
normal distribution with mean 5.6 and standard deviation 1.2:
6.0 6.6 5.9 5.0 6.2 4.8
5.5 6.3 5.9 5.8 6.6 6.2
Test the hypothesis by K-S test that the data have come from the same
normal distribution at 5% level of significance.

We now end this unit by giving a summary of what we have covered in it.
31
Non-Parametric Tests
13.7 SUMMARY
In this unit, we have discussed following points:
1. Need of non-parametric tests.
2. Applications of non-parametric tests in various fields.
3. Advantages and disadvantages of non-parametric tests over parametric
tests.
4. The one-sample non-parametric tests.
5. The suitable non-parametric test is appropriate for a particular situation.
6. Test the hypothesis when we cannot make any assumption about the
distribution from which we are sampling.
7. The tests which are used in place of t-test for mean when assumption(s) of
the t-test is (are) not fulfilled, that is, sign test and Wilcoxon signed-rank
test.
8. The test for randomness of data, that is, run test.
9. The Kolmogorov-Smirnov test for goodness of fit.

13.8 SOLUTIONS /ANSWERS

E1) Refer Section 13.2.
E2) The main difference between t-test and sign test is that the t-test based
on the assumption that the parent population is normal where as sign
test is not based on such assumption.
E3) Here, distribution of the waiting time on a well travelled rout is not
given. So the assumption of normality for t-test is not fulfilled. Also
sample size is small so we can not use Z-test. So we go for sign test.
Here, we wish to test the claim of Metropolitan Area Road Transport
Authority that the average (median) waiting time ( ) on a well travelled
rout is 20 minutes. So our claim is   20 and it complement is   20.
Since claim contains the equality sign so we can take the claim as the
null hypothesis and the complement as the alternative hypothesis. Thus,
H0 :    0  20 and H1 :   20
For applying sign test, we replacing each observation greater than 20
(median value) with a plus sign and each observation less than 20 with
a minus sign and discarding the one observation which equal to 20, we
get
+ − − − + + − + + − −
By counting, we have
S+ = number of plus signs = 5
S− = number of mines signs = 6
n = total number of plus and minus signs = 11
Since alternative hypothesis is two-tailed so the test statistic (S) is given
by

32
 S  min S ,S  One-Sample Tests

 min  5,6   5
Here, n = 11(< 20) so it is a case of small sample. Thus, to take the
decision about the null hypothesis, we determine p-value with the help
of Table I given in Appendix at the end of this block. Here, n = 11,
p = 0.5 and r = 5. Thus, we have
p-value  P S  5  0.50

Since p-value = 0.50 > 0.01 (α). So we do not reject the null hypothesis
i.e. we support the claim at 1% level of significance.
Thus, we conclude that the sample fails to provide us sufficient
evidence against the claim so we may assume that the claim of
Metropolitan Area Road Transport Authority that the average waiting
time is 20 minutes is true.
E4) Here, distribution of the weight of the cadets is not given. So the
assumption of normality for t-test is not fulfilled. So we go for sign test.
Here, we want to test that median weight ( ) of all the cadets of a
college is 50 kg. So our claim is   50 and it complement is   50.
Since claim contains the equality sign so we can take the claim as the
null hypothesis and the complement as the alternative hypothesis. Thus,
H0 :    0  50  median weight is equal to 50kg

H1 :   50  median weight is not equal to 50kg

Since the alternative hypothesis is two-tailed so the test is two-tailed
test.
For applying sign test, we replacing each observation greater than 50
(median value) with a plus sign and each observation less than 50 with
a minus sign and discarding the two observations which are equal to 50,
we get
− − + − − + + − + −
− − + + − + − + + −
+ − + − + + − −
By counting, we have
S+ = number of plus signs = 13
S− = number of minus signs = 15
n = total number of plus and minus signs = 28
Since alternative hypothesis is two-tailed so the test statistic(S) is the
minimum of number of plus signs (S+) and minus signs (S−), that is,
S  minS ,S   min13,15  13

Here, n = 28 ( > 20), so it is a case of large sample. In this case. we the

use Z-test. The test-statistic of Z-test is given by

33
Non-Parametric Tests n 28
S 13 
Z 2  2
n 28
4 4
13  14
  0.38
7
The critical (tabulated) values for two-tailed test at 5% level of
significance are ± zα/2 = ± z0.025 = ± 1.96.
Since calculated value of test statistic Z (= 0.38) is greater than the
critical value (= − 1.96) and less than the critical value (= 1.96), that
means it lies in non-rejection region, so we do not reject the null
hypothesis i.e. we support the claim at 5% level of significance.
Thus, we conclude that the samples do not provide us sufficient
evidence against the claim so we may assume that the median weight of
all the cadets of the centre is 50 kg.
E5) The main difference between sign test and Wilcoxon signed-rank test is
that the sign test utilizes only information whether the difference
between observation and the postulated value of the median is positive
or negative, that is, it considers only sign of differences but ignores the
magnitude of these differences whereas the Wilcoxon signed-rank test
uses both the signs as well as the magnitude of the differences.
E6) We are given that characteristic under study is continuous and
population is symmetric about its median and other assumptions of
Wilcoxon signed-rank test hold. So we can apply this test.
Here, we want to test that sample is taken from the population having
median ( ) greater than 18. So our claim is   18 and it complement
is   18. Since complement contains the equality sign so we can take
the complement as the null hypothesis and the claim as the alternative
hypothesis. Thus,
H 0 :    0  18 and H1 :   18
Since the alternative hypothesis is right-tailed so the test is right-tailed
test and the test statistic (T) is the sum negative ranks (T−).
Calculation for T:
S. No. Sample Difference Absolute Value Rank of Signed
(X) d  X   of Difference d Rank
 X  18 d
1 20.2 2.2 2.2 7 7
2 24.1 6.1 6.1 10 10
3 21.3 3.3 3.3 8 8
4 17.2 −0.8 0.8 2 −2
5 19.8 1.8 1.8 5 5
6 16.5 −1.5 1.5 4 −4
7 21.8 3.8 3.8 9 9
8 18.7 0.7 0.7 1 1
9 17.1 −0.9 0.9 3 −3
10 19.9 1.9 1.9 6 6

34
 Therefore, we have One-Sample Tests

T+ = 7 + 10 + 8 + 5 + 9 + 1 + 6 = 46
T− = 2 + 4 + 3 = 9
n = number of non-zero di’s =10
So the value of test statistic T = 9
The critical value of test statistic T for right-tailed test corresponding
n = 10 at 5% level of significance is 11.
Since calculated value of test statistic T (= 9) is less than the critical
value (= 11) so we reject the null hypothesis and support the alternative
hypothesis i.e. we support the claim at 5% level of significance.
Thus, we conclude that the sample fails to provide us sufficient
evidence against the claim so we may assume that the sample is taken
from the population with median greater than 18.
E7) Here all the assumptions of Wilcoxon signed-rank test hold. So we go
for this test.
Here, we want to test the manufacturer’s claim that the average
(median) breaking strength ( ) of the rods is 30 pounds. So our claim is
  30 and it complement is   30. So we can take claim as the null
hypothesis and complement as the alternative hypothesis. Thus,
H 0 :    0  30  median breakingstrength is30

H 1 :   30  median breakingstrengthis not 30

Since the alternative hypothesis is two-tailed so the test is two-tailed
test and the test statistic is the smaller of sum of positive ranks (T+) and
sum of negative ranks (T−), that is,

T  min T ,T 

Calculation for T:
S. No. Breaking Difference Absolute Value Rank of Signed
Strength d  X   of Difference d Rank
(X)  X  30 d
1 19 −11 11 23.5 −23.5
2 30 Tie --- --- ---
3 28 −2 2 4 −4
4 35 5 5 9.5 9.5
5 23 −7 7 16 −16
6 37 7 7 16 16
7 36 6 6 12.5 12.5
8 32 2 2 4 4
9 40 10 10 21.5 21.5
10 24 −6 6 11.5 −11.5
11 31 1 1 1 1
12 33 3 3 7 7

35
Non-Parametric Tests 13 21 −9 9 20 −20
14 36 6 6 12.5 12.5
15 30 Tie --- --- ---
16 32 2 2 4 4
17 26 −4 4 8 −8
18 40 10 10 21.5 21.5
19 38 8 8 18.5 18.5
20 28 −2 2 4 4
21 30 Tie --- --- ---
22 41 11 11 23.5 23.5
23 22 −8 8 18.5 −18.5
24 43 13 13 25.5 25.5
25 25 −5 5 9.5 −9.5
26 28 −2 2 4 −4
27 24 −6 6 125 −12.5
28 37 7 7 16 16
29 17 −13 13 25.5 −25.5

From the above calculation, we have

T+ = 9.5 + 16 + 12.5 + 4 + 21.5 + 1 + 7 + 12.5 + 4 + 21.5 + 18.5 +
23.5 + 25.5 + 16 = 193.0
T− = 23.5 + 4 + 16 + 12.5 + 20 + 8 + 4 + 18.5 + 9.5 + 4 + 12.5 + 25.5
= 158.0
n = number of non-zero di’s = 29
Putting the values in test statistic, we have

T  min T ,T   min 193.0,158.0  158.0

Also n = 29 (> 20), therefore, it is the case of large sample. So in this

case, we use Z-test. The test statistic of Z-test is given by
T  E T 
Z ~ N  0,1 
SE  T 
n(n  1) 29(29  1)
where, E  T     217.5 and
4 4
n  n  1   2n  1 
SE  T  
24
29  29  1   2  29  1 
  46.25
24
Putting the values in test statistic Z, we have
158.0  217.5
Z  1.29
46.25
The critical (tabulated) values for two-tailed test at 1% level of
significance are ± zα/2 = ± z0.005 = ± 2.58.

36
 Since calculated value of test statistic Z (= −1.29) is greater than the One-Sample Tests
critical value (= − 2.58) and less than the critical value (= 2.58), that
means it lies in non-rejection region so we do not reject the null
hypothesis i.e. we support the claim at 1% level of significance.
Thus, we conclude that the sample fails to provide us sufficient
evidence against the claim so we may assume that the manufacturer’s
claim that the average (median) breaking strength of the rods is 30
pounds is true.
E8) Here, we want to test the randomness of the given data so we go for run
test.
We want to test that given data are in random order. So our claim is
“the data are in random order” and its complement is “the data are not
in random order”. So we can take claim as the null hypothesis and
complement as the alternative hypothesis. Thus,
H0 : The data are in random order
H1 : The data are not in random order
For applying run test, we need given observations in a sequence of two
symbols. In order to convert given observations into a sequence of two
symbols, first of all we replace the given observations by the symbol A
if it is above the median value (= 0) and by symbol B if it is below the
median value and discard the observation equal to median value. In this
way, we get a sequence of symbols A’s and B’s which is given below:
AAA
 BBA BBAA
 BA
 BBBBBAAA
 

1 2 3 4 5 6 7 8 9

Here, we see that there is a first run of three A’s then there is a run of
two B’s and so on. In the last there is a run of three A’s.
Here by counting, we have
n1 = number of symbol A occurs = 10
n 2 = number of symbol B occurs = 10
R = total number of runs in above sequence = 9
For run test the test statistic is the total number of runs in the given
sequence.
Since n1 = 10(< 20) and n2 = 10(< 20), so it is a small sample case.
The lower and upper critical values of runs corresponding to n1 and n 2
at 5% level of significance are R L  6, R U  16.
Since observed number of runs (R = 9) lies between critical values of
runs, R L  6 and R U  16 so we do not reject the null hypothesis i.e. we
support the claim at 5% level of significance.
Thus, we conclude that the given data are in random order.
E9) Here, we want to test the randomness of the cars entered in the parking
so we go for run test.
Also we wish to test that the cars entered in the parking are in random
order. So our claim is “the cars that enter in the parking are in random
order” and its complement is “the cars that enter in the parking are not
37
Non-Parametric Tests in random order”. So we can take claim as the null hypothesis and
complement as the alternative hypothesis. Thus,
H0 : The cars that enter in the parking are in random order
H1 : The cars that enter in the parking are not in random order
For applying run test, we count the number of runs in the given
sequence.
II
 FFF
I F
 I FFFF
 II 
F III
 FF I FF
 I FFFF
 II 
FFF I FF
 IIII
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

We see that there is a first run of two I’s then there is a run of three F’s
and so on in the last there is a run of four I’s. By counting, we have
n1 = number of symbol I occurs = 18
n 2 = number of symbol F occurs = 22
R = total number of runs in above sequence = 19
Since n2 = 22(>20) therefore, we use Z-test. The test statistic of Z-test is
given by
R  E R
Z ~ N  0,1
SE  R 

2n1n 2
where E  R   1
n
2  18  22
  1  20.8 and
40
2n1n 2 (2n1n 2  n)
SE(R) 
n 2 (n  1)

2  18  22  2  18  22  40 
  9.54  3.08
402  40  1

Putting the values in test statistic Z, we have

19  20.8
Z   0.58
3.08
The critical (tabulated) values for two-tailed test at 1% level of
significance are ± zα/2 = ± z0.005 = ± 2.58.
Since calculated value of Z (= − 0.58) is greater than critical value
(= − 2.58) and less than critical value (= 2.58), that means it lies in non-
rejection region, so we do not reject the null hypothesis i.e. we support
the claim at 1% level of significance.
Thus, we conclude that the sample fails to provide us sufficient
evidence against the claim so we may assume that the cars enter in the
parking lot are in random order.

E10) The main difference between chi-square test and Kolmogorov-Smirnov

(K-S) test is that the chi-square test is designed for categorical data
whereas K-S test is designed for the continuous data.
38
E11) Here, we want to test that the data have come from normal distribution One-Sample Tests
with mean 5.6 and standard deviation 1.2. If F0(x) is the cumulative
distribution function of this normal distribution then our claim is
F(x) = F0(x) and its complement is F(x) ≠ F0(x). Since the claim
contains the equality sign so we can take claim as the null hypothesis
and complement as the alternative hypothesis. Thus
H0 : F(x)  F0 (x) for all values of x

H1 : F(x)  F0 (x) for at least one value of x

For testing the null hypothesis, the test statistic is given by

D n  sup S  x   F0  x 
x

where, S(x) is the empirical cumulative distribution function based on

the observed sample.
For obtaining the empirical cumulative distribution function S(x), first
we obtain cumulative frequencies and then by the definition of
empirical cumulative distribution function we obtain S(x) as
The number of sample observations  x
S x  
Total number of observations

Here, theoretical cumulative distribution function F0 (x) can be obtained

with the help of the method as described in Unit 4 of Block 4 of the
course MST-003. Therefore, by the definition of distribution function,
we have

X   x  
F0 x   PX  x   P    PZ  z
   

where, P  Z  z  can be to obtained with he help of Table I given in

Appendix at the end of the Block 1 of this course.
x   4.8  5.6
Thus, when x = 4.8, z     0.67 and
 1.2
F0  4.8   P  Z  z   P  Z  0.67 

 0.5  P  0  Z  0.67 

 0.5  0.2486
 0.2514
5.8  5.6
Similarly, when x = 5.8, z   0.17 and
1.2
F0  5.8   P  Z  z   P  Z  0.17 

 0.5  P  0  Z  0.17 

 0.5  0.0675  0.5675 and so on.

39
Non-Parametric Tests Calculation for S  x   F0  x  :
X Observed Cumulative Observed X  Theoretical Absolute
Frequency Observed Distribution Z Distribution Difference

Frequency Function X  5.6 Function S  x   F0  x 
S(x)  F0 (x)
1.2

4.8 1 1 1/12 = 0.0833 −0.67 0.2514 0.1681

5.0 1 2 2/12 = 0.1667 −0.50 0.3085 0.1418
5.5 1 3 3/12 = 0.2500 −0.08 0.4681 0.2181
5.8 1 4 4/12 = 0.3333 0.17 0.5675 0.2342
5.9 2 6 6/12 = 0.5000 0.25 0.5987 0.0987
6.0 1 7 7/12 = 0.5833 0.33 0.6293 0.0460
6.2 2 9 9/12 = 0.7500 0.50 0.6915 0.0585
6.3 1 10 10/12 = 0.8333 0.58 0.7190 0.1143
6.6 2 12 12/12 = 1.0000 0.83 0.7967 0.2033

From the above calculations, we have

D n  sup S  x   F0  x   0.2342
x
The critical value of test statistic Dn corresponding n = 12 at 5% level
of significance is 0.375.
Since calculated value of test statistic Dn (= 0.2342) is less than critical
value (= 0.375) so we do not reject the null hypothesis i.e. we support
the claim at 5% level of significance.
Thus, we conclude that the sample fails to provide us sufficient
evidence against the claim so we may assume that data have come from
the normal distribution with mean 5.6 and standard deviation 1.2.

40
UNIT I4 TWO-SAMPLE TESTS
Structure
14.1 Introduction
Objectives
14.2 Paired Sign Test
14.3 Wilcoxon Matched-Pair Signed-Rank Test
14.4 Mann-Whitney U Test
14.5 Kolmogorov-Smirnov Two-Sample Test
14.6 Summary
14.7 Solutions / Answers

14.1 INTRODUCTION
In previous unit, we have already highlighted the main difference between
parametric and non-parametric tests. Also we have discussed some well
know/popular/commonly used one-sample non-parametric tests in the same
unit. But comparison of a characteristic in two populations is a common
problem faced by statisticians. We have already addressed this problem when
distributions of the two populations are known to us. But what will do if
assumption(s) of the test is (are) not fulfilled as distributions of the two
populations are not known to us. This unit is devoted to address this problem.
In this unit, we describe some of the frequently used two-sample tests based on
non-parametric methods.
This unit is divided into seven sections. Section 14.1 is described the need of
two-sample non-parametric tests. The paired sign test and Wilcoxon matched-
pair signed-rank test for two samples are described in Sections 14.2 and 14.3
respectively. The Mann-Whitney U test which is used to test the difference in
median of two independent populations is discussed in Section 14.4. In Section
14.5, we describe the Kolmogorov-Smirnov two-sample test. Unit ends by
providing summary of what we have discussed in this unit in Section 14.6 and
solution of exercises in Section 14.7.
Objectives
After completion of this unit, you should be able to:
 describe the need of two-sample non-parametric tests;

 identify which of the two-sample non-parametric tests is appropriate for a

given situation;

 perform the test which is used in place of paired t-test when assumption(s)
of the paired t-test is (are) not fulfilled, that is, paired sign test and
Wilcoxon matched-pair signed-rank test;

 describe the Mann-Whitney U test which is used when the assumption(s) of

two sample t-test is (are) not fulfilled;

 describe the Kolmogorov-Smirnor two-sample test for testing the

hypothesis that two samples come from the populations having the same
distribution.
41
Non-Parametric Tests
14.2 PAIRED SIGN TEST
In social sciences and in marketing research, generally two related groups are
paired and we are interested to examine differences among two related groups.
Paired sign test is the non- This often happens when the observations are recorded as before and after.
parametric version of the Here, the paired observations are recorded on the same individuals or items. In
paired t-test. It is used when these situations, we use paired t-test. If the assumptions of paired t-test is (are)
the assumption(s) of paired t- not fulfilled then we use paired sign test, that means paired sign test is the non
test is (are) not fulfilled. parametric version of the paired t-test. For example, if a college administrator
is interested in knowing how a particular three-weeks training on basic
Statistics affects student’s grades in Statistics paper. In this situation, paired
t-test is not applicable because of grades, that is, observations are given in
ordinal scale as A, A+, B, B+, etc. But paired sign test can be applied because it
needs only to convert the observations into plus and minus signs.
Assumptions
This test needs following assumptions to work:
(i) The pairs of measurements are independent.
(ii) The measurement scale is at least ordinal within each pair.
(iii) The variable under study is continuous.
Let us discuss the general procedure of this test:
Let  X1, Y1  ,(X2 , Y2 ),...,(X n , Yn ) be a random sample of n independent items
/ units / individuals, each with two measurements as before and after the
training, diet, treatment, medicine, etc. Also let µ 1 and µ 2 be the medians of the
populations before and after the training, diet, treatment, medicine, etc. Here,
we want to test whether there is an effect of a diet, training, treatment,
medicine, etc. or not. So we can take the null and alternative hypotheses as
H 0: µ 1 = µ 2 or H 0 : µ 1  µ 2  0
H1:  1   2 or H1:  1  2  0 for two-tailed test
H 0 :  1   2 and H1:  1   2 
or  for one-tailed test 
H 0 :  1   2 and H1:  1   2 

The procedure for paired sign test is similar to sign test for one sample as
discussed in pervious unit. After setting null and alternative hypothesis, paired
sign test involves following steps:
Step 1: First of all, we convert the given observations into a sequence of plus
and minus signs. For this, we find the difference Xi  Yi for all
observations and consider only the sign of difference as plus or minus
or compare X i with Yi for all observations. If X i  Yi then we replace
the pair of observation  Xi , Yi  by a plus sign and if Xi  Yi then we
replace  Xi , Yi  by minus sign. But when Xi  Yi , then
corresponding pair of observations gives no information in terms of
plus or minus signs so we exclude all such pairs from the analysis
part. Due to such observations our common sample size reduces and
let reduced common sample size be denoted by n.
Step 2: After that, we count the number of plus signs and number of minus
signs. Suppose they are denoted by S and S respectively.
42
Step 3: When null hypothesis is true and the population is dichotomised on Two-Sample Tests
the basis of sign of the difference then we expect that the number of
plus signs (success) and number of minus signs (failure)
approximately equal. And number of plus signs or number of minus
signs follows binomial distribution (n, p = 0.5). For convenient
consider the smaller number of plus or minus signs. If S is less than
S then we will take plus sign as success and minus sign as failure.
Similarly, if S is less than S  then we will assume minus sign as
success and plus sign as failure.
Step 4: To take the decision about the null hypothesis, we use concept of p- The critical values of this
value as we have discussed in sign test for one sample. For p-value test are not generally in
we determine the probability that test statistic is less than or equal to tabular form and slightly
the actually observed of plus or minus signs. Since distribution of difficult to obtain whereas
number of plus or minus signs is binomial (n, p = 0.5) therefore, this p-value is easy to
probability can be obtained with the help of Table I given in obtained with the help of
Appendix at the end of this block which provide the cumulative cumulative binomial table
binomial probability and compare this probability with the level of so we use concept of p-
significance (α). Here, test statistic depends upon the alternative value for take the decision
hypothesis so the following cases arise: about the null hypothesis.

For one-tailed test:

Case I: When H 0 :  1   2 and H1 :  1   2 (right-tailed test)
In this case, we expect that number of minus signs (S−) is smaller than
number of plus signs (S+) therefore, the test statistic(S) is the number
of minus signs (S−). The p-value is determined as
p-value = P  S  S  
If p-value is less than or equal to α, that is, P  S  S     then we
reject the null hypothesis at α level of significance and if the p-value
is greater than α then we do not reject the null hypothesis.
Case II: When H 0 :  1   2 and H1 :  1   2 (left-tailed test)
In this case, we expect that number of plus signs (S+) is smaller than
number of minus signs (S−) therefore, the test statistic(S) is the
number of plus signs (S+). The p-value is determined as
p-value = P  S  S  
If p-value is less than or equal to α, that is, P  S  S     then we
reject the null hypothesis at α level of significance and if the p-value
is greater than α then we do not reject the null hypothesis.
For two-tailed test: When H 0 :  1   2 and H1 :  1   2
For two tailed test, the test statistic (S) is the smaller of number of
plus signs (S+) and minus signs (S−), that is,
S  min S ,S 
and approximate p-value is determined as
p-value = 2P  S  S   ; if S  is small

p-value = 2P  S  S   ; if S  is small

43
Non-Parametric Tests If p-value is less than or equal to α then we reject the null hypothesis
at α level of significance and if the p-value is greater than α then we
do not reject the null hypothesis.
For large sample  n  20  :
For a large sample size n greater than and equal to 20, we use normal
approximation to binomial distribution with mean
1 n
E(S)  np  n   … (1)
2 2
and variance
1 1 n
Var(S)  npq  n    … (2)
2 2 4
Therefore, we use normal test (Z-test). The test statistic of Z-test is
given by
S  E S
Z
SE  S
n
S
 2 ~ N  0,1 [Using equations (1) and (2)] … (3)
n
4
After that, we calculate the value of test statistic Z and compare it
with the critical value(s) given in Table 10.1 at prefixed level of
significance α. Take the decision about the null hypothesis as
described in Section 10.2 of Unit 10 of this course.
Now, let us do some examples based on paired sign test.
Example 1: To compare the softness of the paper of two types A and B, a
sample of each product of A and B is given to ten judges. Each judge rates the
softness of each product on a scale from 1 to 10 with higher ratings implying a
softer product. The results of the experiment are shown below. The first rating
in the pair is of product A and the second is of product B.
(4, 5) (6, 4) (8, 5) (9, 8) (4, 1) (7, 9) (6, 2) (5, 3) (8, 2) (6, 7)
Apply the paired sign test to test the hypothesis that the products A and B are
equally soft at 1% level of significance.
Solution: Here, the ratings are given in pair form so it can be considered as the
case of dependent samples. But assumption of normality is not given also the
given data are in the form of ranks so we cannot use the paired t-test in this
case. So we go for paired sign test.
Here, we wish to test that the product A and B are equally soft. If  1 and  2
denote the average rank of softness of products A and B respectively then our
claim is  1   2 and it complement is  1   2 . Since claim contains the
equality sign so we can take the claim as the null hypothesis and the
complement as the alternative hypothesis. Thus,
H 0 :  1   2 and H 1 :  1   2
Since the alternative hypothesis is two-tailed so the test is two-tailed test and
the test statistic (S) is the minimum of plus signs (S+) and minus signs ( S  ),
that is,
44
 S  min S ,S  Two-Sample Tests

Calculation for S:
Judge Product Product Sign of Difference
A B (A−B)
1 4 5 –
2 6 4 +
3 8 5 +
4 9 8 +
5 4 1 +
6 7 9 –
7 6 2 +
8 5 3 +
9 8 2 +
10 6 7 –

From the above, we have

n = number of non zero differences = 10
S‒ = number of plus signs = 7
S‒ = number of minus signs = 3
Therefore,
S  min S ,S   min 7, 3  3
To take the decision about the null hypothesis, we determine p-value with the
help of Table I given in Appendix at the end of this block. Here, n = 10,
p = 0.5 and r = 3. Thus, we have
p-value  2P S  3   2  0.1719  0.3438

Since p-value = 0.3438 > 0.01(= α) so we do not reject the null hypothesis i.e.
support the claim.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so we may assume that products A and B are equally soft.
Example 2: The following are the weight (in pounds) before and after a dieting
programme of 18 persons for one month:
Before 148.0 181.5 150.5 230.2 196.5 206.3 170.0 214.4 140.6
After 139.2 172.4 154.2 212.6 193.2 200.0 163.0 218.0 140.0
Before 166.8 150.3 197.2 168.0 172.4 160.3 210.0 172.4 180.2
After 166.8 156.2 185.4 158.5 167.2 162.3 202.3 166.2 170.5

Using the paired sign test to test whether the dieting programme is effective or
not at 5% level of significance.
Solution: It is case of before and after study, so either we can go for paired
t-test or paired sign test. But t-test requires assumptions of normality which is
not given. So we go for paired sign test.
If µ 1 and µ 2 denote the median weights of the two depending populations
before and after the dieting programme respectively. Here, we want to test
whether the dieting programme is effective or not. So we can take our claim
that the dieting programme is effective that means average (median) weight
before dieting programme is greater than average (median) weight after the
programme, that is,  1   2 and it complement is  1   2 . Since complement
contains the equality sign so we can take the complement as the null hypothesis
and the claim as the alternative hypothesis. Thus,
45
Non-Parametric Tests H 0 :  1   2 and H1 :  1   2
Since the alternative hypothesis is right-tailed so the test is right-tailed test.
Since alternative hypothesis is right-tailed so the test statistic (S) is the number
of minus signs ( S  )
Calculation for S:
Weight before Weight after Sign of Difference
S. No
diet (X) diet(Y) (X – Y)
1 148.0 139.2 +
2 181.5 172.4 +
3 150.5 154.2 –
4 230.2 212.6 +
5 196.5 193.2 +
6 206.3 200.0 +
7 170.0 163.0 +
8 214.4 218.0 –
9 140.6 140.0 +
10 166.8 166.8 Tie
11 150.3 156.2 –
12 197.2 185.4 +
13 168.0 158.5 +
14 172.4 167.2 +
15 160.3 162.3 –
16 210.0 202.3 +
17 172.4 166.2 +
18 180.2 170.5 +

From the above, we have

n = number of non-zero difference = 17
S+ = number of plus signs = 13
S− = number of minus signs = 4
Therefore, value of test statistic(S) is S− = 4
To take the decision about the null hypothesis, we determine p-value with the
help of Table I given in Appendix at the end of this block. Here, n = 17,
p = 0.5 and r = 4. Thus, we have
p-value  P S  4   0.0245
Since p-value = 0.0245 < 0.05 (α) so we reject the null hypothesis and support
the alternative hypothesis i.e. we support the claim.
Thus, we conclude that the samples fail to provide us sufficient evidence
against the claim so may assume that the dieting programme is effective.
Now, you can try the following exercises.
E1) Write one difference between paired sign test and paired t-test.
E2) Students are given a course designed to improve their IQ on a standard
test. Their scores before and after the course are recorded as follows:
Before 95 112 128 96 97 117 105 95 99 86 75 82
After 99 115 133 96 99 120 106 106 100 89 90 85

With the help of paired sign test, examine whether IQ is affected due to
course at 1% level of significance?
46
 E3) A randomly chosen group of 15 consumers was asked to rate the two Two-Sample Tests
products A and B on a scale of 1 to 5 (5 being the highest). For each
consumer, the first rating in the pair is of product A and the second is of
product B. The data are:
(5, 2) (2, 3) (4, 1) (4, 2) (5, 3) (4, 4) (5, 4) (5, 1) (2, 4)
(3, 2) (3, 2) (4, 3) (3, 3) (3, 4) (2, 3)
The aim of study was to show that product A is favoured over product B.
Do these data present evidence to justify this claim at 1% level of
significance?

14.3 WILCOXON MATCHED-PAIR SIGNED-RANK

TEST
We have already discussed Wilcoxon signed-rank test for one-sample in Unit
13. The Wilcoxon signed-rank test is also useful in comparing two populations
for which we have paired observations. In this case, it is known as Wilcoxon
matched-pair signed rank-test. As such, the test is a good alternative to the
paired t-test in cases where the differences between paired observations are not
believed to be normally distributed. Paired sign test discussed in previous
section of this unit is also for the same situation. The paired sign test uses only
the information whether X is larger or smaller than Y. This test is fine if the
information about the observation sample is available in ordinal scale only. But
if the measurement of the observation is available interval or ratio scales then
choice of paired sign test is not recommended. Because this test do not take
into account the information available in terms of the magnitude of the
of differences. To overcome this drawback of the paired sign test Wilcoxon
matched-pair signed-rank test do the job for us, which takes into account the
o be information of signs as well as of magnitude of differences. The Wilcoxon
mean. matched-pair signed-rank test is exactly the same as Wilcoxon signed-rank test
described in Section 13.4 of previous unit of this course when di  Xi   0 is
replaced by di  Xi  Yi .
Assumptions
This test needs following assumptions to work:
(i) The distribution of differences di 's is symmetric.
(ii) The differences di ' s are mutually independent.
(iii) The measurement scale is at least interval within each pair.
(iv) The variable under study is continuous.
Let us discuss the general procedure of this test:
Let  X1, Y1  ,(X2 , Y2 ),...,(X n , Yn ) be a random sample of n independent items
/ units / individuals each with two measurements as before and after the
training, diet, treatment, medicine, etc. Also let µ 1 and µ 2 be the medians of the
populations before and after the training, diet, treatment, medicine, etc. Here,
we want to test whether there is an effect of a diet, training, treatment,
medicine, etc. or not. So we can take the null and alternative hypotheses as
H 0 : µ 1 = µ 2 or H 0 : µ 1  µ 2  0
H1:  1   2 or H1:  1  2  0 for two-tailed test 
47
Non-Parametric Tests H 0 :  1   2 and H1:  1   2 
or  for one-tailed test 
H 0 :  1   2 and H1:  1   2 
The procedure of this test is similar to Wilcoxon signed-rank test for one-
sample as discussed in pervious unit. After setting null and alternative
hypotheses, this test involves following steps:
Step 1: First of all, we obtain the difference di for all observations with their
plus and minus sign as
di  Xi  Yi for all observations
But when the observation X i equal to Yi then the difference
di  Xi  Yi gives no information in terms of signs as well as in terms
of magnitude of the difference. So we exclude all such observations
from the analysis part. Due to these observations our sample size
reduces and let reduced sample size be denoted by n.
Step 2: After that, we find the absolute value of these di’s obtained in Step1as
d 1 , d 2 ,..., d n .

For more detailed about Step 3: In this step, we rank all di ’s (obtained in Step 2) with respect to
repeated (tied) rank please their magnitudes from smallest to largest, that is, the rank 1 is given
go through Section 7.4 of to the smallest of di ’s, rank 2 is given to the second smallest and so
the Unit 2 of MST-002
on up to the largest di ’s. If several values are same (tied), we assign
each the average of ranks they would have received if there were no
repetition.
Step 4: Now, assign the signs to the ranks which the original differences
have.
Step 5: Finally, we calculate the sum of the positive ranks (T+) and sum of
negative ranks (T−) separately.
Under H0, we expect approximately equal number of positive and
negative ranks. And under the assumption that distribution of
differences is symmetric about its median, we expect that sum of the
positive ranks (T+) and sum of negative ranks (T−) are equal.
Step 6: Decision Rule:
To take the decision about the null hypothesis, the test statistic (T) is
the smaller of T+ and T−. And the test statistic is compared with the
critical (tabulated) value for a given level of significance (α) under the
condition that the null hypothesis is true. Table II given in Appendix
at the end of this block provides the critical values of test statistic at α
level of significance for both one-tailed and two-tailed tests. Here, test
statistic depends upon the alternative hypothesis so the following
cases arise:
For one-tailed test:
Case I: When H 0 :  1   2 and H1 :  1   2 (right-tailed test)
In this case, we expect that sum of negative ranks (T−) is smaller than
sum of positive ranks (T+) therefore, the test statistic (T) is the sum of
negative ranks (T−).

48
 If computed value of test statistic (T) is less than or equal to the Two-Sample Tests
critical value T at α level of significance, that is, T  T then we
reject the null hypothesis at α level of significance, otherwise we do
not reject the null hypothesis.
Case II: When H 0 :  1   2 and H1 :  1   2 (left-tailed test)
In this case, we expect that sum of positive ranks (T+) is smaller than
sum of negative ranks (T−) therefore, the test statistic (T) is the sum of
positive ranks (T+).
If computed value of test statistic (T) is less than or equal to the
critical value T at α level of significance, that is, T  T then we
reject the null hypothesis at α level of significance, otherwise we do
not reject the null hypothesis.
For two-tailed test:
When H 0 :  1   2 and H1 :  1   2
In this case, the test statistic (T) is the smaller of sum of positive
ranks (T+) and sum of negative ranks (T−), that is,
T  min T ,T 
If computed value of test statistic (T) is less than or equal to the
critical value Tα/2 at α level of significance, that is, T  T / 2 then we
reject the null hypothesis at α level of significance, otherwise we do
not reject the null hypothesis.
For large sample (n > 25):
For a large sample size n greater than 25, the distribution of test
statistic (T) approximated by a normal distribution with mean
n ( n  1)
E T   … (4)
4

and variance
n n  12n  1
Var T   … (5)
24
The proof of mean and variance of test statistic T is beyond the scope
of this course.
Therefore in this case, we use normal test (Z-test). The test statistic of
Z-test is given by
T  ET  T  ET 
Z 
SE  T  Var  T 
n  n  1
T
 4 ~ N  0,1  Using equations  … (6)
n  n  1 2n  1  (4) and (5) 
24
After that, we calculate the value of test statistic Z and compare it
with the critical value(s) given in Table 10.1 at prefixed level of
significance α. Take the decision about the null hypothesis as
described in Section 10.2 of Unit 10 of this course.
49
Non-Parametric Tests Now, it is time to do some examples based on above test.
Example 3: A group of 12 children was tested to find out how many digits
they would repeat from memory after hearing them once. They were given
practice session for this test. Next week they were retested. The results
obtained were as follows:
Child Number 1 2 3 4 5 6 7 8 9 10 11 12
Recall Before 6 4 5 7 6 4 3 7 8 4 6 5
Recall After 6 6 4 7 6 5 5 9 9 7 8 7

Assuming that the distribution of the differences of the scores before and after
the practice session is symmetrical about its median, can the memory practice
session improve the performance of children at 5% level of significance using
Wilcoxon matched-pair signed-rank?
Solution: It is case of before and after study so we can go for paired t-test or
paired sign test or Wilcoxon matched-pair signed-rank. But t-test requires
assumptions of normality which is not given. Also measurement of
observations is available in interval scale, so we will go for Wilcoxon matched-
pair signed-rank test.
Here, we wish to test that memory practice session improves the performance
of children. If  1 and  2 denote the average (median) score before and after the
practice session so our claim is  1   2 and its complement is  1   2 . Since
complement contains the equality sign so we can take the complement as the
null hypothesis and the claim as the alternative hypothesis. Thus,
H0 :  1   2
H 1 :  1   2  memory practice session improves the performance 
Since the alternative hypothesis is left-tailed so the test is left-tailed test and the
test statistic (T) is the sum of positive ranks (T+).
Calculation for T:
Child Recall Recall Difference Absolute Rank of Signed
Number Before After (X−Y) Difference d d Rank
(X) (Y)
1 6 6 Tie --- ---- ----
2 4 6 −2 2 6 −6
3 5 4 −1 1 2 −2
4 7 7 Tie --- ---- ----
5 6 6 Tie --- ---- ----
6 4 5 1 1 2 2
7 3 5 −2 2 6 −6
8 7 9 −2 2 6 −6
9 8 9 −1 1 2 −2
10 4 7 −3 3 9 −9
11 6 8 −2 2 6 −6
12 5 7 −2 2 6 −6

From the above calculations, we have

T+ = sum of positive ranks = 2
n = number of non-zero di’s = 9

50
So the value of test statistic (T) is T+ = 2 Two-Sample Tests

The critical (tabulated) value of test statistic for one-tailed test corresponding
n = 9 at 5% level of significance is 9.
Since calculated value of test statistic T( = 2) is less than the critical value (= 9)
so we reject the null hypothesis and support the alternative hypothesis i.e. we
support the claim at 5% level of significance.
Thus, we conclude that samples fail to provide us sufficient evidence against
the claim so we may assume that memory practice session improves the
performance of children.
Example 4: A drug is given to 30 patients and the increment in their blood
pressures were recorded to be 2, 0, − 4, −1, 2, 3, −2, 5, − 6, 6, 2, 4, 3, 5, 4, 2, 2,
−1, −3, −5, 4, 6, −8, − 4, 2, 4, 0, − 6, 3, 1. Assuming that the distribution of the
increment in the blood pressures is symmetrical about its median, is it
reasonable to believe that the drug is effective to reduce the blood pressure at
1% level of significance?
Solution: We are given that the increment in blood pressures of 30 patients
after given a drug to the patients. The assumption of normality for increment in
blood pressure is not given so we will not go for paired t-test. Also
measurement of observations is available in interval scale, so we will go for
Wilcoxon matched pair signed-rank test.
If µ 1 and µ 2 denote medians of blood pressures of two populations before and
after the drug respectively and we want to test that drug is effective to reduce
the blood pressure so our claim is  1   2 and its complement is  1   2 . Since
complement contains the equality sign so we can take the complement as the
null hypothesis and the claim as the alternative hypothesis. Thus,
H 0 :  1   2
H1 :  1   2 The drug is efective
Since the alternative hypothesis is right-tailed so the test is right-tailed test.
Since in this exercise the increment in the blood pressures is given that means
we are given that the difference of after on before that is Y − X instead of
X – Y therefore, for testing the null hypothesis, the test statistic is the sum of
positive ranks (T+) instead of sum of negative ranks (T−).
Calculation for T:
Patient Increment Absolute Rank of Signed
Number  d  X  Y Difference d d Rank
1 2 2 7 7
2 0 Tie --- ---
3 −4 4 17.5 −17.5
4 −1 1 2 −2
5 2 2 7 7
6 3 3 12.5 12.5
7 −2 2 7 −7
8 5 5 22 22
9 −6 6 25.5 −25.5
10 6 6 25.5 25.5
11 2 2 7 7
12 4 4 17.5 17.5
13 3 3 12.5 12.5
51
Non-Parametric Tests 14 5 5 22 22
15 4 4 17.5 17.5
16 2 2 7 7
17 2 2 7 7
18 −1 1 2 −2
19 −3 3 12.5 −12.5
20 −5 5 22 −22
21 4 4 17.5 17.5
22 6 6 25.5 25.5
23 −8 8 28 −28
24 4 4 17.5 17.5
25 2 2 7 7
26 4 4 17.5 17.5
27 0 Tie --- ---
28 −6 6 25.5 −25.5
29 3 3 12.5 12.5
30 1 1 2 2

From the above calculations, we have

T+ = 264
n = number of non-zero di’s = 28
Therefore, the value of test statistic (T) is T+ = 264
Also, n = 28 (> 25) therefore, it is the case of large sample. So in this case, we
use Z-test. The test statistic of Z-test is given by
T  ET
Z ~ N  0,1 
SE  T 
n(n  1) 28(28  1)
where, E  T     203 and
4 4
n  n  1  2n  1  28  28  1  2  28  1 
SE  T     43.91
24 24
Putting the values in test statistic Z, we have
264  203
Z  1.38
43.91
The critical (tabulated) value of test statistic for right-tailed test at 1% level of
significance is zα = z0.01 = 1.645.
Since calculated value of Z (= 1.38) is less than the critical value (= 1.645), that
means it lies in non-rejection region, so we do not reject the null hypothesis
and reject the alternative hypothesis i.e. we reject the claim at 1% level of
significance.
Thus, we conclude that the sample provide us sufficient evidence against the
claim so we can say that drug is not effective to reduce the blood pressure.
Now, you can try the following exercises.
E4) Write one assumption of the Wilcoxon matched-pair singed-rank test
which is not required in paired sign test.
E5) A group of 10 students were given a test in Statistics and after one
month’s coaching they were given another test of the similar nature. The
following table gives the increment in their marks in the second test over
the first:
Roll No. 1 2 3 4 5 6 7 8 9 10
Increase in Marks 6 −2 8 −4 10 2 5 −4 6 0
52
 Assuming that the distribution of the increment in the marks is Two-Sample Tests
symmetrical about its median, do the marks indicate that students have
benefited by coaching. Test it at 1% level of significance?
E6) To compare the softness of the paper products, a sample of two products
A and B is given to ten judges. Each judge rates the softness of each
product on a scale from 1 to 10 with higher ratings implying a softer
product. The results of the experiment are shown below. The first rating
in the pair is of product A and the second is of product B.
(4, 5) (6, 4) (8, 5) (9, 8) (4, 1) (7, 9) (6, 2) (5, 3) (8, 2)
(6, 7)
Assume that the distribution of the differences of the ranks of products A
and B is symmetrical about its median. Apply the Wilcoxon matched pair
signed-rank test to test the hypothesis that the products A and B are
equally soft at 1% level of significance.

14.4 Mann-Whitney U Test

In Section 14.2 and 14.3, we have discussed the paired sign test and Wilcoxon
matched-pair signed-rank test respectively which are used when the
observations are dependent, that is, observations are paired. When we are
interested in testing of difference in means of two independent populations then
we use two sample t-test. To use the t-test however, it is necessary to make a
set of assumptions (as described in Unit 11 of this course). In particular, it is
necessary that two independent samples be randomly drawn from normal
populations having equal variances and the data be measured in at least of an
interval scale. But in studies of consumer behaviour, marketing research,
experiment of psychology, etc. generally the data are collected in ordinal scale
and the form of the population is not known. Since the parametric t-test could
not be used in such situation, an appropriate non-parametric technique is
needed. In such a circumstance a very simple non-parametric test known as
Mann-Whitney U test may be used.
This test was developed jointly by Mann, Whitney and Wilcoxon. Wilcoxon
considered only the case of equal sample sizes while Mann and Whitney seem
to have been the first to treat the case of unequal sample sizes. Therefore, it is
sometimes called the Mann-Whitney U test and sometimes the Wilcoxon ranks
sum test. Thus, this test may also be viewed as a substitute for the parametric t-
test for the difference between two population means. When assumptions of the
two-sample t-test are fulfilled then this test is slightly weaker than t-test.
Assumptions
This test work under the following assumptions:

(i) The two samples are randomly and independently drawn from their
respective populations.

(ii) The variable under study is continuous.

(iii) The measurement scale is at least ordinal.

(iv) The distributions of two populations differ only with respect to location
parameter.

53
Non-Parametric Tests Let us discuss the general procedure of this test:
Let us suppose that we have two independent random samples X1 , X 2 ,..., X n1
and Y1 , Y2 ,..., Yn 2 drawn from two populations having medians  1 and  2
respectively. Here, we want to test the hypothesis about the medians of two
populations so we can take the null and alternative hypotheses as
H 0 : µ 1 = µ 2 and H 1: µ 1  µ 2 for two-tailed test 
H 0 :  1   2 and H1:  1   2 
or  for one-tailed test 
H 0 :  1   2 and H1:  1   2 

We can also form the null and alternative hypotheses in the form of distribution
functions. If F1(x) and F2(x) are the distribution functions of the first and
second populations respectively and we want to test that the two independent
samples come from the populations that are identical with respect to location,
that is, median.
Therefore, we can take the null and alternative hypotheses as
H1 : F1 (x)  F2 (x) for at least one value of x 
 for two-tailed test 
H1 : F1 (x)  F2 (X) for at least one value of x 
H 0 : F1 (x)  F2 (x) for at least one value of x 
or  for right-tailed test 
H 1 : F1 (x)  F2 (X) for at least one value of x 
H 0 : F1 (x)  F2 (x) for at least one value of x 
or  for left-tailed test 
H 1 : F1 (x)  F2 (X) for at least one value of x 
After setting null and alternative hypotheses, this test involves following steps:
Step 1: First of all, we combine the observations of two samples.
Step 2: After that, ranking all these combined observations from smallest to
largest, that is, the rank 1 is given to the smallest of the combined
observations, rank 2 is given to the second smallest and so on up to
the largest observation. If several values are same (tied), we assign
each the average of ranks they would have received if there were no
repetition.
Step 3: If null hypothesis is true then we can expect that the sum of ranks of
the two samples are equal. Now for convenience, we consider the
smaller sized sample and calculate the sum of ranks of the
observations of this sample. Let S be the sum of the ranks assigned to
the sample observations of smaller sized sample then for testing the
null hypothesis the test statistic is given by
n1 n1  1
U S ; if n1 is small
2
n2  n2  1
U  S ; if n2 is small
2
Step 4: Obtain critical value(s) of test statistic U at given level of significance
under the condition that null hypothesis is true. Table VII in the
Appendix at the end of this block provides the lower and upper
critical values for a given combination of n1 and n2 at α level of
significance for one-tailed and two-tailed test.
54
Step 5: Decision rule: Two-Sample Tests

To take the decision about the null hypothesis, the calculated value of
test statistic U (computed in Step 3) is compared with the critical
(tabulated) value (obtained in Step 4) at a given level of significance
(α) under the condition that null hypothesis is true. Since test may be
one-tailed or two-tailed so following cases arise:
For two-tailed test: When H0 :  1   2 and H1 :  1   2
For two-tailed test, we see critical values at α/2 for α level of
significance. If calculated value of test statistic U is either less than or
equal to the lower critical value (U L,  / 2 ) or greater than or equal to
the upper critical value (U U,  / 2 ) , that is, U  U L,  / 2 or U  U U,  / 2 then
we reject the null hypothesis at α level of significance. However, if
computed U lies between these critical values, that is,
U L,  / 2  U  U U,  / 2 , then we do not reject the null hypothesis at α%
level of significance.
For one-tailed test:
For one-tailed test, we see the critical value at α for α level of significance.
Case I: When H 0 :  1   2 and H1 :  1   2 (right-tailed test)
If calculated value of test statistic U is greater than or equal to the
upper critical value (U U,  ), that is, U  U U,  then we reject the null
hypothesis at α level of significance. However, if computed U is less
than upper critical value (U U,  ) , that is, U  U U,  then we do not
reject the null hypothesis at α% level of significance.
Case II: When H 0 :  1   2 and H1 :  1   2 (left-tailed test)
If calculated value of test statistic U is less than or equal to the lower
critical value (U L,  ), that is, U  U L,  then we reject the null
hypothesis at α level of significance. However, if computed U is
greater than lower critical value (U L,  ), that is, U  U L,  then we do
not reject the null hypothesis at α% level of significance.
For large (n1 or n2 >20):
When either n1 or n2 exceeds 20, the statistic U is approximately
normally distributed with mean
n 1n 2
E U   … (7)
2
and variance
n1n 2  n1  n 2  1
Var  U   … (8)
12
The proof of mean and variance of test statistic(T) is beyond the
scope of this course.
Therefore in this case, we use normal test (Z-test) (described in Unit
10 of Block 3 of this course.). The test statistic of Z-test is given by
U  E U  U  E U 
Z  ~ N  0, 1 
SE  U  Var  U 
55
Non-Parametric Tests n1 n 2
U
 2 ~ N  0,1  Using equations  … (9)
n 1 n 2  n 1  n 2  1  (7) and (8) 
12
After that, we calculate the value of test statistic Z and compare it
with the critical value(s) given in Table 10.1 at prefixed level of
significance α. Take the decision about the null hypothesis as
described in Section 10.2 of Unit 10 of this course.
Let us do some examples to become more user friendly with the test explained
above:
Example 5: A Statistics professor taught two special sections of a basic course
in which students in each section were considered outstanding. He used a
“traditional” method of instruction (T) in one section and an “experimental”
method of instruction (E) in the other. At the end of the semester, he ranked the
students based on their performance from 1 (worst) to 20 (best).
T 1 2 3 5 8 10 12 13 14 15
E 4 6 7 9 11 16 17 18 19 20

Test whether there is any evidence of a difference in performances based on the

two methods at 5% level of significance.
Solution: It is case of two independent populations and the assumption of
normality of both the populations is not given also the given data are in the
form of ranks so we cannot use two-sample t-test in this case. Also sample
sizes are small so we cannot use the Z-test. Therefore, we go for Mann-
Whitney U test.
Here, we want to test that performance of the students based on the two
methods is different. If  1 and  2 denote the average (median) ranks of the
students who taught by traditional” and “experimental” methods respectively
so our claim is  1   2 and its complement is  1   2. Since complement
contains the equality sign so we can take the complement as the null hypothesis
and the claim as the alternative hypothesis. Thus,
H0 : 1   2
H 1 :  1   2  both methods are different 
Since the alternative hypothesis is two-tailed so the test is two-tailed test.
Here, combined ranks are given. Also n1 = n2 so we take any sum of ranks(S)
of first sample or second sample. We consider first sample. Thus,
S = 1 + 2 + 3 + 5 + 8 + 10 + 12 + 13 + 14 + 15 = 83
Here, we consider the first sample so the test statistic is given by
n1 n1  1 10  (10  1)
U S  83   28
2 2
Since test is two-tailed so the lower and upper critical values of test statistic for
two-tailed test corresponding n1  10 and n2  10 at 5% level of significance are
U L,  / 2  U L, 0.025  79 and U U,  / 2  U U, 0.025  131.
Since calculated value of test statistic U (= 28) is less than critical values
(= 79 and 131) so we reject the null hypothesis and support the alternative
56
hypothesis i.e. we support the claim at 5% level of significance. Two-Sample Tests
Thus, we conclude that the samples fail to provide us sufficient evidence
against the claim so we can say that performances based on two methods are
different.
Example 6: The following data show the median amount to be spent on a
birthday gift by students who are unemployed and employed:
Unemployed 200 350 500 250 150 275
Employed 550 500 1000 450 700 200 500 600 1000

Examine that the median amount to be spent on a birthday gift by students who
are unemployed is lower than for students who are employed by Mann-
Whitney U test at 1% level of significance.
Solution: It is case of two independent populations and the assumption of
normality of both the populations is not given so we cannot use t-test in this
case. Also sample sizes are small so we cannot use the Z-test. Therefore, we go
for Mann-Whitney U test.
Here, we want to test that the median amount to be spent on a birthday gift by
students who are unemployed is lower than for students who are employed. If
 1 and  2 denote the average (median) amount to be spent on a birthday gift by
students who are unemployed and employed respectively so our claim is
 1   2 and its complement is  1   2 . Since complement contains the equality
sign so we can take the complement as the null hypothesis and the claim as the
alternative hypothesis. Thus,
H0 :  1   2

H1 :  1   2  median amount to be spent by students who are 

 unemployed is lower than the students who are employed 
Since the alternative hypothesis is left-tailed so the test is left-tailed test.
To perform the Mann-Whitney U test, we combine the observations of two
samples and for convenient, we arrange the data in ascending order. After that
we rank all these combined observations from smallest to largest as:
Unemployed Ranks Unemployed Ranks
150 1 200 2.5
200 2.5 450 7
250 4 500 9
275 5 500 9
350 6 550 11
500 9 600 12
700 13
1000 14.5
1000 14.5
Total 27.5

Since n1 (= 6) is small than n2 (= 9) so we take sum of the ranks(S) assigned to

the sample of unemployed students. Thus, S = 27.5.

57
Non-Parametric Tests Here, we consider the first sample so the test statistic is given by

n1  n1  1
U S
2
6   6  1
 27.5   6.5
2
To decide about the null hypothesis, the calculated value of test statistic U is
compared with the lower critical (tabulated) value at 1% level of significance.
Since test is left-tailed so the lower critical value of test statistic for left-tailed
corresponding to n1= 6 and n2 = 9 at 1% level of significance is
U L,   U L, 0.01  29.

Since calculated value of test statistic U (= 6.5) is less than lower critical value
(= 29) so we reject the null hypothesis and support the alternative hypothesis
i.e. we support the claim at 1% level of significance.
Thus, we conclude that the samples fail to provide us sufficient evidence
against the claim so we may assume that the median amount to be spent on a
birthday gift by students who are unemployed is lower than for students who
are employed.
Now, you can do following exercises in same manner.
E7) Write one difference between two-sample t-test and Mann-Whitney U
test.
E8) Write one difference between Wilcoxon matched-pair singed-rank test
and Mann-Whitney U test.
E9) The senior class in a particular high school had 25 boys. Twelve boys
lived in villages and other thirteen lived in a town. A test was conduct
to see that village boys in general were physically fit than the town
boys. Each boy in the class was given a physical fitness test in which a
low score indicates poor physical condition. The scores of the village
boys (V) and the town boys(T) are as follows:

Village 15.7 8.2 6.5 7.2 9.0 4.5 10.6

Boys(V) 12.4 16.2 12.9 11.4 5.6

Town 12.7 3.2 11.8 7.9 5.6 6.7 12.6

Boys(T) 7.9 2.7 6.1 3.6 6.5 2.8

Test whether the village boys are more fit than town boys at 5% level of
significance.
E10) The following data represent lifetime (in hours) of batteries for two
different brands A and B:
Brand A 40 30 55 40 40 35 30 40 50 45 40 35

Brand B 45 60 50 60 35 50 55 60 50 50 40 55

Examine, is the average life of two brands is same at 1% level of

significance?

58
 Two-Sample Tests
14.5 KOLMOGOROV-SMIRNOV TWO-SAMPLE
TEST
In Mann-Whitney U test, we have tested the hypothesis that the two
independent samples come from the populations that are identical with respect
to location whereas Kolmogorov-Smirnov two-sample test is sensitive to
differences of all types that may exist between two distributions, that is,
location, dispersion, skewness, etc. Therefore, it is referred as a general test.
This test was developed by Smirnov. This test also carries the name of
Kolmogorov because of its similarity to the one-sample test developed by
Kolmogorov. In Kolmogorov-Smirnov one sample test, the observed (sample
or empirical) cumulative distribution function is compared with the
hypothesized cumulative distribution function whereas in two-sample case the
comparison is made between the empirical cumulative distributions functions
of the two samples.
Assumptions
The assumptions necessary for this test are:
(i) The two samples are randomly and independently drawn from their
respective populations.
(ii) The variable under study is continuous.
(iii) The measurement scale is at least ordinal.
Let us discuss the general procedure of this test:
Let us suppose that we have two independent random samples X1 , X 2 ,..., X n1
and Y1 , Y2 ,..., Yn 2 drawn from first and second populations with distribution
functions F1(x) and F2(x) respectively. Also let S1(x) and S2(x) be the sample or
empirical cumulative distribution functions of samples drawn from first and
second populations respectively.
Generally, we want to test whether independent random samples come from
populations having the same distribution functions in all respect or the
distribution functions of two populations differ with respect to location,
dispersion, skewness, etc. So we can take the null and alternative hypotheses as
H 0 : F1 (x)  F2 (x) for all values of x
H1 : F1 (x)  F2 (x) for at least one value of x
After setting null and alternative hypotheses, the procedure of Kolmogorov -
Smirnov two-sample test summarise in the following steps:
Step 1: This test is based on the comparison of the empirical (sample)
cumulative distribution functions, therefore, first of all we compute
sample cumulative distribution functions S1(x) and S2(x) from the
sample data as the proportion of the number of sample observations
less than or equal to some number x to the number of observations,
that is,
The number of sample observations less than or equal to x
S1  x  
n1
and
59
Non-Parametric Tests The number of sample observations less than or equal to x
S2  x  
n2
Step 2: After finding the empirical cumulative distribution functions S1(x)
and S2(x) for all possible values of x, we find the deviation between
the empirical cumulative distribution functions for all x. That is,
S1 (x)  S2 (x) for all x
Step 3: If the two samples have been drawn from identical populations then
S1(x) and S2(x) should be fairly close for all value of x. Therefore, we
find the point at which the two functions show the maximum
deviation. So we take the test statistic which calculate the greatest
vertical deviation between S1(x) and S2(x), that is,
D  sup S1 (x)  S2 (x)
x

which is read as “D equals the supreme over all x, of the absolute

value of the difference S1 (x)  S2 (x) ”
Step 4: Obtain critical value of test statistic at α % level of significance under
the condition that null hypothesis is true. Table V and VI in
Appendix provides the critical values of test statistic for equal and
unequal sample sizes respectively at different level of significance.
Step 5: Decision Rule:
To take the decision about the null hypothesis, the test statistic
(calculated in Step 3) is compared with the critical (tabulated) value
(obtained in Step 4) for a given level of significance (α).
If computed value of D is greater than or equal to critical value (D  ) ,
that is, D  D then we reject the null hypothesis at α level of
significance, otherwise we do not reject H0.
For large sample (n1 = n2 = n > 40 and n1 >16 for unequal):
For a sample size n1 = n2 = n > 40 and n1 >16 for unequal sample
sizes, the critical value of test statistic at given level of significance is
approximated by the formula given in last row of Table V & VI in
the Appendix.
Now, it is time to do some examples based on above test.
Example 7: The following data represent lifetime (in hours) of batteries for
two different brands A and B:
Brand A 40 30 55 40 40 35 30 40 50 45 40 35
Brand B 45 60 50 60 35 50 55 60 50 50 40 55

Examine by the Kolmogorov-Smirnov two-sample test, is the average life of

two brands same at 5% level of significance?

Solution: Here, we want to test that the average life of batteries of two brands
is same. If F1(x) and F2(x) are cumulative distribution functions of the life of
batteries of brand A and brand B respectively then our claim is F1(x) = F2(x)
and its complement is F1(x) ≠ F 2(x). Since the claim contains the equality sign
so we can take claim as the null hypothesis and complement as the alternative
hypothesis. Thus,
60
 H 0 : F1 (x)  F2 (x) for all values of x Two-Sample Tests

H1 : F1 (x)  F2 (x) for at least one value of x

So the test statistic is given by
D  sup S1  x   S2  x 
x
where, S1(x) and S2(x) are the empirical cumulative distribution functions of
the lives of the batteries of brand A and brand B drawn in the samples
respectively which are defined as:
number of batteries of brand A whose average life  x
S1  x  
n1
number of batteries of brand B whose average life  x
S2  x  
n2
Calculation for S1  x   S2  x  :
Average Frequency C.F. SA(x) Frequency C.F. SB(x) S1  x   S2  x 
Life (Brand A) Brand A (Brand B) Brand B

30 2 2 2/12 0 0 0 2/12
35 2 4 4/12 1 1 1/12 3/12
40 5 9 9/12 1 2 2/12 7/12
45 1 10 10/12 1 3 3/12 7/12
50 1 11 11/12 4 7 7/12 4/12
55 1 12 12/12 2 9 9/12 3/12
60 0 12 12/12 3 12 12/12 0
Total n1 = 12 n2 = 12
From the above calculation, we have
7
D  sup S1  x   S2  x    0.58
x 12
The critical value of test statistic for equal sample sizes n1 = n2 = 12 at 5%
level of significance is 6/12 = 0.5.
Since calculated value of test statistic (= 0.58) is greater than critical value
(= 0.5) so we reject the null hypothesis i.e. we reject the claim at 5% level of
significance.
Thus, we conclude that the samples provide us sufficient evidence against the
claim so the average life of batteries of two brands is different.
Example 8: The following are the marks in Statistics of B.Sc. students taken
randomly from two colleges A and B:
Marks 0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100
College (A) 2 2 4 6 3 3 4 8 7 5
College (B) 1 1 2 5 7 3 3 2 6 6

Apply Kolmogorov-Smirnov test to examine that the distribution of marks in

college A and college B is same at 1% level of significance.

Solution: Here, we wish to test that the distribution of marks in Statistics of

B.Sc. students in college A and college B is same. If F1(x) and F2(x) are
distribution functions of the marks in college A and college B respectively then

61
Non-Parametric Tests our claim is F1(x) = F2(x) and its complement is F1(x) ≠ F2(x). Since the claim
contains the equality sign so we can take claim as the null hypothesis and
complement as the alternative hypothesis. Thus,
H 0 : F1 (x)  F2 (x) for all values of x

H1 : F1 (x)  F2 (x) for at least one value of x

So the test statistic is given by
D  sup S1  x   S2  x 
x

where, S1(x) and S2(x) are the empirical distribution functions of samples
drawn from college A and college B respectively which are defined as:
number of students of college A whose marks  x
S1  x  
n1
number of students of college B whose marks  x
S2  x  
n2

Calculation for S1  x   S2  x  :
Marks Frequency C.F. SA(x) Frequency C. F. SB(x) S A x   S B x 
(College A) (College A) (College B) (College B)

0-10 2 2 0.0455 1 1 0.0278 0.0177

10-20 2 4 0.0909 1 2 0.0556 0.0354
20-30 4 8 0.1818 2 4 0.1111 0.0707
30-40 6 14 0.3182 5 9 0.2500 0.0682
40-50 3 17 0.3864 7 16 0.4444 0.0581
50-60 3 20 0.4545 3 19 0.5278 0.0732
60-70 4 24 0.5455 3 22 0.6111 0.0657
70-80 8 32 0.7273 2 24 0.6667 0.0606
80-90 7 39 0.8864 6 30 0.8333 0.0530
90-100 5 44 1 6 36 1 0
Total n1 = 44 n2 = 36

From the above calculation, we have

D  sup S1  x   S2  x 
x

 0.0732
Since n1 = 44 > 16 so we can calculate the critical value of test statistic for
unequal sample sizes n1 = 44 and n2 = 36 at 1% level of significance by the
formula
n1  n 2 44  36
D n,  1.63  1.63  0.366
n 1n 2 44  36
Since calculated value of test statistic D (= 0.0732) is less than critical value
(= 0.366) so we do not reject the null hypothesis i.e. we support the claim at
1% level of significance.

62
Thus, we conclude that the samples fail to provide us sufficient evidence Two-Sample Tests
against the claim so we may assume that the distribution of marks in Statistics
of B.Sc. students in college A and college B are same.
Now, you can try the following exercises.
E11) What is the main difference between Mann-Whitney U test and
Kolmogorov-Smirnov two-sample tests?
E12) The following are the grades in Mathematics of High school students
taken randomly from two School I and School II:
Grades A A+ B B+ C C+ D D+
Number of Students in School- I 2 2 4 6 3 3 0 0
Number of Students in School- II 1 1 2 3 4 2 4 3

Apply Kolmogorov -Smirnov test to examine that the performance in

Mathematics of High school students in School-I and School-II is same at
5% level of significance.
We now end this unit by giving a summary of what we have covered in it.

14.6 SUMMARY
In this unit, we have discussed following points:
1. Need of two-sample non-parametric tests.
2. Which non-parametric two-sample test is appropriate for a particular
situation.
3. The test which is used in place of paired t-test when assumption(s) of the
test is (are) not fulfilled, that is, paired sign test and Wilcoxon matched-
pair signed-rank tests.
4. The Mann-Whitney U test which is used when the assumption(s) of two
sample t-test is (are) not fulfilled.
5. The Kolmogorov-Smirnor two-sample test for testing the hypothesis that
two samples come from the populations having the same distribution.

14.7 SOLUTIONS / ANSWERS

E1) The main difference between paired t-test and paired sign test is that the
paired t-test based on normality assumption where as sign test is not
based on such assumption.
E2) Here, the given dada are in the form of before and after study but
assumption of normality is not given so we cannot use paired t-test. So
we go for paired sign test instead of paired t-test.
Here, we wish to test that IQ is improved due to course designed that
means average score before the course is less than average score after
the course. If  1 and  2 denote the average (median) score before and
after the course so our claim is  1   2 and its complement is  1   2 .
Since complement contains the equality sign so we can take the
complement as the null hypothesis and the claim as the alternative
hypothesis. Thus,
H0 :  1   2
63
Non-Parametric Tests H 1 :  1   2 course improves the IQ 
Since the alternative hypothesis is left-tailed so the test is left-tailed test
and the test statistic (S) is the number of plus signs (S+).
Calculation for S:
S. IQ Before IQ After Sign of Difference
No (X) (Y) (X – Y)
1 95 99 –
2 112 115 –
3 128 133 –
4 96 96 Tie
5 97 99 –
6 117 120 –
7 105 106 –
8 95 106 –
9 99 100 –
10 86 89 –
11 75 90 –
12 82 85 –

From the above calculation, we have

n = number of non-zero difference = 11
S+ = number of plus signs = 0
S− = number of minus signs = 11
Therefore, the value of the test statistic is
S = number of plus signs (S+) = 0
To take the decision about the null hypothesis, we determine p-value
with the help of Table I given in Appendix at the end of this block.
Here, n = 11, p = 0.5 and r = 0. Thus, we have
p-value  P S  0   0.0005
Since p-value = 0.0005< 0.05 (= α) so we reject the null hypothesis and
support the alternative hypothesis i.e. we support the claim at 5% level
of significance.
Thus, we conclude that the samples fail to provide us sufficient
evidence against the claim so we may assume that IQ of students is
improved due to course.
E3) Here, the ratings are given in the form of pairs so it can be considered
as the case of dependent samples. But assumption of normality is not
given also the given data are in the form of ranks so we cannot use the
paired t-test in this case. So we go for paired sign test.
Here, we want to test that the product A is favoured over product B that
means average (median) rating consumption of product A is greater
than B. If  1 and  2 denote the average (median) rating consumption of
product A and product B respectively then our claim is  1  2 and its
complement is  1   2 . Since complement contains the equality sign so
we can take the complement as the null hypothesis and the claim as the
alternative hypothesis. Thus,

64
 H0 :  1   2 Two-Sample Tests

H 1 :  1   2  the product A is favoured over product B 

Since the alternative hypothesis is right-tailed so the test is right-tailed
test so the test statistic (S) is the number of minus signs (S−).
Calculation for S:
S. No Favour of Product Favour of Product Sign of Difference
A (X) B (Y) ( X – Y)
1 5 2 +
2 2 3 –
3 4 1 +
4 4 2 +
5 5 3 +
6 4 4 Tie
7 5 4 +
8 5 1 +
9 2 4 –
10 3 2 +
11 3 2 +
12 4 3 +
13 3 3 Tie
14 3 4 –
15 2 3 –

From the above calculation, we have

n = number of non-zero differences = 13
S+ = number of plus signs = 9
S− = number of minus signs = 4
Therefore, value of test statistic is
S = number of minus signs (S−) = 4
To take the decision about the null hypothesis, we determine p-value
with the help of Table I given in Appendix at the end of this block.
Here, n = 13, p = 0.5 and r = 4. Thus, we have
p-value  P S  4   0.1334

Since p-value = 0.1334 > 0.01(= α) so we do not reject the null

hypothesis and reject the alternative hypothesis i.e. we reject the claim
at 1% level of significance.
Thus, we conclude that the samples provide us sufficient evidence
against the claim so product A is not favoured over product B.
E4) Wilcoxon matched-pair singed-rank test is based on the assumption that
the distribution of differences is symmetric whereas paired sign test
does not require this assumption.
E5) It is case of two dependent populations and the assumption of normality
for the differences is not given so we will not go for paired t-test. Also
measurement of observations is available in interval scale so we will go
for Wilcoxon matched pair signed-rank test.

65
Non-Parametric Tests Here, we wish to test that the students have benefited by coaching that
means average score before the coaching is less than average score after
the coaching. If  1 and  2 denote the average (median) score before and
after the coaching then our claim is  1   2 and its complement is
 1   2 . Since complement contains the equality sign so we can take the
complement as the null hypothesis and the claim as the alternative
hypothesis. Thus,
H0 :  1   2
H1 :  1   2 students have benefited by coaching 
Since the alternative hypothesis is left-tailed so the test is left-tailed
test.
Since in this exercise the increment in marks is given that means we are
given that the difference of after on before, that is, Y − X instead of
X – Y therefore, for testing the null hypothesis, the test statistic is the
sum of negative ranks (T −) instead of sum of positive ranks (T+).
Calculation for T:
Roll Difference Absolute Rank of Signed
No. d Difference d d Rank
1 6 6 6.5 6.5
2 −2 2 1.5 −1.5
3 8 8 8 8
4 −4 4 3.5 −3.5
5 10 10 9 9
6 2 2 1.5 1.5
7 5 5 5 5
8 −4 4 3.5 −3.5
9 6 6 6.5 6.5
10 Tie --- --- ---

From the above calculations, we have

T− = 8.5
n = number of non-zero di’s = 9
So the value of test statistic(T) is T− = 8.5
The critical (tabulated) value of test statistic for one-tailed test
corresponding to n = 9 at 5% level of significance is 9.
Since calculated value of test statistic T (= 8.5) is less than the critical
value (= 9) so we reject the null hypothesis and support the alternative
hypothesis i.e. we support the claim at 5% level of significance.
Thus, we conclude that the sample fails to provide us sufficient
evidence against the claim so we may assume that the coaching is
effective.
E6) Here, the data are given in the form if pairs so it can be considered as
the case of dependent samples. But assumption of normality is not
given also the given data are in the form of ranks so we cannot use the
paired t-test in this case. So we go for Wilcoxon matched-pair signed-
rank test.
66
 Here, we wish to test that the product A and B are equally soft. If Two-Sample Tests
 1 and  2 denote the average rank of softness of products A and B
respectively then our claim is  1   2 and it complement is  1   2 .
Since claim contains the equality sign so we can take the claim as the
null hypothesis and the complement as the alternative hypothesis. Thus,
H 0 :  1   2 and H1 :  1   2
Since the alternative hypothesis is two-tailed so the test is two-tailed
test so the test statistic is the minimum of the sum of positive ranks(T+)
and negative ranks (T−), that is,
T  min T  , T  
Calculation for T:
Judge Product Product Difference Absolute Rank of Signed
A B (d) Difference d d Rank

1 4 5 –1 1 2 –2
2 6 4 2 2 5 5
3 8 5 3 3 7.5 7.5
4 9 8 1 1 2 2
5 4 1 3 3 7.5 7.5
6 7 9 –2 2 5 –5
7 6 2 4 4 9 9
8 5 3 2 2 5 5
9 8 2 6 6 10 10
10 6 7 –1 1 2 –2

From the above calculations, we have

T   46 and T   9
n = number of non-zero di’s = 10
Putting the values in test statistic, we have
T  min T ,T   min 46,9  9

The critical value of test statistic for two-tailed test corresponding

n = 10 at 1% level of significance is 4.
Since calculated value of test statistic T (= 9) is greater than the critical
value (= 4) so we do not reject the null hypothesis i.e. we support the
claim at 1% level of significance.
Thus, we conclude that the sample fail to provide us sufficient evidence
against the claim so we may assume that the products A and B are
equally soft.
E7) Two-sample t-test is based on the assumptions that populations under
study are normally distributed and the observations are measured in at
least of an interval scale whereas Mann-Whitney U test is non required
such assumptions.
E8) Wilcoxon matched-pair singed-rank test is used when observations are
paired whereas the Mann-Whitney U test is used when two independent
samples are randomly drawn two populations.

67
Non-Parametric Tests E9) It is case of two independent populations and the assumption of
normality of both the populations is not given so we cannot use t-test in
this case. So we go for Mann-Whitney U test.
Here, we want to test that village boys are more fit than town boys that
means average (median) score of village boys is greater than average
(median) score of town boys. If µ 1 and µ 2 denote the average (median)
score of the village boys and town boys respectively then our claim is
 1  2 and its complement is  1   2 . Since complement contains the
equality sign so we can take the complement as the null hypothesis and
the claim as the alternative hypothesis. Thus,
H0 :  1   2
H 1 :  1   2  village boys are more fit than town boys 
To perform the Mann-Whitney U test, we combine the observations of
both samples and for convenient, we arrange the observations in
ascending order. Then rank all these combined observations from
smallest to largest as:
Village Boys(V) Ranks Town Boys (T) Ranks
4.5 5 2.7 1
5.6 6.5 2.8 2
6.5 9.5 3.2 3
7.2 12 3.6 4
8.2 15 5.6 6.5
9.0 16 6.1 8
10.6 17 6.5 9.5
11.4 18 6.7 11
12.4 20 7.9 13.5
12.9 23 7.9 13.5
15.7 23 11.8 19
16.2 25 12.6 21
12.7 22
Total 190

Here, n1 =12 and n2 = 13. Since n1 is less than n2 therefore we take sum
of ranks corresponding to village boys.
Since we consider the first sample so test statistic U is given by
n1  n 1  1 12  12  1
U S  190   112
2 2
The upper critical value of test statistic for right-tailed corresponding to
n1 = 12 and n2 = 13 at 5% level of significance is U U,   U U, 0.01  198.
Since calculated value of test statistic (=112) is less than upper critical
value (= 198) so we do not reject the null hypothesis and reject the
alternative hypothesis i.e. we support the claim at 5% level of
significance.
Thus, we conclude that the samples provide us sufficient evidence
against the claim so village boys are not more fit than town boys.

68
E10) It is case of two independent populations and the assumption of Two-Sample Tests
normality of both the populations is not given so we cannot use two
sample t-test in this case. So we go for Mann-Whitney U test.
Here, we want to test that average life of two brands of batteries is
same. If µ 1 and µ 2 denote the average (median) life of batteries of
brands A and B respectively then our claim is  1   2 and it
complement is  1   2 . Since claim contains the equality sign so we can
take the claim as the null hypothesis and the complement as the
alternative hypothesis. Thus
H 0 :  1   2 average life of two brands of batteries is same 
H1 :  1   2 average life of two brands of batteries is not same 
To perform the Mann-Whitney U test, we combine the observations of
both samples and for convenient, we arrange the observations in
ascending order. Then rank all these combined observations from
smallest to largest as:
Brand A Ranks Brand B Ranks
30 1.5 35 4
30 1.5 40 8.5
35 4 45 12.5
35 4 50 16
40 8.5 50 16
40 8.5 50 16
40 8.5 50 16
40 8.5 55 20
40 8.5 55 20
45 12.5 60 23
50 16 60 23
55 20 60 23
Total 102

Here, n1 = n2 so we can take any the sum of ranks(S) of first sample or

second sample. We consider first sample. Therefore, the test statistic U
is given by
n  n  1 12  12  1
U S 1 1  102   24
2 2
The critical values of test statistic for two-tailed test corresponding
n1= n2 = 12 at 1% level of significance are U L,  / 2  U L, 0.005  106 and
U U,  / 2  U U, 0.005  194.
Since calculated value of test statistic U (= 24) is less than both critical
values (= 106 and 194) so we reject the null hypothesis i.e. we reject the
claim at 1% level of significance.
Thus, we conclude that the samples provide us sufficient evidence
against the claim so average life of two brands of batteries is not same.
E11) Mann-Whitney U test is used to test the hypothesis that two
independent samples come from the populations follow distributions
that are identical with respect to location whereas Kolmogorov-
Smirnov two-sample test is sensitive to differences of all types that may

69
Non-Parametric Tests exist between two distributions that is, location, dispersion, skewness,
etc.

E12) Here, we want to test that performance of students in School-I and

School-II are same. If F1(x) and F2(x) are distribution functions of
grades in Mathematics of High school students in School-I and
School-II then our claim is F1(x) = F2(x) and its complement is
F1(x) ≠ F 2(x). Since the claim contains the equality sign so we can take
claim as the null hypothesis and complement as the alternative
hypothesis. Thus,
H0 : F1 (x)  F2 (x) for all values of x
H1 : F1 (x)  F2 (x) for at least one value of x
For testing the null hypothesis, the test statistic is given by
D  sup S1  x   S2  x 
x
where, S1(x) and S2(x) are the empirical cumulative distribution
functions of samples drawn from School-I and School-II respectively
which are defined as:
number of students of School-I whose grade  x
S1  x  
n1
and
number of students of School-II whose grade  x
S2  x  
n2
Calculation for S1  x   S 2  x  :
Grade Frequency C.F. SA(x) Frequency C. F. SB(x) S A x   S B x 
(School-I) (School-I) (School-II) (School-II)

A 2 2 0.10 1 1 0.05 0.05

A+ 2 4 0.20 1 2 0.10 0.10
B 4 8 0.40 2 4 0.2 0.20
B+ 6 14 0.70 3 7 0.35 0.35
C 3 17 0.85 4 11 0.55 0.30
C+ 3 20 1.00 2 13 0.65 0.35
D 0 20 1.00 4 17 0.85 0.15
D+ 0 20 1.00 3 20 1 0.00
Total n1 = 20 n2 = 20

From the above calculation, we have

D  sup S1  x   S2  x   0.35
x

The critical value of test statistic for equal sample sizes n1 = n2 = 20 at

5% level of significance is 8/20 = 0.4.
Since calculated value of test statistic (= 0.35) is less than critical value
(= 0.4) so we do not reject the null hypothesis i.e. we support the claim
at 5% level of significance.
Thus, we conclude that the samples fail to provide us sufficient
evidence against the claim so we may assume that the performance of
students in School-I and School-II is same.

70
UNIT 15 k-SAMPLE TESTS
Structure
15.1 Introduction
Objectives
15.2 Kruskal-Wallis Test
15.3 Friedman Test
15.4 Summary
15.5 Solutions /Answers

15.1 INTRODUCTION
In Unit 13 of this block, we have studied some one-sample non-parametric tests
for testing the hypothesis about the population median and in Unit 14, we have
studied some two-sample non-parametric tests for testing the hypothesis about
the equality of two population medians. If there is a situation where three or
more populations are under study and we are interested to test the equality of
their mean or medians. In such situations, generally, we use analysis of
variance (ANOVA) technique that will describe in Block 2 of MST-005. But
ANOVA procedure requires the assumption that the populations being
compared are all normally distributed with equal variances. When the
populations under study are not normally distributed then we cannot use the
ANOVA procedure. Therefore, we require non-parametric tests that not based
on any assumption about the shape (form) of the population distributions. This
unit is devoted to discuss some such tests that use in these situations as
Kruskal-Wallis test and Friedman test.
This unit is divided into five sections. Section 15.1 is described the need of
k-sample non-parametric tests. The Kruskal-Wallis test is discussed in Section
15.2 in details whereas Friedman test is discussed in Section 15.3. Unit ends by
providing summary of what we have discussed in this unit in Section 15.4 and
solution of exercises in Section 15.5.
Objectives
After studying this unit, you should be able to:
 describe the need of k-sample test;
 perform the Kruskal-Wallis test for testing the hypothesis about equality of
more than two population medians; and
 perform the Friedman test.

15.3 KRUSKAL-WALLIS TEST

Kruskal-Wallis test is the
As we have discussed in previous section that analysis of variance (ANOVA) non-parametric alternative to
test whether three or more independent samples have been drawn from the the one way analysis of
variance or completely
populations having the same mean or not. Also it is considered as the extension randomised design. Also this
of two-sample t-test for equality of two means. But ANOVA procedure test represents a
requires the assumption that the populations being compared are all normally generalisation of the two-
distributed with equal variances. When the populations under study are not sample Mann-Whitney U
normally distributed then we cannot use the ANOVA procedure. Therefore, we test.
require a non-parametric test that not based on any assumption about the shape
71
Non-Parametric Tests of the population distribution and use in testing of such situations. Such a test is
the Kruskal-Wallis test. This test is a non-parametric alternative to the one-way
analysis of variance or completely randomised design. Kruskal-Wallis test also
a generalisation of the two-sample Mann-Whitney U test.
Assumptions
This test works under the following assumptions:
(i) k samples are randomly and independently drawn from their respective
populations.
(ii) The variable under study is continuous.
(iii) The measurement scale is at least ordinal.
(iv) The populations are identical in shape except for a possible difference in
location.
Let us suppose that we have k independent random samples of sizes n1, n2,...,
nk taken from k populations respectively which are shown as below:
Sample (Treatment) Sample Observation
Number
1 X11 X12 … X1n1
2 X21 X 22 … X 2n 2
. . . … .
. . . … .
. . . … .
k Xk1 Xk2 … X knk

Generally, we are interested to test whether all k populations have same median
or not so we can take the null and alternative hypotheses as
H 0 : All k populations have same median
H1 : All k populations have not same median

A treatment is an object or In ANOVA generally, we test the effect of treatments so we can also take the
a factor which is allocated null and alternative hypotheses as
to one or more units to
estimate its effect. For H 0 : All k treatments have same effect
example, if some fertilizers H1 : All k treatments have not same effect
are applied on experimental
fields to know the effect to After setting the null and alternative hypotheses, this test involves following
these then fertilizers are steps:
known as treatments.
Step 1: First of all, we combine the observations of all samples.
Step 2: After that, ranking all these combined observations from smallest to
largest, that is, the rank 1 is given to the smallest of the combined
observations, rank 2 is given to the second smallest and so on up to
the largest observation. If several values are same (tied), we assign
each the average of ranks they would have received if there were no
repetition.
Step 3: If the null hypothesis is true, we expect that sum of ranks of all
samples to be equal. Therefore, in this step we find the sum of ranks
for each sample. Let R1, R2, ..., Rk represent the sum of ranks of 1st
sample, 2 nd sample and so on kth sample respectively.

72
Step 4: To test the null hypothesis, the Kruskal-Wallis test statistic is defined k-Sample Tests
as:
12  k R i2 
H    3n  1
n n  1  i1 n i 
where, n = n1 + n2 + ... + nk
If tie occurs, an adjustment in Kruskal-Wallis test statistic has to be
made. The adjustment factor is
r
1
C  1   t 3i  t i 
 n  n  i1
3

where, r is the number of groups of different tied ranks, and ti is the

number of tied values within ith group that are tied at a particular
value. Therefore, adjusted Kruskal-Wallis test statistic is
HC  H / C
Step 5: Obtain critical value of Kruskal-Wallis test statistic at given level of
significance under the condition, that null hypothesis is true. Table
VIII in the Appendix at the end of this block provides critical values
for three samples (k = 3) and each sample has five or fewer
observations (ni ≤ 5) at α level of significance.
Step 6: Decision rule:
To take the decision about the null hypothesis, the calculated value of
Kruskal-Wallis test statistic (computed in Step 4) is compared with
the critical (tabulated) value (obtain in Step 5) at given level of
significance(α) under the condition that null hypothesis is true.
If calculated value of test statistic is greater than critical (tabulated)
value at α level of significance then we reject the null hypothesis at α
level of significance otherwise we do not reject the null hypothesis
H0.
For large samples
For large number of samples (k > 3) or for large sample sizes (ni > 5),
the distribution of the test statistic under the null hypothesis is well
approximated by the chi-square distribution with (k –1) degrees of
freedom. So when number of samples or observations per sample are
such that we cannot use Table VIII then we use χ2-table (given in the
Appendix of the Block 1 of this course) for critical value of the test
statistic.
Now, it is time to do some examples based on above test.
Example 1: Three different feeds are to be compared to determine that they
have same distribution of weight gains on experimental animals (such as pig).
Suppose 12 animals are divided at random into three groups of four animals
and each group at different feed. The following results are obtained:
Feed Weight Gains
1 104 110 106 102
2 112 117 115 114
3 120 126 121 128

Test that the median weight gains due to three feeds are same by Kruskal-
Wallis test at 5% level of significance.
73
Non-Parametric Tests Solution: Here, we want to test that the median weight gains due to three feeds
are same. Therefore, our claim is “the median weight gains due to three feeds
are same” and its complement is “the median weight gains due to three feeds
are not same”. So we can take claim as the null hypothesis and complement as
the alternative hypothesis. Thus,
H1 : The median weight gains due to three feeds are same
H1 : The median weight gains due to three feeds are not same
Here, the distributions of weight gains due to different feeds are not given so
the assumption of normality for one-way ANOVA is not fulfilled. Therefore,
we go for Kruskal-Wallis test.
To perform the Kruskal-Wallis test, the test statistic is given by
12 k
 R i2 
H     3  n  1 … (1)
n(n  1) i 1  n i 

where, Ri is the sum of ranks of ith group.

Therefore, first we rank all the observations of three groups and then find the
sum of ranks (Ri) of each group.
Calculation for Ri:
Weight Gains
Feed I Rank Feed II Rank Feed III Rank
104 2 112 5 120 9
110 4 117 8 126 11
106 3 115 7 121 10
102 1 114 6 128 12
Total R1 = 10 R2 = 26 R3 = 42

Putting the values in equation (1), we have

12  10 2  26  2  42 2 
H      3(12  1)
12(12  1)  4 4 4 

12 100 676 1764 

    39
12  13  4 4 4 
 48.85  39  9.85
The critical (tabulated) value corresponding to k = 3, n1 = 4, n2 = 4, n3 = 4 at 5%
level of significance is 5.692.
Since calculated value of test statistic H (= 9.85) is greater than critical value
(= 5.692) so we reject the null hypothesis i.e. we reject the claim at 5% level of
significance.
Thus, we conclude that samples provide us sufficient evidence against the
claim so the median weight gains due to feeds are not same.
Example 2: An analyst in the publishing industry wants to test whether the
price of newspaper advertisement of a given size is about the same in four large
newspaper groups. Random sample of seven newspapers from each group is

74
selected and the price (in thousand rupees) of an ad is recorded which are given k-Sample Tests
below:
Group A: 65 62 45 70 48 57 50
Group B: 81 64 72 55 90 68 75
Group C: 42 58 48 59 61 60 64
Group D: 85 92 69 82 94 64 73
Do you believe that there is significant difference in the price of an ad across
the four groups by Kruskal-Wallis test at 1% level of significance?
Solution: Here, we want to test that price of an ad across the four groups is
different. Therefore, our claim is “the price of an ad across the four groups is
not same” and its complement is “the price of an ad across the four groups is
same”. Therefore, we can take complement as the null hypothesis and claim as
the alternative hypothesis. Thus,
H0 : The price of an ad across the four newspapers is same
H1 : The price of an ad across the four newspapers is not same
Here the assumption of normality is not fulfilled, so we cannot use one-way
ANOVA. Therefore, we go for Kruskal-Wallis test.
To perform the Kruskal-Wallis test, the test statistic is given by
12 k
 R i2 
H     3n  1
n (n  1) i1  n i 
… (2)

where, Ri is the sum of ranks of ith group.

Therefore, first we rank all the observations of three groups. For convenient,
we arrange the data of each group in ascending order and then find the sum of
ranks (Ri) of each group.
Calculation for Ri:
Group Rank Group Rank Group Rank Group Rank
A B C D
45 2 55 6 42 1 64 14
48 3.5 64 14 48 3.5 69 18
50 5 68 17 58 8 73 21
57 7 72 20 59 9 82 24
62 12 75 22 60 10 85 25
65 16 81 23 61 11 92 27
70 19 90 26 64 14 94 28
Total R1 = 64.5 R2 =128 R3 =56.5 R4 =157

Putting the values in equation (2), we have

12   64.5 2 128  2  56.52 157 2 
H       3(28  1)
28(28  1)  7 7 7 7 

12  4160.25  16384  3192.25  24649 

    87
28  29  7 
12
  6912.21  87  102.15  87  15.15
28  29
From the above calculations, we see that there are tied ranks therefore, we have
Number of groups of tied ranks = r = 2 (3.4 and 14)
75
Non-Parametric Tests Number of times group 3.4 of tied ranks occurs = t1(3.4) = 2
Number of times group 14 of tied ranks occurs = t2(14) = 3
Since ties occur, therefore, the correction factor is
r
1 1
C 1   t 3i  t i   1    2 3  2    33  3
  i1  28  28  
3 3
n  n
1
 1  30  1  0.0013  0.9987
21924
Therefore, adjusted Kruskal-Wallis test statistic is
H C  H / C  15.15 / 0.9987  15.17
Since all sample sizes n1 = 7, n2 = 7, n3 = 7 exceed 5 therefore, the test statistic
Hc is approximated by chi square with k − 1 = 4 − 1 = 3 degrees of freedom.
Thus, we compare calculated test statistic Hc with critical value given in
χ2-table.
The critical (tabulated) value of 2 with 3 degrees of freedom at 1% level of
significance is 11.34.
Since calculated value of test statistic Hc (= 15.17) is greater than critical value
(= 11.34) so we reject the null hypothesis and support the alternative
hypothesis i.e. we support the claim at 1% level of significance.
Thus, we conclude that the samples fail to provide us sufficient evidence
against the claim so we may assume that the price of an ad across the four
newspapers is different.
Now, you can try the following exercises in same manner.
E1) Suppose dairy farmer wishes to compare the effect of four different diets
on the amount of milk produced by cattle. He randomly selects 12 cows
and assigns four cows randomly to each diet for eight week. The
following table gives the rank assigned to each cows.
Diet Rank Total Ri
A 2 5 4 9 20
B 7 1 8 3 19
C 6 10 12 11 39

Test that the diets are equally productive by Kruskal-Wallis test at 1%

level of significance.
E2) A manufacturer tried three alternative manufacturing processes and
recorded the number of defective articles in each process as follows:
Number of Defectives per Batch
Process I Process II Process III
12 10 13
5 15 11
2 2 6
6 7 16
0 8 14
4 6 20
Assuming that same number of articles is produced in each batch. Can
you conclude that there is no difference in the three processes in terms of
number of defective per batch by Kruskal-Wallis test at the 1% level of
significance?

76
 k-Sample Tests
15.4 FRIEDMAN TEST
In the previous section, we have studied Kruskal-Wallis test which is a non-
parametric version of one-way ANOVA or completed randomised design and Kruskal-Wallis test is a non-
parametric version of one
also an extension of Mann-Whitney U test. The Friedman test is a non- way ANOVA or completely
parametric version of two-way ANOVA or randomised block design with one randomised design whereas
item per call that will be discussed in Block 2 of MST-005. This test is also the Friedman test is a non-
based on ranks and may be viewed as an extension of the Wilcoxon signed- parametric version of the
rank test for two samples. randomised block design or
two way ANOVA with one
Assumptions item per call.

This test work under the following assumptions:

(i) The observations are independent within and between samples.
(ii) The treatments are randomly assigned to experimental units within each
block.
(iii) The measurement scale is at least ordinal.
(iv) The variable under study is continuous.
Suppose, we have k mutually independent random samples of equal size n, that
is Xi1, X i2 ,..., X in where, i = 1, 2, …, k. In other words, we can say that there are
n treatments and k blocks of each of size n and each treatment occurs once in
each block. The data for k blocks and n treatments can be presented in the
following two-way table:
Block Treatment
(Sample)
1 2 … j … n
1 X11 X12 … X1j … X1n
2 X21 X22 … X2j … X2n
. . . … . … .
. . . … . … .
. . . … . … .
i Xi1 Xi2 … Xij … Xin
. . . … . … .
. . . … . … .
. . . … . … .
k Xk1 Xk2 … Xkj … Xkn

Generally, we are interested to test whether all treatments have same effect or
not so we can take the null and alternative hypotheses as
H 0 : All treatments have same effect
H1 : All treatments have same not effect
After setting null and alternative hypotheses, this test involves following steps:
Step 1: To perform the Friedman test, we rank the observations within each In Kruskal-walls test, we
block (sample) independently from smallest to largest. The rank 1 is rank all observations which
given to the smallest observation of a block, rank 2 is given to the are combined from all
second smallest and so on up to the largest observation of the block samples whereas in
Friedman test we rank the
(sample). If several values are same (tied) we assign each the average observations within each
of ranks they would have received if there were no repetition. Then sample or block.
we replace each observation of the block (sample) by corresponding
rank. This process is repeated for each block (sample).
77
Non-Parametric Tests Step 2: After that, we sum all the ranks for each treatment. Let R1 be the sum
of all ranks for 1st treatment, R 2 be the sum of all ranks for 2nd
treatment and so on, R n be the sum of all ranks for nth treatment.
Step 3: If null hypothesis is true, that is, all treatments have same effects then
we expect that the sum of the ranks for each treatment would be
approximately equal. The sum of squares of differences among the
sums of the ranks will be indicative of the difference in the treatment
effects. These sum of squares are measured by the Friedman and give
in the simple form which is known as Friedman test statistic (F) and
given by
n
12
F  R 2j  3k(n  1)
kn(n  1) j1
The distribution of Friedman test statistic F is approximated by
chi-square distribution with (n – 1) degrees of freedom. In order to the
reasonably good approximation, we require that either number of
blocks (k) or number of treatments (n) exceed 5.
If tie occurs within blocks, an adjustment in Friedman F-statistic has
to be made. The adjustment factor is
k
1
C  1 
k  n  1 i1
2  t 3i  t i 

where t i is the number of tied observations in the ith block (sample).

Therefore, adjusted Friedman statistic is
FC  F / C
Step 4: Obtain critical value of Friedman test statistic at given level of
significance under the condition that null hypothesis is true. Table IX
in the Appendix at the end of this block provides critical values for 3
to 6 (k = 3 to 6) samples and each sample has n = 20 or fewer
observations at α level of significance.
Step 5: Decision rule:
To take the decision about the null hypothesis, the calculated value of
Friedman test statistic (computed in Step 3) is compared with the
critical (tabulated) value (obtain in Step 4) at a given level of
significance(α) under the condition that null hypothesis is true.
If calculated value of test statistic is greater than critical (tabulated)
value at α level of significance then we reject the null hypothesis at α
level of significance otherwise we do not reject the null hypothesis
H0.
For large samples
For large number of treatments or for large sample sizes the distribution
of the test statistic under the null hypothesis is well approximated by
the chi-square distribution with (n –1) degrees of freedom. So when
number of samples and observations per sample are such that we cannot
use Table IX then we use χ2-table (given in the Appendix of the Block
1 of this course) for critical values of the test statistic.
78
Now, it is time to do some examples based on Friedman test. k-Sample Tests

Example 3: The table given below shows reaction time (in second) data from 6
objects (treatments) each of which was tested under three conditions I, II and
III:
Object A B C D E F

Condition
I 356 590 665 450 540 500
II 401 564 667 560 570 525
III 455 570 650 575 560 550

Apply the Friedman test to test, whether reaction time of each object is same or
not at 5% level of significance?
Solution: Here, we want to test that reaction time of each object is same.
Therefore, our claim is “reaction time of each object is same” and its
complement is “reaction time of each object is not same”. So we can take claim
as the null hypothesis and complement as the alternative hypothesis. Thus,
H 0 : Reaction time of each object is same
H1 : Reaction time of each object is not same
Here, two factors (object and condition) are under study so we can use two-
way ANOVA if the distributions of both factors follow normal distributions.
But it is not the case so we go for Friedman test.
For testing the null hypothesis, we have Friedman test statistic as
n
12
F  R 2j  3k(n  1) … (3)
kn(n  1) j1
To perform the Friedman test, we rank the observations within each condition
(sample or block) and then calculate the sum of ranks of each object
(treatment) as:
Condition Object
A B C D E F
I 1 5 6 2 4 3
II 1 4 6 3 5 2
III 1 4 6 5 3 2
Sum 3 13 18 10 12 7

Here, k = 3 and n = 6, therefore, by putting the values in equation (3), we have

12
F  32  132  182  102  12 2  7 2   3  3   6  1
3  6  (6  1)
2
  795  63  75.71  63  12.71
21
The critical (tabulated) value of test statistic F corresponding k = 3 and n = 6 at
5% level of significance is 7.0.
Since calculated value of test statistic F (= 12.71) is greater than critical value
(= 7.0) so we reject the null hypothesis i.e. we reject the claim at 5% level of
significance.
79
Non-Parametric Tests Thus, we conclude that the samples provide us sufficient evidence against the
claim so reaction time of each object is not same.
Example 4: An experiment is conducted to investigate the toxic effect of three
chemicals A, B and C on the skin of rats. Three adjacent 1-inch squares are
marked on the backs of eight rats and each of the three chemicals is applied to
each rat. The squares of skin are the scored from 0 to 10 depending on the
degree of imitation. The data are given in the following table:
Chemical A B C
Rat

1 6 5 3
2 9 8 4
3 6 9 6
4 5 8 6
5 7 8 9
6 5 7 6
7 6 7 5
8 6 5 7

Test there is sufficient evidence that effect of all chemicals is same by the
Freidman test at 1% level of significance.
Solution: Here, we want to test that effect of all chemicals is same. So our
claim is “effect of all chemicals is same” and its complement is “effect of all
chemicals is not same”. So we can take claim as the null hypothesis and
complement as the alternative hypothesis. Thus,
H 0 : Effect of all chemicals is same

H1 : Effect of all chemicals is not same

Here, two factors (chemical and rat) are under study, so we can use two-way
ANOVA if the distributions of both factors follow normal distributions. But it
is not the case, so we go for Friedman test.
For testing the null hypothesis, we have Friedman test statistic as
n
12
F  R 2j  3k(n  1) … (4)
kn(n  1) j1
To perform the Friedman test, we rank the observations within each rat (block)
and then calculate the sum of ranks of each chemical (treatment) as:
Chemical A B C
Rat

1 3 2 1
2 3 2 1
3 1.5 3 1.5
4 1 3 2
5 1 2 3
6 1 3 2
7 2 3 1
8 2 1 3
Total 14.5 19 14.5

80
Here, k = 8 and n = 3, therefore, by putting the values in equation (4), we have k-Sample Tests

12 14.5 2  19  2  14.5 2   3  8   3  1

F
8  3  (3  1)  

1
 210.25  361  210.25  96
8
 97.69  96  1.69
Since tie occurs within 3 rd rat (block) therefore, an adjustment in Friedman F
statistic has to be made. The adjustment factor is
k
1
C  1   t3i  t i 
k  n 2  1 i1

where, t i is the number of tied observations in the ith rat (sample or block).
Here, t1 = 2 therefore,

C  1
 2  2   1  0.094  0.91
3

8  3  1 2

Adjusted Friedman statistic is

FC  F / C  1.69 / 0.91  1.86
Since number of rats (samples or blocks) k = 8 > 6 so the distribution of the
test statistic under the null hypothesis is well approximated by the chi-square
distribution with n –1 = 2 – 1 = 1 degree of freedom. Thus, we compare
calculated test statistic Fc with critical value given in χ2-table.
The critical (tabulated) value of 2 with 1 degree of freedom at 1% level of
significance is 6.63.
Since calculated value of test statistic Fc (= 1.86) is less than critical value
(= 6.63) so we do not reject the null hypothesis i.e. we support the claim at 1%
level of significance.
Thus, we conclude that samples do not provide us sufficient evidence against
the claim so we may assume that effect of all chemicals on rats is same.
Now, you can try the following exercises.
E3) An experiment was conducted using randomised block design with four
treatments and six blocks. The ranks of the measurements within each
block are shown in the following table:
Block
1 2 3 4 5 6
Treatment
A 3 3 2 3 2 3
B 1 1 1 2 1 1
C 4 4 3 4 4 4
D 2 2 4 1 3 2

Use the Friedman test for a randomised block design to test that effect of
all treatments is same at 5% level of significance.

81
Non-Parametric Tests E4) A random sample of 10 consumers asked to rank their preferences of four
new smells that a perfume manufacturer wants to introduce to the market
in the coming fall. The data are as follows where best is ranked by 1,
second best by 2 and worst by 4.
Respondent Smell I Smell II Smell III Smell IV
1 1 2 4 3
2 1 3 4 2
3 1 3 2 4
4 2 1 3 4
5 1 3 4 2
6 1 3 2 4
7 2 1 4 3
8 1 3 4 2
9 1 3 2 4
10 1 4 3 2

Applying Freidman test, do you believe that all four smells are equally
liked at 1% level of significance?
We now end this unit by giving a summary of what we have covered in it.

15.4 SUMMARY
In this unit, we have discussed following points:
1. Need of k-sample non-parametric tests.
2. The Kruskal-Wallis test for testing the hypothesis about equality of more
than two population medians and is known as the non-parametric version of
one-way ANOVA or completely randomised design.
3. Friedman test which is known as the non-parametric version of for
randomised block design.

15.5 SOLUTIONS / ANSWERS

E1) Here, we want to test that diets are equally productive of milk. So our
claim is “the diets are equally productive of milk” and its complement
is “diets are not equally productive of milk”. So we can take claim as
the null hypothesis and complement as the alternative hypothesis. Thus,
H0 : The diets are equally productive of milk
H1 : The diets are not equally productive of milk
Here, the assumption of normality is not fulfilled and the data are given
in the form of ranks so we cannot use one-way ANOVA. Therefore, we
go for Kruskal-Wallis test.
To perform the Kruskal-Wallis test, the test statistic is given by
12 k
 R i2 
H     3n  1
n (n  1) i1  n i 
… (5)

where, Ri is the sum of ranks of ith feed.

Since data are given in the form of ranks so first we find the sum of
rank (Ri) of each feed.

82
 Calculation for Ri: k-Sample Tests

Feed A Feed B Feed C

2 7 6
5 1 10
4 8 12
9 3 11
R1 = 20 R1 =19 R1 = 39

Putting the values in equation (5), we have

12   20  2 19  2  39  2 
H      3(12  1)
12(12  1)  4 4 4 

1  400 361 1521 

      39  43.88  39  4.88
13  4 4 4 
The critical (tabulated) value corresponding k = 3, n1 = 4, n2 = 4, n3 = 4
at 1% level of significance is 7.654.
Since calculated value of test statistic H (= 4.88) is less than critical
value (= 7.654) so we do not reject the null hypothesis i.e. we support
the claim at 1% level of significance.
Thus, we conclude that the samples do not provide us sufficient
evidence against the claim so we may assume that the diets are equally
productive of milk.
E2) Here, we want to test that there is no difference in the three processes in
terms of number of defective articles per batch, that is, three
manufacturing processes are identical. Therefore, our claim is “three
manufacturing processes are identical” and its complement is “three
manufacturing processes are not identical”. So we can take claim as the
null hypothesis and complement as the alternative hypothesis. Thus,
H0 : The three manufacturing processes are identical
H1 : The three manufacturing processes are not identical
Here, the assumption of normality is not fulfilled so we cannot use one-
way ANOVA. Therefore, we go for Kruskal-Wallis test.
To perform the Kruskal-Wallis test, the test statistic is given by
12 k
 R i2 
H     3n  1
n (n  1) i1  n i 
… (6)

where, Ri is the sum of ranks of ith processes.

Therefore, we first rank all the number of defectives articals
manufactured by three processes and then find the sum of ranks (Ri) of
each process.
Calculation for Ri:
Process I Rank Process II Rank Process III Rank
12 13 10 11 13 14
5 5 15 16 11 12
2 2.5 2 2.5 6 7
6 7 7 9 16 17
0 1 8 10 14 15
4 4 6 7 20 18
Total R1 = 32.5 R2 = 55.5 R3 = 83

83
Non-Parametric Tests Putting the values in equation (6), we have
12   32.5 2  55.5 2  83 2 
H      3(18  1)
18(18  1)  6 6 6 

12  1056.25  3080.25  6889 

    57
18 19  6 
12 11025.5
   57  64.48  57  7.48
18  19 6
From the above calculations, we see that there are tied ranks therefore,
we have
Number of groups of tied ranks = r = 2 (2.5 and 7)
Number of times group 2.5 of tied ranks occurs = t1(2.5) = 2
Number of times group 7 of tied ranks occurs = t2(7) = 3
Since ties occur so the correction factor is given by
r
1
C  1
n 3  n   t 3i  ti 
i 1

1   23  2    32  3 
 1
18   18  
3

1
 1  30  0.995
5814
Therefore, adjusted Kruskal-Wallis test statistic is
H C  H / C  7.48 / 0.995  7.52

Since all sample sizes exceed 5, therefore, the test statistic Hc is

approximated by chi square with k−1 = 3−1 = 2 degrees of freedom.
Thus, we compare calculated test statistic Hc with critical value given in
χ2-table with 2 degrees of freedom and 1% level of significance.
The critical (tabulated) value of 2 with 2 degrees of freedom at 1%
level of significance is 9.21.
Since calculated value of test statistic Hc (= 7.52) is less than critical
value (= 9.21) so we do not reject the null hypothesis i.e. we support the
claim at 1% level of significance.
Thus, we conclude that the samples fail to provide us sufficient
evidence against the claim so we may assume that three manufacturing
process are identical.
E3) Here, we want to test that the effect of four treatments is same.
Therefore, our claim is “effect of four treatments is same” and its
complement is “effect of four treatments is not same”. So we can take
claim as the null hypothesis and complement as the alternative
hypothesis. Thus,
H 0 : Effect of four treatments is same

H1 : Effect of four treatments is not same

84
 Here, two factors (treatment and block) are under study so we can use k-Sample Tests
two-way ANOVA if the distributions of both factors follow normal
distributions but it is not the case. Also data are given in the form of
ranks so we go for Friedman test.
For testing the null hypothesis, we have Friedman test statistic as
n
12
F  R 2j  3k(n  1)
kn(n  1) j1
… (7)

In this exercise, ranks are given therefore to perform the Friedman test
we calculate sum of ranks of each treatment as:
Block 1 2 3 4 5 6 Total
Treatment

A 3 3 2 3 2 3 16
B 1 1 1 2 1 1 7
C 4 4 3 4 4 4 23
D 2 2 4 1 3 2 14

Here k = 6 and n = 4, therefore, by putting the values in equation (7),

we have
12
F
6  4  (4  1)
16 2  7 2  232  14 2   3  6   4  1

1
  256  49  529  196   90  13
10
The critical value of test statistic F corresponding k = 6 and n = 4 at 5%
level of significance is 10.29.
Since calculated value of test statistic F (= 13) is greater than critical
value (= 10.29) so we reject the null hypothesis i.e. we reject the claim
at 5% level of significance.
Thus, we conclude that data provide us sufficient evidence against the
claim so effect of four treatments is not same.
E4) Here, we want to test that the four smells are equally liked. Therefore,
our claim is “four smells are equally liked” and its complement is “four
smells are not equally liked”. So we can take claim as the null
hypothesis and complement as the alternative hypothesis. Thus,
H 0 : Four smells are equally liked

H1 : Four smells are not equally liked

Here, two factors (consumer and smell) are under study so we can use
two-way ANOVA if the distributions of both factors follow normal
distributions but it is not the case. Also data are given in the form of
ranks so we go for Friedman test.
For testing the null hypothesis, we have Friedman test statistic as
n
12
F  R 2j  3k(n  1) … (8)
kn(n  1) j1

85
Non-Parametric Tests In this exercise, ranks are given therefore to perform the Friedman test
we calculate sum of ranks of each smell as:
Respondent Smell I Smell II Smell III Smell IV
1 1 2 4 3
2 1 3 4 2
3 1 3 2 4
4 2 1 3 4
5 1 3 4 2
6 1 3 2 4
7 2 1 4 3
8 1 3 4 2
9 1 3 2 4
10 1 4 3 2
Total 12 26 32 30

Here k = 10 and n = 4 so by putting the values in equation (8), we have

12
F
10  4  (4  1)
122  262  322  302   3 10   4  1
12
 144  676  1024  900   150
200
12
  2744  150  14.64
200
Since k > 6, therefore, the Friedman test statistic F is approximated by
chi square with n − 1 = 4 − 1 = 3 degrees of freedom. Thus, we
compare calculated test statistic with critical value given in χ2-table at 3
degrees of freedom and 1% level of significance.
The critical value of 2 with 3 degrees of freedom at 1% level of
significance is 11.34.
Since calculated value of test statistic (= 14.64) is less than critical
value (= 11.34) so we do not reject the null hypothesis i.e. we support
the claim at 1% level of significance.
Thus, we conclude that the samples fail to provide us sufficient
evidence against the claim so we may assume that four smells are
equally liked.

86
UNIT 16 ANALYSIS OF FREQUENCIES
Structure
16.1 Introduction
Objectives
16.2 Chi-Square Test for Goodness of Fit
16.3 Chi-Square Test for Independence of Attributes
16.4 Summary
16.5 Solutions / Answers

16.1 INTRODUCTION
In testing of hypothesis and estimation of parameter(s), we generally assume
that the random variable follows a distribution such as normal, binomial,
Poisson distribution, etc., but often the need is felt to confirm whether our
assumption is true or not. Therefore, on the basis of observational data for such
testing we use Kolmogorov-Smirnov test (described in Unit 13). But in many When the data are
real world situations, like in business and other areas, the data are collected in classified into different
categories or groups
the form of counts. In some cases, the collected data are classified into different
according to one or more
categories or groups according to one or more attributes. Such type of data is attributes such type of data
known as categorical data. For example, the number of people of a colony can known as categorical data.
be classified into different categories according to age, sex, income, job, etc. or
the books of library can be classified according to their subjects such as books
of Science, Commerce, Art, etc. Now, the question arises “how we tackle the
inference problems arising out of categorical data?” The chi square test is
usually used in such situations. In this unit, we shall discuss two most widely
used applications of chi-square test as goodness of fit and independence of
attributes.
This unit is divided into five sections. Section 16.1 is described the need of
tests which are applied for categorical data. The chi-square test for goodness of
fit is discussed in Section 16.2 in details whereas chi-square test for
independence of attributes is discussed in Section 16.3. Unit ends by providing
summary of what we have discussed in this unit in Section 16.4 and solution of
exercises in Section 16.5.
Objectives
After studying this unit, you should be able to:
 define the categorical data;
 describe the need of tests which are applied for categorical data.
 apply chi-square test for goodness of fit in different situations;
 define the contingency table; and
 apply chi-square test for independence of two attributes.

16.2 CHI-SQUARE TEST FOR GOODNESS OF FIT

In the previous block, we have discussed the parametric tests in which we first
assume the form of the parent population and then perform a test about some
parameter(s) of the population(s) such as mean, variance, proportion, etc.
87
Non-Parametric Tests Generally, the parametric tests are based on the assumption of a normal
population. The suitability of a normal distribution or some other distribution
may itself be verified by means of a goodness of fit test. The chi-square ( χ 2 )
test for goodness of fit was given by Karl Pearson in 1900. It is the oldest non-
parametric test. With the help of this test, we test whether the random variable
under study follows a specified distribution such as binomial, Poisson, normal
or any other distribution when the data are in categorical form. Here, we
compare the actual or observed frequencies in each category with theoretically
expected frequencies that would have occurred if the data followed a specified
or assumed or hypothesized probability distribution. This test is known as
“goodness of fit test” because we test how well an observed frequency
distribution (distribution from which the sample is drawn) fit to the theoretical
distribution such as normal, uniform, binomial, etc.
Assumptions
This test works under the following assumptions:
(i) The sample observations are random and independent.
(ii) The sample size is large.
(iii) The observations may be classified into non-overlapping categories.
(iv) The expected frequency of each class is greater than five.
(v) Sum of observed frequencies is equal to sum of expected frequencies, i.e.,
 O   E.
As usual first step in testing of hypothesis is to setup null and alternative
hypotheses, so our null and alternative hypotheses are setup, below in Step 1.
Step 1: Generally, we are interested to test whether data follow a specified or
assumed or hypothesized distribution F0(x) or a sample has come
from a specified distribution or not. So here we consider only two-
tailed case. Thus, we can take the null and alternative hypotheses as
H0 : Data follow a specified distribution
H1 : Data does not follow a specified distribution
In symbolical form
Here, we take the notation
for sample size is ‘N’
H 0 : F(x)  F0 (x) for all values of x
instead of ‘n’ since in this
test generally we deal with
H1 : F(x)  F0 (x) for at least one value of x
frequencies of the
Step 2: After setting null and alternative hypotheses, our next set up is to
observations and to
represent the sun of draw a random sample. So, let a random sample of size N be drawn
frequencies we take from a population with unknown distribution function F(x) and the
notation ‘N’. data are categorized into k groups or classes. Also let O1, O2,…,Ok are
the observed frequencies and E1, E2, …, Ek are the corresponding
expected frequencies. If the parameter (s) of assumed distribution is
(are) unknown then in this step, we estimate the value of each
parameter of the assumed distribution with the help of sample data by
calculating sample mean, variance, proportion, etc as may be the case.
Step 3: After that, we find the probability of each category or group in which
an observation falls with the help of the assumed probability
distribution.
Step 4: If pi (i =1, 2, …, k) is the probability that an observation falls in ith
88
 category then we find the expected frequency by the formula given Analysis of Frequencies
below
Ei  Npi ; for all i =1, 2, …, k
Note 1: Sometimes, (generally, when the expected frequencies are come in
the form of decimal) it is observed that sum of expected frequencies
not come equal to sum of observed frequencies yet it is necessary
condition for this test so in such cases, last expected frequency is
obtained by subtracting sum of all expected frequencies except last
expected frequency from the sum of observed frequencies, that is,
Ek  N  (E1  E2  ...  Ek 1 )
Step 5: Test statistic:
Since this test compares observed frequencies with the corresponding
expected frequencies, therefore, we are interested in the magnitudes
of the differences between the observed and expected frequencies.
Specifically, we wish to know whether the differences are small
enough to be attributed to chance or they are large due to some other
factors. With the help of observed and expected frequencies we may
compute a test statistic that reflects the magnitudes of differences
between these two quantities when H0 is true. The test statistic is
given by
2
2
 
k
 Oi  Ei  ~ 2 k 1 under H0
i 1 Ei
where, k represents the number of classes. If any expected frequency
is less than 5 then it is pooled or combined with the preceding or
succeeding class then k represents the number of classes that remain
after the combining classes.
The χ2-statistic follows approximately chi-square distribution with
(k –1) degrees of freedom.
If the parameter (s) of the distribution to be fitted is (are) unknown,
that is, not specified in null hypothesis then test statistic χ2 follows
approximately chi-square distribution with (k – r – 1) degrees of
freedom, that is,
2
2
 
k
 Oi  E i  ~ 2k r 1 under H0
i 1 Ei
where, r is the number of unknown parameters which are estimated
from the sample.
Step 6: Obtain critical value of test statistic at given level of significance
under the condition that null hypothesis is true. Table III in the
Appendix at the end of Block I of this course provides critical values
of the test statistic χ2 for various degrees of freedom and different
level of significance.
Step 7: Take the decision about the null hypothesis as:
To take the decision about the null hypothesis, the test statistic
(calculated in Step 5) is compared with chi-square critical (tabulated)
value (observed in Step 6) for a given level of significance (α) under
the condition that the null hypothesis is true.

89
Non-Parametric Tests If calculated value of test statistic is greater than or equal to critical
value with (k  r  1) degrees of freedom at α level of significance
then we reject the null hypothesis H0 at α level of significance,
otherwise we do not reject H0.
Let us do some example based on above test.
Example 1: The following data are collected during a test to determine
consumer preference among five leading brands of bath soaps:
Brand Preferred A B C D E Total
Number of Customers 194 205 204 196 201 1000

Test that the preference is uniform over the five brands at 5% level of
significance.
Solution: Here, we want to test that the preference of customers over five
brands is uniform. So our claim is “the preference of customers over five
brands is uniform” and its complement is “the preference of customers over
five brands is not uniform”. So we can take claim as the null hypothesis and
complement as the alternative hypothesis. Thus,
H0 : The preference of customers over the five brands of bath soap is
uniform
H1 : The preference of customers over the five brands of bath soap is not
uniform
In other words, we can say that
H0 : The probability distribution is uniform
H1 : The probability distribution is not uniform
Since the data are given in the categorical form and we are interested to fit a
distribution, so we can go for chi- square goodness of fit test.
If X denotes preference of customers over the five brands of bath soap that
follows uniform distribution then the probability mass function of uniform
distribution is given by
1
P X  x   ; x  0, 1, 2,..., N
N
The uniform distribution has a parameter N which is given so for testing the
null hypothesis, the test statistic is given by
2
2
 
k
 Oi  Ei 
~  2 k 1 under H 0
i 1 Ei
where, Oi and Ei are the observed and expected frequencies of ith brand of bath
soap respectively.
Here, we want to test the null hypothesis that the preference of the customers is
uniform i.e. follows uniform distribution. Uniform distribution is one in which
all outcomes have equal (or uniform) probability. Therefore, the probability
that the customers prefer one of any brand is same. Thus,
1
p1  p2  p3  p4  p5  p 
5
The theoretical or expected number of customers or frequency for each brand is
obtained by multiplying the appropriate probability by total number of
customers, that is, sample size N. Therefore,
90
 1 Analysis of Frequencies
E1  E 2  E 3  E 4  E5  Np  1000   200
5
2

Calculations for
O  E  :
E
Soap Observed Expected (O−E) (O−E)2 O  E 2
Brand Frequency (O) Frequency (E)
E
A 194 200 −6 36 0.18
B 205 200 5 25 0.13
C 204 200 4 16 0.08
D 196 200 −4 16 0.08
E 201 200 1 1 0.01
Total 1000 1000 0.48

From the above calculations, we have

k
2   i
O  Ei  2  0.48
i 1 Ei
The critical value of chi-square with k  1  5  1  4 degrees of freedom at 5%
level of significance is 9.49.
Since calculated value of test statistic (= 0.48) is smaller than critical value
(= 9.49) so we do not reject the null hypothesis i.e. we support the claim at 5%
level of significance.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so we may assume that the preference of customers over the
five brands of bath soap is uniform.
Example 2: The following data give the number of weekly accidents occurring
on a mile stretch of a particular road:
Number of Accidents 0 1 2 3 4 5 6 or more
Frequency 10 12 12 9 5 3 1

A highway engineer wants to know whether the data follow Poisson

distribution or not at 5% level of significance.

Solution: Here, highway engineer wants to test that the number of accidents
follows Poisson distribution. So our claim is “the number of accidents follows
Poisson distribution” and its complement is “the number of accidents does not
follow Poisson distribution”. So we can take claim as the null hypothesis and
complement as the alternative hypothesis. Thus, Here, we can not use the
Kolmogorov-Smirnov test
H0 : The number of accidents follows Poisson distribution for goodness of fit because
this test is applied only
H1 : The number of accidents does not follow Poisson distribution when all the parameters of
Since the data are given in the categorical form and we are interested to fit a the fitting distribution are
known.
distribution, so we can go for chi-square goodness of fit test.
Here, the parameter of Poison distribution is unknown so testing the null
hypothesis, the test statistic is given by
2
2
 
k
 Oi  Ei  ~  2k r 1 under H 0
i 1 Ei
where, k = number of classes
r = number of parameters estimated from the sample data
91
Non-Parametric Tests If X denotes the number of accidents per week that follows Poisson distribution
then the Poisson probability mass function is given by
e  x
PX  x   ; x  0, 1, 2, ...
x!
where, λ is the mean number of accidents per week.
The difficulty here is that the parameter λ is unknown (it is not specified in null
hypothesis), therefore, it is estimated from sample data. Since λ represents the
mean of the population so it can be estimated by value of sample mean. Thus,
first we find the sample mean and the value of sample mean would be taken as
the estimate of the mean of the Poisson distribution.
S. No. Number of Accidents(X) Frequency(f) fX
1 0 10 0
2 1 12 12
3 2 12 24
4 3 9 27
5 4 5 20
6 5 3 15
7 6 1 6
N = 52  fX  104
The formula for calculating mean is
1 1
X
N
 fX 
52
 104  2

Therefore, ˆ  X  2 and Poisson distribution is

e 2 2x
PX  x   ; x  0, 1, 2, ... … (1)
x!
Now, to find the expected or theoretical number of accidents, we first find the
probability of each class (X = 0, 1, 2,…,6) by putting X = 0, 1,…,6 in equation
(1) respectively. Table X in the Appendix given at the end of this block
provides the Poisson probabilities at a specified value of X and at various
values of λ. Therefore, with the help of this table, we can also find these
probabilities by taking X = 0, 1, …,6 and λ = 2 as:
p1 = P[X = 0] = 0.1353, p 2 = P[X = 1] = 0.2707, p3 = P[X = 2] = 0.2707,
p4 = P[X = 3] = 0.1804, p 5 = P[X = 4] = 0.0902, p6 = P[X = 5] = 0.0361,
p7 = P[X = 6 or more] = 1−P[X < 6] =1− [P[X = 0]+…+ P[X = 5]]
=1 − (0.1353 + 0.2707 + 0.2707 + 0.1804 + 0.0902 + 0.0361) = 0.0166
Thus, expected number of accidents is obtained by multiplying the appropriate
probability by total number of accidents, that is, N. Therefore,
E1  Np1  52  0.1353  7.0356
Similarly,
E 2  Np2  14.0764, E 3  Np3  14.0764, E 4  Np4  9.3808,
E5  Np5  4.6904, E6  Np6  1.8772,
E 7  N   E1  E 2  ...  E 6 
 52   7.0356  14.0764  ...  1.8772   0.8632
92
Since the expected frequencies in the last three classes are < 5, therefore, we Analysis of Frequencies
combine the last three classes in order to realize a cell total of at least 5.
2

Calculations for
O  E  :
E
S.No. Observed Expected (O − E) (O − E)2 O  E 2
Frequency Frequency
E

1 10 7.0356 2.9644 8.7877 1.2490

2 12 14.0764 −2.0764 4.3114 0.3063
3 12 14.0764 −2.0764 4.3114 0.3063
4 9 9.3808 −0.3808 0.1450 0.0155
5 9 7.4308 1.5692 2.4624 0.3314
Total 52 52 2.2085

From above calculations, we have

2
5
 Oi  E i 
2    2.2085
i 1 Ei
Here, we combined the classes so
k = number of classes that remain after the combining classes = 5

r = number of parameters estimated from the sample data = 1

The critical value of χ2 with k  r  1  5  1  1  3 degrees of freedom at 5%
level of significance is 7.81.
Since calculated value of test statistic (= 2.2085) is less than critical value
(= 7.81) so we do not reject the null hypothesis i.e. we support the claim at 5%
level of significance.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so the number of accidents follows Poisson distribution.
Example 3: A random sample of 400 car batteries revealed the following
distribution of battery life (in years):
Life 0-1 1-2 2-3 3-4 4-5 5-6
Frequency 16 90 145 112 27 10

Do these data follow a normal distribution at 1% level of significance?

Solution: Here, we want to test that life of car batteries follows a normal
distribution. So our claim is “the life of car batteries follows normal
distribution” and its complement is “the life of car batteries does not follow
normal distribution”. So we can take claim as the null hypothesis and
complement as the alternative hypothesis. Thus,
H0 : The life of car batteries follows normal distribution
H1 : The life of car batteries does not follow normal distribution
Since the data are given in the categorical form and we are interested to fit a
distribution, so we can go for chi- square goodness of fit test.
Here, the parameters of normal distribution is not given so for testing the null
hypothesis, the test statistic is given by
93
Non-Parametric Tests 2
2
 
k
 Oi  Ei  ~  2k r 1 under H 0
Ei
i 1

where, k = number of classes

r = number of parameters estimated from the sample data
If X represents the life of car battery that follows normal distribution then the
probability density function of normal distribution is given by
1
1  2  x  2
f x   e 2 ;    x  ,      ,   0
 2
There are two parameters (mean µ and standard deviation σ) in the normal
distribution that can be estimated from the sample data as:
Life Mid Value Frequency fX fX2
(X) (f)
0-1 0.5 16 8.0 4.00
1-2 1.5 90 135.0 202.50
2-3 2.5 145 362.5 906.25
3-4 3.5 112 392.0 1372.00
4-5 4.5 27 121.5 546.75
5-6 5.5 10 55.0 302.50
Total N = 400 1074 3330.00

The formula for calculating mean is

1 1
ˆ  X   fX   1074  2.68
N 400
The formula for calculating standard deviation is
2
  fX  
1  2 
ˆ  S    fX  
N 1  N 

2
1  1074  
 3330    1.199  1.06
399  400 
The next step is to find the expected frequency for each class under the
assumption of normally distributed population. The expected frequencies of
normal distribution can be obtained with the help of the method as described in
Unit 4 of MST-003. Therefore, first we find the standard normal veriate
X  ˆ
Z corresponding to each lower limit of the class and then find
ˆ
Fx   PX  x  PZ  z for each value with the help of Table I given in the
Appendix at the end of Block 1of this course.
x  ˆ 0  2.68
Thus, when x = 0, z    2.53 and
ˆ 1.06
F  0  P  Z  z  P  Z  2.53  0.5  P  2.53  Z  0
 0.5  P  0  Z  2.53  0.5  0.4943  0.0057
3  2.68
Similarly, when x = 3, z   0.30 and
1.06
F  3  P  Z  z  P  Z  0.30  0.5  P 0  Z  0.30
 0.5  0.1179  0.6179
94
Therefore, expected frequencies are calculated as follows: Analysis of Frequencies
Life Lower Standard Area Under Normal Difference between Expected
Limit Normal Curve to the Left of Successive Areas Frequency
Variate 
Z i.e. P Z  z   400  col .5
X  ˆ
Z
ˆ
Below 0.0057 − 0
−∞ −∞ 0 2.28
0 = 0.0057
0.0571 − 0.0057
0-1 0 −2.53 0.0057 20.56
= 0.0514
0.2611 − 0.0571
1-2 1 −1.58 0.0571 81.60
= 0.2040
0.6179 − 0.2611
2-3 2 −0.64 0.2611 142.72
= 0.3568
0.8944 − 0.6179
3-4 3 0.30 0.6179 110.6
= 0.2765
0.9706 − 0.8944
4-5 4 1.25 0.8944 30.48
= 0.0762
0.9991 − 0.9706
5-6 5 1.89 0.9706 11.76
= 0.0285
6 and
6 3.13 0.9991 -- --
above

Here, last cell expected frequency is obtained by the formula given by

E 7  N   E1  E 2  ...  E 6   11.76
Since expected frequency in the first cell is < 5, therefore, we combine the first
two classes in order to realize a cell total of at least 5.
2
O  E :
Calculations for
E
Life Observed Expected (O−E) (O−E)2 O  E 2
Frequency Frequency
E
0-1 16 22.84 −6.84 46.7856 2.0484
1-2 90 81.60 8.40 70.5600 0.8647
2-3 145 142.72 2.28 5.1984 0.0364
3-4 112 110.60 1.40 1.9600 0.0177
4-5 27 30.48 −3.48 12.1104 0.3973
5-6 10 11.76 −1.76 3.0976 0.2634
Total 400 400 3.6279
Therefore, from above calculations, we have
2
2
 
6
 Oi  E i 
 3.6279
Ei
i 1

Here, we combined the classes so

k = number of classes that remain after the combining classes = 6
r = number of parameters estimated from the sample data = 2
The critical value of χ2 with k  r  1  6  2  1  3 degrees of freedom at 1%
level of significance is 11.34.
Since calculated value of test statistic (= 3.6279) is less than critical value
(= 11.34) so we do not reject the null hypothesis i.e. we support the claim at
1% level of significance.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so we may assume that the life of car batteries follows normal
distribution.
95
Non-Parametric Tests Example 4: The following data represent the results of an investigation of the
sex distribution of the children of 30 families containing 4 children each:
Number of Sons 0 1 2 3 4
Number of Families 4 8 8 7 3
Apply the chi-square test to see whether the number of sons in a family follows
binomial distribution with probability that a child to be a son is 0.5 at 5% level
of significance.
Solution: Here, we want to test whether the number of sons in a family
follows binomial distribution with p = 0.5. So our claim is “the number of sons
in a family follows binomial distribution with p = 0.5” and its complement is
“the number of sons in a family does not follow binomial distribution with
p = 0.5”. So we can take claim as the null hypothesis and complement as the
alternative hypothesis. Thus,
H0 : The number of sons in a family follows binomial distribution with
p = 0.5
H1 : The number of sons in a family does not follow binomial
distribution with p = 0.5
Since the data are given in the categorical form and we are interested to fit a
distribution, so we can go for chi-square goodness of fit test.
Here, the parameter p of binomial distribution is given so for testing the null
hypothesis, the test statistic is given by
2
2
 
k
 Oi  E i  ~  2k 1
i 1 Ei
where, k = number of classes
If X denotes the number of sons in a family that follows binomial distribution
then the binomial probability mass function is given by
PX  x  n C x p x 1  p 
nx
; x  0,1, 2..., n
Here, total number of children is 4 and p = 0.5. Therefore, we have
4 x
P  X  x   4 Cx (0.5) x 1  0.5 ; x  0,1,2,3,4
4 x
 4 Cx (0.5) x  0.5   4 Cx (0.5) 4 ; x  0,1, 2, 3, 4 …(2)
Now, to find the expected or theoretical number of families, we first find the
probability of each class by putting X = 0, 1, 2, 3, 4 in equation (2)
respectively. Thus,
4
p1  P  X  0  4 C0  0.5  0.0625
4
p2  P  X  1  4 C1  0.5  4  0.0625  0.2500
4
p3  P  X  2  4 C2  0.5  6  0.0625  0.3750
4
p 4  P  X  3  4 C3  0.5  4  0.0625  0.2500
4
p5  P  X  4  4 C4  0.5  1 0.0625  0.0625
The expected number of families is obtained by multiplying the appropriate
probability by total number of families, that is, sample size N. Therefore,

96
 E1  Np1  30  0.0625  1.875 Analysis of Frequencies
Similarly,
E 2  Np 2  7.5, E 3  Np 3  11.25, E 4  Np 4  7.5,
E 5  N   E1  E 2  E 3  E 4   30  1.875  7.5  11.25  7.5  1.875
Since the expected frequency in the last cell is less than 5 therefore, we
combine the last two classes in order to realize a cell total of at least 5.
2

Calculation for
O  E :
E
S. No. Observed Expected (O−E) (O−E)2 O  E 2
Frequency Frequency
E
1 4 1.875 2.125 4.5156 2.4083
2 8 7.5 0.500 0.2500 0.0333
3 8 11.25 −3.250 10.5625 0.9389
4 10 9.375 0.625 0.3906 0.0417
Total 30 30 3.4222

From above calculations, we have

2

 2
5
Oi  Ei 
 3.4222
Ei
i 1

Here, we combined the classes so

k = number of classes that remain after the combining classes = 4
The tabulated value of χ2 with k  1  4  1  3 degrees of freedom and at 5%
level of significance is 7.81.
Since calculated value of test statistic (= 3.4222) is less than critical value
(= 7.81) so we do not reject the null hypothesis i.e. we support the claim at 5%
level of significance.
Thus, we conclude that the sample fails to provide us sufficient evidence
against the claim so we may assume that the number of sons in a family
follows binomial distribution with probability that a child to be a son is 0.5.
Now, you can try the following exercises.
E1) Write one difference between chi-square test and Kolmogorov-Smirnov
test for goodness of fit.
E2) The following table gives the numbers of road accidents that occurred
during the various days of the week:
Days Mon Tue Wed Thu Fri Sat Sun
Number of Accidents 14 15 8 20 11 9 14

Test whether the accidents are uniformly distributed over the week by
chi-square test at 1% level of significance.
E3) The number of customers waiting for service on the checkout counter
line of a large supermarket is examined at random on 64 occasions
during a period. The results are as follows:
Waiting Time
0 or 1 2 3 4 5 6 7 8 or more
(in minutes)
Frequency 5 8 10 11 10 9 7 4

Does the number of customers waiting for service per occasion follows
Poisson distribution with mean 8 minutes at 5% level of significance?
97
Non-Parametric Tests
16.3 CHI-SQUARE TEST FOR INDEPENDENCE OF
ATTRIBUTES
There are many situations where we need to test the independency of two
The chi-square test for
characteristic or attributes of categorical data. For example, a sociologist may
independence of attributes
can be used in the situation wish to know whether level of formal education is independent with income,
in which the data classified whether height of sons depending on height of their fathers or not, etc. If there
according to two attributes is no association between two variables, we say that they are independent. In
or characteristics. other words, we can say that two variables are independent if the distribution of
one is not depending on the distribution of other. To test the independence of
two variables when observations in a population are classified according to
some attributes we may use chi-square test for independence. This test will
indicate only whether or not any association exists between attributes.
To conduct the test, a sample is drawn from the population and the observed
A contingency table is an frequencies are cross-classified according to the two characteristics so that each
arrangement of data into observation belongs to one and only one level of each characteristic. The cross-
a two-way classification.
classification can be conveniently displayed by mean of a table called a
One of the classifications
is entered in rows and the contingency table. Therefore, a contingency table is an arrangement of data
other in columns. into a two-way classification. One of the classifications is entered in rows and
the other in columns.
Assumptions
This test work under the following assumptions:
(i) The sample observations are random and independent.
(ii) The observations may be classified into non-overlapping categories.
(iii) The observed and expected frequencies of each class are greater than 5.
(iv) Sum of observed frequencies is equal to sum of expected frequencies,
i.e.,  O   E.
(v) Each observation in the sample may be classified according to two
characteristics so that each observation belongs to one and only one level
of each characteristic.
Let us discuss the procedure of this test:
Step 1: As usual, first step is to setup null and alternative hypotheses.
Generally, we are interested to test whether two characteristics or
attributes, say, A and B are independent or not. So here we consider
only two-tailed case. Thus, we can take the null and alternative
hypotheses as
H0 : The two characteristics of classification, say, A and B are
independent
H1 : They are not independent
Step 2: After setting null and alternative hypotheses, next step is to draw a
random sample. So, let a sample of N observations is drawn from a
population and they are cross-classified according to two
characteristics, say, A and B. Also let the characteristic A be assumed
to have ‘r’ categories A1, A2, …, Ar and characteristic B be assumed
to have ‘c’ categories B1, B2, …, Bc. The various observed
frequencies in different classes can be expressed in the form of a table
known as contingency table.
98
 B B1 B2 … Bj … Bc Total Analysis of Frequencies

A
A1 O11 O12 … O1j … O1c R1
A2 O21 O22 … O2j … O2c R2
. . . . . .
. . . . . .
. . . . . .
Ai Oi1 Oi2 … Oij … Oic Ri
. . . . . .
. . . . . .
. . . . . .
Ar Or1 Or2 … Orj … Orc Rr
Total C1 C2 … Cj … Cc N
Here, Oij represents the number of observations corresponding to ith
level of characteristic A and jth level of characteristic B. Ri and Cj
represent the sum of number of observations corresponding to ith level
of characteristic A, i.e. ith row and jth level of characteristic B, i.e. jth
column respectively.
Step 3: The chi-square test of independence of attributes compares observed
frequencies with frequencies that are expected when H0 is true.To
calculate the expected frequencies, we first find the probabilities that
an observation lies in a particular cell for all i =1, 2, …, r and j =1, 2,
…, c. These probabilities are calculated according to the
multiplication law of probability (described in Section 3.4 of Unit 3 of
the course MST-003). According to this law, if two events are
independent then the probability of their joint occurrence is equal to
the product of their individual probabilities. Therefore, the probability
that an observation falls in cell (i, j) is equal to the probability that
observation falls in the ith row multiplied by the probability of falling
in the jth column. We estimate these probabilities from the sample
data by Ri/N and Cj/N, respectively. Thus,
R C
P  Ai B j   i . j
N N
Step 4: To obtain expected frequency Eij for (i, j) cells, we multiply this
estimated probability by the total sample size. Thus,
R C
Eij  N. i . j
N N
This equation reduces to
R  C j Sum of i th row  Sum of jth column
E ij  i 
N Total sample size
This form of the equation indicates that we can easily compute an
expected cell frequency for each cell by multiplying together the
appropriate row and column totals and dividing the product by the
total sample size.
Step 5: Test statistic:
Since this test also compares the observed cell frequencies with the
corresponding expected cell frequencies, therefore, we are interested
in the magnitudes of the differences between the observed and
expected cell frequencies. Specifically, we wish to know whether the
differences are small enough to be attributed to chance or they are
99
Non-Parametric Tests large due to some other factors. With the help of observed and
expected frequencies, we may compute a test statistic that reflects the
magnitudes of differences between these two quantities. When H0 is
true, the test statistic is
2
2
  
r c
O ij  Eij 
~ 2r 1c1 under H0
i 1 j1 Eij
The test statistic follows as a chi-square distribution with (r −1)(c −1)
degrees of freedom.
Step 6: Obtain critical value of test statistic corresponding (r −1)(c −1) df at
given level of significance under the condition that null hypothesis is
true. Table III in the Appendix at the end of Block I of this course
provides critical values of the test statistic χ2 for various df and
different level of significance.
Step 7: Take the decision about the null hypothesis as:
To take the decision about the null hypothesis, the test statistic
(calculated in Step 5) is compared with chi-square critical (tabulated)
value (observed in Step 6) for a given level of significance (α) under
the condition that the null hypothesis is true.
If calculated value of test statistic is greater than or equal to tabulated
value with (r −1)(c −1) degrees of freedom at α level of significance,
we reject the null hypothesis at α level of significance otherwise we
do not reject it.
Let us do some examples to become more user friendly with this test.
Example 5: 1000 students at college level were graded according to their IQ
level and the economic condition of their parents.
Economic IQ level
Condition High Low Total
Poor 240 160 400
Rich 460 140 600
Total 700 300 1000

Test that IQ level of students is independent of the economic condition of their

parents at 5% level of significance.
Solution: Here, we want to test that IQ level of students is independent of the
economic condition of their parents. So our claim is “IQ level and economic
condition are independent” and its complement is “IQ level and economic
condition are not independent”. So we can take claim as the null hypothesis
and complement as the alternative hypothesis. Thus,
H0 : IQ level and economic condition are independent
H1 : IQ level and economic condition are not independent
Here, we are interesting to test the independence of two characteristics IQ and
economic condition, so we go for chi-square test of independence.
For testing the null hypothesis, the test statistic is
2

2  
r c
Oij  Eij 
~ 2r 1c1 under H0
i 1 j1 Eij
Now, under H0, the expected frequencies can be obtained as:
100
 Eij = Expected frequency of the ith row and jth column Analysis of Frequencies

R i  C j Sum of i th row  Sum of jth column

= 
N Total sample size

Therefore,
R 1  C1 400  700 R  C 2 400  300
E11    280, E12  1   120
N 1000 N 1000
R  C1 600  700 R  C 2 600  300
E 21  2   420, E 22  2   180
N 1000 N 1000
2

Calculations for
O  E :
E
Observed Expected (O – E) (O – E)2 O  E 2
Frequency (O) Frequency (E)
E
240 280 −40 1600 5.71
160 120 40 1600 13.33
460 420 40 1600 3.81
140 180 −40 1600 8.89
Total = 1000 1000 31.74
Therefore, from above calculations, we have
2
2
  
r c
O ij  Eij 
 31.74
i 1 j1 Eij
The degrees of freedom will be (r –1)(c –1) = (2 – 1)(2 – 1) = 1.
The critical value of χ2 with 1 degree of freedom at 5% level of significance is
3.84.
Since calculated value of test statistic (= 31.74) is greater than critical value
(= 3.84) so we reject the null hypothesis i.e. we reject the claim at 5% level of
significance.
Thus, we conclude that sample provides us sufficient evidence against the
claim so IQ level of students is not independent of the economic condition of
their parents.
Example 6: Calculate the expected frequencies for the following data
presuming the two attributes and check that condition of home and condition of
the child are independent at 5% level of significance.
Condition of Child Condition of Home
Clean Dirty
Clear 70 50
Fairly Clean 80 20
Dirty 35 45

Solution: Here, we want to test that condition of home and condition of the
child are independent. So our claim is “condition of home and condition of
child are independent” and its complement is “condition of home and condition
of child are not independent”. So we can take claim as the null hypothesis and
complement as the alternative hypothesis. Thus,
H0 : Condition of home and condition of child are independent
H1 : Condition of home and condition of child are not independent

101
Non-Parametric Tests Here, we are interesting to test the independence of two characteristics,
condition of home and condition of the child, so we go for chi-square test of
independence.
For testing the null hypothesis, test statistic is
2

2  
r c
Oij  Eij 
~ 2r 1c1 under H0
i 1 j1 Eij
Now, under H0, the expected frequencies can be obtained as:
Condition of Condition of Home Total
Child Clean Dirty
Clear 70 50 120
Fairly Clean 80 20 100
Dirty 35 45 80
Total 185 115 300

Eij = Expected frequency of the ith row and jth column

Ri  Cj Sum of i th row  Sum of jth column
= 
N Total sample size
Therefore,
R 1  C1 120  185 R  C2 120  115
E11    74; E12  1   46;
N 300 N 300
R 2  C1 100  185 R  C2 100  115
E 21    61.67; E 22  2   38.33;
N 300 N 300
R  C1 80  185 R  C2 80  115
E 31  3   49.33; E 32  3   30.67
N 300 N 300
2

Calculations for
O  E :
E
Observed Expected (O – E) (O – E)2 O  E 2
Frequency (O) Frequency (E)
E
70 74.00 −4.00 16.00 0.22
50 46.00 4.00 16.00 0.35
80 61.67 18.33 335.99 5.45
20 38.33 −18.33 335.99 8.77
35 49.33 −14.33 205.35 4.16
45 30.67 14.33 205.35 6.70
Total = 300 300 25.64
Therefore, from above calculations, we have
2
2
  
r c
O ij
 25.64
 Eij 
Eij
i 1 j1

The degrees of freedom will be (r –1)(c –1) = (3 – 1)(2 – 1) = 2.

The critical value of χ2 with 2 degrees of freedom at 5% level of significance is
5.99.
Since calculated value of test statistic (= 25.64) is greater than critical value
(= 5.99) so we reject the null hypothesis i.e. we reject the claim at 5% level of
significance.
Thus, we conclude that the sample provides us sufficient evidence against the
claim so condition of home and condition of the child are not independent.
102
Now, you can try the following exercises. Analysis of Frequencies

E4) 1500 families were selected at random in a city to test the belief that
high income families usually send their children to public schools and
low income families often send their children to government schools.
The following results were obtained in the study conducted.
Income Public School Government School Total
Low 300 600 900
High 435 165 600
Total 735 765 1500

Use chi-square test at 1% level of significance to test whether the two

attributes are independent.
E5) The following contingency table presents the analysis of 300 persons
according to hair colour and eye colour:
Hair Eye Colour
Colour Blue Grey Brown Total
Fair 30 10 40 80
Brown 40 20 40 100
Black 50 30 40 120
Total 120 60 120 300

Test the hypothesis that there is an association between hair colour and
eye colour at 1% level of significance.
We now end this unit by giving a summary of what we have covered in it

16.4 SUMMARY
In this unit, we have discussed following points:
1. When the data are classified into different categories or groups according to
one or more attributes such type of data known as categorical data.
2. Needs of tests which are applied for categorical data.
3. The chi-square test for goodness of fit.
4. A contingency table is an arrangement of data into a two-way
classification. One of the classifications is entered in rows and the other in
columns.
5. The chi-square test for independence of two attributes.

16.5 SOLUTIONS / ANSWERS

E1) The main difference between chi-square test and Kolmogorov-Smirnov
(K-S) test for goodness of fit is that the chi-square test is designed for
categorical data whereas K-S test is designed for the continuous data.
E2) Here, we want to test that road accidents are uniformly distributed over
the week. So our claim is “the accidents are uniformly distributed over
the week” and its complement is “the accidents are not uniformly
distributed over the week”. So we can take claim as the null hypothesis
and complement as the alternative hypothesis. Thus,
H0 : The accidents are uniformly distributed over the week
H1 : The accidents are not uniformly distributed over the week
103
Non-Parametric Tests Since the data are given in the categorical form and we are interested to
fit a distribution, so we can go for chi- square goodness of fit test.
If X denotes the number of accidents per day that follows uniform
distribution then the probability mass function of uniform distribution is
given by
1
P X  x   ; x  0, 1, 2,..., N
N
The uniform distribution has a parameter N which is given so for
testing the null hypothesis, the test statistic is given by
k
Oi  E i  2
2   ~ 2k 1 
i 1 Ei
Since the uniform distribution is one in which all outcomes considered
have equal or uniform probability. Therefore, the probability that the
accident occurs in any day is same. Thus,
1
p1  p 2  p 3  p 4  p5  p 
7
The theoretical or expected frequency for each day is obtained by
multiplying the appropriate probability by the total number of
accidents, that is, sample size N. Therefore,
1
E1  E 2  E 3  E 4  E 5  Np  91   13
7
2

Calculations for
O  E  :
E
Days Observed Expected (O−E) (O−E) 2 O  E 2
Frequency (O) Frequency (E)
E
Mon 14 13 1 1 0.0769
Tue 15 13 2 4 0.3077
Wed 8 13 −5 25 1.9231
Thu 20 13 7 49 3.7692
Fri 11 13 −2 4 0.3077
Sat 9 13 −4 16 1.2308
Sun 14 13 1 1 0.0769
Total 91 91 7.6923

From the above calculation, we have

k
Oi  E i  2
2    7.6923
i 1 Ei
The critical value of chi-square with k  1  7  1  6 degrees of freedom
at 1% level of significance is 16.81.
Since calculated value of test statistic (= 7.6923) is less than critical
value (= 16.81) so we do not reject the null hypothesis i.e. we support
the claim at 1% level of significance.
Thus, we conclude that the sample fails to provide us sufficient
evidence against the claim so we may assume that the accidents are
uniformly distributed over the week.

104
E3) Here, we want to test that the number of costumers waiting for Analysis of Frequencies
servicing follow a Poisson distribution with mean waiting time 8
minutes. So our claim is “the number of costumers waiting for servicing
follow a Poisson distribution with mean waiting time 8 minutes” and its
complement is “the number of costumers waiting for servicing does not
follow a Poisson distribution with mean waiting time 8 minutes”. So we
can take claim as the null hypothesis and complement as the alternative
hypothesis. Thus,
H0 : The number of costumers waiting for servicing follows
Poisson distribution with mean waiting time 8 minutes
H1 : The number of costumers waiting for servicing does not
follow Poisson distribution with mean waiting time 8
minutes
Since the data are given in the categorical form and we are interested to
fit a distribution, so we can go for chi- square goodness of fit test.
Here, the parameter λ of Poison distribution is given so for testing the
null hypothesis, the test statistic is given by
2
2
 
k
 Oi  E i  ~ 2k 1
i 1 Ei
where, k = number of classes
If X denotes the number of costumers waiting for servicing per
occasion that follows the Poisson distribution then the Poisson
probability mass function is given by
e  x
PX  x   ; x  0, 1, 2, ...
x!
where, λ is the mean number of costumers waiting for servicing per
occasion. Since it is specified 8 minutes in null hypothesis therefore, the
Poisson probability mass function is,
e 8 8x
PX  x   ; x  0, 1, 2, ... … (3)
x!
Now, to find the expected number of costumers waiting for servicing
per occasion, we first find the probability of each class by putting
X = 0, 1, … in equation (3). With the help of Table X in the Appendix
at the end of this block, we can also find these probabilities by taking
X = 0, 1, … and λ = 8 as:
p 1 = P[X = 0 or 1] = P[X = 0] + P[X =1] = 0.0003 + 0.0027 = 0.0030,
p 2 = P[X = 2] = 0.0107, p 3 =P[X =3] = 0.0286, p4 = P[X = 4] = 0.0573,
p 5 = P[X = 5]= 0.0916, p6 = P[X = 6] = 0.1221, p7 = P[X = 7] = 0.1396,
p 8 = P[X = 8 or more] = 1−P[X < 8] =1− [P[X = 0]+…+ P[X = 7]]
=1 − (0.0003 + 0.0027 + 0.0107 + 0.0286 + 0.0573 + 0.0916 +
0.1221 + 0.1396) = 0.5471
The expected number of costumers waiting for servicing is obtained by
multiplying the appropriate probability by total number of accidents,
that is, N. Therefore,
105
Non-Parametric Tests E1  Np1  64  0.0030  0.1920
E 2  Np 2  0.6848, E 3  Np3  1.8304, E 4  Np 4  3.6672,
E5  Np5  5.8624, E6  Np6  7.8144, E 7  Np7  8.9344,
E8  N   E1  E 2  ...  E 7   35.0144
Since the expected frequencies in the first four classes are less than 5
therefore, we combine the first four classes in order to realize a cell
total of at least 5.
2

Calculations for
 O  E
:
E
S. Observed Expected (O−E) (O−E)2 O  E 2
No. Frequency Frequency
E
1 34 6.3744 27.6256 763.1738 119.7248
2 10 5.8624 4.1376 17.1197 2.9203
3 9 7.8144 1.1856 1.4056 0.1799
4 7 8.9344 1.9344 3.7419 0.4188
5 4 35.0144 −31.0144 961.8930 27.4714
64 64 150.7151

From the above calculations, we have

2
2
 
5
 Oi  Ei 
 150.7151
Ei
i 1

Here, we combined the classes so Here,

param
k = number of classes that remain after the combining classes = 5 we tak
The critical value of chi-square with k  1  5  1  4 degrees of freedom
at 5% level of significance is 9.49.
Since calculated value of test statistic (= 150.7151) is greater than
critical value (= 9.49) so we reject the null hypothesis i.e. we reject the
claim at 5% level of significance.
Thus, we conclude that the sample provides us sufficient evidence
against the claim so the number of costumers waiting for servicing per
occasion does not follow Poisson distribution.

E4) Here, we want to test that the family income and selection of school are
independent. So our claim is “the family income and selection of school
are independent” and its complement is “the family income and
selection of school are not independent”. So we can take claim as the
null hypothesis and complement as the alternative hypothesis. Thus,
H0 : Family income and selection of school are independent
H1 : Family income and selection of school are not independent
Here, we are interesting to test the independence of two attributes
family income and selection of school, so we go for chi-square test of
independence.
For testing the null hypothesis, the test statistic is
2
2
  
r c
Oij  Eij 
~ 2r 1c1 under H0
i1 j1 Eij
106
 Now, under H0, the expected frequencies can be obtained as: Analysis of Frequencies

Eij = Expected frequency of the ith row and jth column

Ri  Cj Sum of i th row  Sum of jth column
= 
N Total sample size
Therefore,
R 1  C1 900  735 R  C 2 900  765
E11    441, E12  1   459,
N 1500 N 1500
R 2  C1 600  735 R  C2 600  765
E 21    294, E 22  2   306
N 1500 N 1500
2

Calculations for
 O  E
:
E
Observed Expected (O – E) (O – E)2 O  E 2
Frequency (O) Frequency (E)
E
300 441 −141 19881 45.08
600 459 141 19881 43.31
435 294 141 19881 67.62
165 306 −141 19881 64.97
Total = 1500 1500 220.98
Therefore, from above calculations, we have
2
2
  
r c
O ij  E ij 
 220.98
i 1 j1 Eij
The degrees of freedom will be (r –1)(c –1) = (2 – 1)(2 – 1) = 1.
The critical value of chi square with 1 df at 1% level of significance is
6.63.
Since calculated value of test statistic (= 220.98) is greater than critical
value (= 6.63) so we reject the null hypothesis i.e. we reject the claim at
1% level of significance.
Thus, we conclude that the sample provides us sufficient evidence
against the claim so two attributes family income and selection of
school are not independent.
E5) Here, we want to test that there is an association between hair colour
and eye colour that means we want to test that hair colour and eye
colour are not independent. So our claim is “hair colour and eye colour
are not independent” and its complement is “hair colour and eye colour
are independent”. So we can take complement as the null hypothesis
and claim as the alternative hypothesis. Thus,
H0 : Hair and eye colour are independent
H1 : Hair and eye colour are associated
Here, we are interesting to test the independence of two attributes hair
colour and eye colour, so we go for chi-square test of independence.
For testing the null hypothesis, the test statistic is
2
2
  
r c
O
ij  Eij 
~ 2r 1c1 under H0
i1 j1 Eij
107
Non-Parametric Tests Now, under H0, the expected frequencies can be obtained as:
Eij = Expected frequency of the ith row and jth column
Ri  Cj Sum of i th row  Sum of jth column
= 
N Total sample size
Therefore,
R 1  C1 80  120 R  C 2 80  60
E11    32; E12  1   16;
N 300 N 300
R  C3 80  120 R  C1 100  120
E13  1   32; E 21  2   40;
N 300 N 300
R  C2 100  60 R  C3 100  120
E 22  2   20; E 23  2   40;
N 300 N 300

R 3  C1 120  120 R  C 2 120  60

E 31    48; E 32  3   24;
N 300 N 300
R  C3 120  120
E 33  3   48
N 300
2

Calculations for
O  E  :
E
Observed Expected (O – E) (O – E)2 O  E 2
Frequency (O) Frequency (E)
E
30 32 −2 4 0.13
10 16 −6 36 2.25
40 32 8 64 2
40 40 0 0 0
20 20 0 0 0
40 40 0 0 0
50 48 2 4 0.08
30 24 6 36 1.50
40 48 −8 64 1.33
Total = 300 300 7.29
Therefore, from above calculations, we have
2
2
  
r c
O ij  Eij 
 7.29
i 1 j1 Eij
The degrees of freedom will be (r –1)(c –1) = (3 – 1)(3 – 1) = 4.
The critical value of chi square with 4 df at 1% level of significance is
13.28.
Since calculated value of test statistic chi-square (= 7.29) is less than
critical value (= 13.28) so we do not reject the null hypothesis and
reject the alternative hypothesis i.e. we reject the claim at 1% level of
significance.
Thus, we conclude that the sample provide us sufficient evidence
against the claim so hair colour is independent of eye colour.

108
APPENDIX
Table-I: Standard Normal Distribution (Z-table)
The first column and first row of the table indicate the values of standard normal variate Z
at first and second place of decimal. The entry represents the upper tail area under the
curve or probability i.e. P[0  Z  z] for different values of Z.

Z 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09

0 0.0000 0.0040 0.0080 0.0120 0.0160 0.0199 0.0239 0.0279 0.0319 0.0359
0.1 0.0398 0.0438 0.0478 0.0517 0.0557 0.0596 0.0636 0.0675 0.0714 0.0753
0.2 0.0793 0.0832 0.0871 0.0910 0.0948 0.0987 0.1026 0.1064 0.1103 0.1141
0.3 0.1179 0.1217 0.1255 0.1293 0.1331 0.1368 0.1406 0.1443 0.1480 0.1517
0.4 0.1554 0.1591 0.1628 0.1664 0.1700 0.1736 0.1772 0.1808 0.1844 0.1879
0.5 0.1915 0.1950 0.1985 0.2019 0.2054 0.2088 0.2123 0.2157 0.2190 0.2224
0.6 0.2257 0.2291 0.2324 0.2357 0.2389 0.2422 0.2454 0.2486 0.2517 0.2549
0.7 0.2580 0.2611 0.2642 0.2673 0.2704 0.2734 0.2764 0.2794 0.2823 0.2852
0.8 0.2881 0.2910 0.2939 0.2967 0.2995 0.3023 0.3051 0.3078 0.3106 0.3133
0.9 0.3159 0.3186 0.3212 0.3238 0.3264 0.3289 0.3315 0.3304 0.3365 0.3389
1.0 0.3413 0.3438 0.3461 0.3485 0.3508 0.3531 0.3554 0.3577 0.3599 0.3621
1.1 0.3643 0.3665 0.3686 0.3708 0.3729 0.3749 0.3770 0.3790 0.3810 0.3830
1.2 0.3849 0.3869 0.3888 0.3907 0.3925 0.3944 0.3962 0.3980 0.3997 0.4015
1.3 0.4032 0.4049 0.4066 0.4082 0.4099 0.4115 0.4131 0.4147 0.4162 0.4177
1.4 0.4192 0.4207 0.4222 0.4236 0.4251 0.4265 0.4279 0.4292 0.4306 0.4319
1.5 0.4332 0.4345 0.4357 0.4370 0.4382 0.4394 0.4406 0.4418 0.4429 0.4441
1.6 0.4452 0.4463 0.4474 0.4484 0.4495 0.4505 0.4515 0.4525 0.4535 0.4545
1.7 0.4554 0.4564 0.4573 0.4582 0.4591 0.4599 0.4608 0.4616 0.4625 0.4633
1.8 0.4641 0.4649 0.4656 0.4664 0.4671 0.4678 0.4686 0.4693 0.4699 0.4706
1.9 0.4713 0.4719 0.4726 0.4732 0.4738 0.4744 0.4750 0.4756 0.4761 0.4767
2.0 0.4772 0.4778 0.4783 0.4788 0.4793 0.4798 0.4803 0.4808 0.4812 0.4817
2.1 0.4821 0.4826 0.4830 0.4834 0.4838 0.4842 0.4846 0.4850 0.4854 0.4857
2.2 0.4861 0.4864 0.4868 0.4871 0.4875 0.4878 0.4881 0.4884 0.4887 0.4890
2.3 0.4893 0.4896 0.4898 0.4901 0.4904 0.4906 0.4909 0.4911 0.4913 0.4916
2.4 0.4918 0.4920 0.4922 0.4925 0.4927 0.4929 0.4931 0.4932 0.4934 0.4936
2.5 0.4938 0.4940 0.4941 0.4943 0.4945 0.4946 0.4948 0.4949 0.4951 0.4952
2.6 0.4953 0.4955 0.4956 0.4957 0.4959 0.4960 0.4961 0.4962 0.4963 0.4964
2.7 0.4965 0.4966 0.4967 0.4968 0.4969 0.4970 0.4971 0.4972 0.4973 0.4974
2.8 0.4974 0.4975 0.4976 0.4977 0.4977 0.4978 0.4979 0.4979 0.4980 0.4981
2.9 0.4981 0.4982 0.4982 0.4983 0.4984 0.4984 0.4985 0.4985 0.4986 0.4986
3.0 0.4987 0.4987 0.4987 0.4988 0.4988 0.4989 0.4989 0.4989 0.4990 0.4990
3.1 0.4990 0.4991 0.4991 0.4991 0.4992 0.4992 0.4992 0.4992 0.4993 0.4993
3.2 0.4993 0.4993 0.4994 0.4994 0.4994 0.4994 0.4994 0.4995 0.4995 0.4995
3.3 0.4995 0.4995 0.4995 0.4996 0.4996 0.4996 0.4996 0.4996 0.4996 0.4997
3.4 0.4997 0.4997 0.4997 0.4997 0.4997 0.4997 0.4997 0.4997 0.4997 0.4998
3.5 0.4998 0.4998 0.4998 0.4998 0.4998 0.4998 0.4998 0.4998 0.4998 0.4998
3.6 0.4998 0.4998 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999
3.7 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999
3.8 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999 0.4999
3.9 0.5000 0.5000 0.5000 0.5000 0.5000 0.5000 0.5000 0.5000 0.5000 0.5000
Sampling Distributions

Table-II: Student’s t-Distribution (t-table)

The first column of this table indicates the degrees of freedom and first
row a specified upper tail area (α). The entry represents the value of
t-statistic such that the area under the curve of t-distribution to its
upper tail is equal to α.

One- tail α = 0.10 0.05 0.025 0.01 0.005

ν =1 3.078 6.314 12.706 31.821 63.657
2 1.886 2.920 4.303 6.965 9.925
3 1.638 2.353 3.182 4.541 5.841
4 1.533 2.132 2.776 3.747 4.604
5 1.476 2.015 2.571 3.365 4.032
6 1.440 1.943 2.447 3.143 3.707
7 1.415 1.895 2.365 2.998 3.499
8 1.397 1.860 2.306 2.896 3.355
9 1.383 1.833 2.262 2.821 3.250
10 1.372 1.812 2.228 2.764 3.169
11 1.363 1.796 2.201 2.718 3.106
12 1.356 1.782 2.179 2.681 3.055
13 1.350 1.771 2.160 2.650 3.012
14 1.345 1.761 2.145 2.624 2.977
15 1.341 1.753 2.131 2.602 2.947
16 1.337 1.746 2.120 2.583 2.921
17 1.333 1.740 2.110 2.567 2.898
18 1.330 1.734 2.101 2.552 2.878
19 1.328 1.729 2.093 2.539 2.861
20 1.325 1.725 2.086 2.528 2.845
21 1.323 1.721 2.080 2.518 2.831
22 1.321 1.717 2.074 2.508 2.819
23 1.319 1.714 2.069 2.500 2.807
24 1.318 1.711 2.064 2.492 2.797
25 1.316 1.708 2.060 2.485 2.787
26 1.315 1.706 2.056 2.479 2.779
27 1.314 1.703 2.052 2.473 2.771
28 1.313 1.701 2.048 2.467 2.763
29 1.311 1.699 2.045 2.462 2.756
30 1.310 1.697 2.042 2.457 2.750
40 1.303 1.684 2.021 2.423 2.704
60 1.296 1.671 2.000 2.390 2.660
120 1.289 1.658 1.980 2.358 2.617
∞ 1.282 1.645 1.960 2.326 2.576
 Appendix
2
Table-III: The Chi-square Distribution ( -table)
The first column of this table indicates the degrees of freedom and first row a
specified upper tail area (α). The entry represents the value of chi-square statistic
such that the area under the curve of chi square distribution to its upper tail is
equal to α
.

α= 0.995 0.99 0.975 0.95 0.90 0.10 0.05 0.025 0.01 0.005
ν=1 --- --- --- --- 0.02 2.71 3.84 5.02 6.63 7.88
2 0.01 0.02 0.05 0.10 0.21 4.61 5.99 7.38 9.21 10.60
3 0.07 0.11 0.22 0.35 0.58 6.25 7.81 9.35 11.34 12.84
4 0.21 0.30 0.48 0.71 1.06 7.78 9.49 11.14 13.28 14.86
5 0.41 0.55 0.83 1.15 1.61 9.24 11.07 12.83 15.09 16.75
6 0.68 0.87 1.24 1.64 2.20 10.64 12.59 14.45 16.81 18.55
7 0.99 1.24 1.69 2.17 2.83 12.02 14.07 16.01 18.48 20.28
8 1.34 1.65 2.18 2.73 3.49 13.36 15.51 17.53 20.09 21.96
9 1.73 2.09 2.70 3.33 4.17 14.68 16.92 19.02 21.67 23.59
10 2.16 2.56 3.25 3.94 4.87 15.99 18.31 20.48 23.21 25.19
11 2.60 3.05 3.82 4.57 5.58 17.28 19.68 21.92 24.72 26.76
12 3.07 3.57 4.40 5.23 6.30 18.55 21.03 23.34 26.22 28.30
13 3.57 4.11 5.01 5.89 7.04 19.81 22.36 24.74 27.69 29.82
14 4.07 4.66 5.63 6.57 7.79 21.05 23.68 26.12 29.14 31.32
15 4.60 5.23 6.26 7.26 8.55 22.31 25.00 27.49 30.58 32.80
16 5.14 5.81 6.91 7.96 9.31 23.54 26.30 28.85 32.00 34.27
17 5.70 6.41 7.56 8.67 10.09 24.77 27.59 30.19 33.41 35.72
18 6.26 7.01 8.23 9.39 10.86 25.99 28.87 31.53 34.81 37.16
19 6.84 7.63 8.91 10.12 11.65 27.20 30.14 32.85 36.19 38.58
20 7.43 8.26 9.59 10.85 12.44 28.41 31.41 34.17 37.57 40.00
21 8.03 8.90 10.28 11.59 13.24 29.62 32.67 35.48 38.93 41.40
22 8.64 9.54 10.98 12.34 14.04 30.81 33.92 36.78 40.29 42.80
23 9.26 10.20 11.69 13.09 14.85 32.01 35.17 38.08 41.64 44.18
24 9.89 10.86 12.40 13.85 15.66 33.20 36.42 39.36 42.98 45.56
25 10.52 11.52 13.12 14.61 16.47 34.38 37.65 40.65 44.31 46.93
26 11.16 12.20 13.84 15.38 17.29 35.56 38.89 41.92 45.64 48.29
27 11.81 12.88 14.57 16.15 18.11 36.74 40.11 43.19 46.96 49.64
28 12.46 13.56 15.31 16.93 18.94 37.92 41.34 44.46 48.28 50.99
29 13.12 14.26 16.05 17.71 19.77 39.09 42.56 45.72 49.59 52.34
30 13.79 14.95 16.79 18.49 20.60 40.26 43.77 46.98 50.89 53.67
40 20.71 22.16 24.43 26.51 29.05 51.81 55.76 59.34 63.69 66.77
50 27.99 29.71 32.36 34.76 37.69 63.17 67.50 71.42 76.15 79.49
60 35.53 37.48 40.48 43.19 46.46 74.40 79.08 83.30 88.38 91.95
70 43.28 45.44 48.76 51.74 55.33 85.53 90.53 95.02 100.42 104.22
80 51.17 53.54 57.15 60.39 64.28 96.58 101.88 106.63 112.33 116.32
90 59.20 61.75 65.65 69.13 73.29 107.56 113.14 118.14 124.12 128.30
100 67.33 70.06 74.22 77.93 82.36 118.50 124.34 129.56 135.81 140.17
120 83.85 86.92 91.58 95.70 100.62 140.23 146.57 152.21 158.95 163.64
 Sampling Distributions
Table-IV: F-Distribution (F-table)
94

F-table contains the values of F-statistic for different set of degrees of freedom  1 ,  2  of
numerator and denominator such that the area under the curve of F-distribution to its right (upper
tail) is equal to α.
F values for α = 0.1
Degrees of
Degrees of Freedom for Numerator(ν1)
Freedom for
Denominator
(ν2)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 24 30 40 60 120 ∞
1 39.86 49.50 53.60 55.83 57.23 58.21 58.91 59.44 59.86 60.20 60.47 60.70 60.91 61.07 61.22 61.35 61.47 61.57 61.66 61.74 62.00 62.26 62.53 62.79 63.06 63.33

2 8.53 9.00 9.16 9.24 9.29 9.33 9.35 9.37 9.38 9.39 9.40 9.41 9.41 9.42 9.42 9.43 9.43 9.44 9.44 9.44 9.45 9.46 9.47 9.47 9.48 9.49
3 5.54 5.46 5.39 5.34 5.31 5.28 5.27 5.25 5.24 5.23 5.22 5.22 5.21 5.20 5.20 5.20 5.19 5.19 5.19 5.18 5.18 5.17 5.16 5.15 5.14 5.13
4 4.54 4.32 4.19 4.11 4.05 4.01 3.98 3.96 3.94 3.92 3.91 3.90 3.89 3.88 3.87 3.86 3.86 3.85 3.85 3.84 3.83 3.82 3.80 3.79 3.78 3.76
5 4.06 3.78 3.62 3.52 3.45 3.40 3.37 3.34 3.32 3.30 3.28 3.27 3.26 3.25 3.24 3.23 3.22 3.22 3.21 3.21 3.19 3.17 3.16 3.14 3.12 3.11
6 3.78 3.46 3.29 3.18 3.11 3.05 3.01 2.98 2.96 2.94 2.92 2.90 2.89 2.88 2.87 2.86 2.86 2.85 2.84 2.84 2.82 2.80 2.78 2.76 2.74 2.72
7 3.59 3.26 3.07 2.96 2.88 2.83 2.79 2.75 2.72 2.70 2.68 2.67 2.65 2.64 2.63 2.62 2.61 2.61 2.60 2.59 2.58 2.56 2.54 2.51 2.49 2.47
8 3.46 3.11 2.92 2.81 2.73 2.67 2.62 2.59 2.56 2.54 2.52 2.50 2.49 2.48 2.46 2.45 2.45 2.44 2.43 2.42 2.40 2.38 2.36 2.34 2.32 2.29
9 3.36 3.01 2.81 2.69 2.61 2.55 2.51 2.47 2.44 2.42 2.40 2.38 2.36 2.35 2.34 2.33 2.32 2.31 2.31 2.30 2.28 2.25 2.23 2.21 2.18 2.16
10 3.29 2.92 2.73 2.61 2.52 2.46 2.41 2.38 2.35 2.32 2.30 2.28 2.27 2.26 2.24 2.23 2.22 2.22 2.21 2.20 2.18 2.16 2.13 2.11 2.08 2.06
11 3.23 2.86 2.66 2.54 2.45 2.39 2.34 2.30 2.27 2.25 2.23 2.21 2.19 2.18 2.17 2.16 2.15 2.14 2.13 2.12 2.10 2.08 2.05 2.03 2.00 1.97
12 3.18 2.81 2.61 2.48 2.39 2.33 2.28 2.24 2.21 2.19 2.17 2.15 2.13 2.12 2.10 2.09 2.08 2.08 2.07 2.06 2.04 2.01 1.99 1.96 1.93 1.90
13 3.14 2.76 2.56 2.43 2.35 2.28 2.23 2.20 2.16 2.14 2.12 2.10 2.08 2.07 2.05 2.04 2.03 2.02 2.01 2.01 1.98 1.96 1.93 1.90 1.88 1.85
14 3.10 2.73 2.52 2.39 2.31 2.24 2.19 2.15 2.12 2.10 2.07 2.05 2.04 2.02 2.01 2.00 1.99 1.98 1.97 1.96 1.94 1.91 1.89 1.86 1.83 1.80
15 3.07 2.70 2.49 2.36 2.27 2.21 2.16 2.12 2.09 2.06 2.04 2.02 2.00 1.99 1.97 1.96 1.95 1.94 1.93 1.92 1.90 1.87 1.85 1.82 1.79 1.76
16 3.05 2.67 2.46 2.33 2.24 2.18 2.13 2.09 2.06 2.03 2.01 1.99 1.97 1.95 1.94 1.93 1.92 1.91 1.90 1.89 1.87 1.84 1.81 1.78 1.75 1.72
17 3.03 2.64 2.44 2.31 2.22 2.15 2.10 2.06 2.03 2.00 1.98 1.96 1.94 1.93 1.91 1.90 1.89 1.88 1.87 1.86 1.84 1.81 1.78 1.75 1.72 1.69
18 3.01 2.62 2.42 2.29 2.20 2.13 2.08 2.04 2.00 1.98 1.95 1.93 1.92 1.90 1.89 1.87 1.86 1.85 1.85 1.84 1.81 1.78 1.75 1.72 1.69 1.66
19 2.99 2.61 2.40 2.27 2.18 2.11 2.06 2.02 1.98 1.96 1.93 1.91 1.89 1.88 1.86 1.85 1.84 1.83 1.82 1.81 1.79 1.76 1.73 1.70 1.67 1.63
20 2.97 2.59 2.38 2.25 2.16 2.09 2.04 2.00 1.96 1.94 1.91 1.89 1.87 1.86 1.84 1.83 1.82 1.81 1.80 1.79 1.77 1.74 1.71 1.68 1.64 1.61
21 2.96 2.57 2.36 2.23 2.14 2.08 2.02 1.98 1.95 1.92 1.90 1.88 1.86 1.84 1.83 1.81 1.80 1.79 1.78 1.78 1.75 1.72 1.69 1.66 1.62 1.59
22 2.95 2.56 2.35 2.22 2.13 2.06 2.01 1.97 1.93 1.90 1.88 1.86 1.84 1.83 1.81 1.80 1.79 1.78 1.77 1.76 1.73 1.70 1.67 1.64 1.60 1.57
23 2.94 2.55 2.34 2.21 2.11 2.05 1.99 1.95 1.92 1.89 1.87 1.85 1.83 1.81 1.80 1.78 1.77 1.76 1.75 1.74 1.72 1.69 1.66 1.62 1.59 1.55
24 2.93 2.54 2.33 2.19 2.10 2.04 1.98 1.94 1.91 1.88 1.85 1.83 1.81 1.80 1.78 1.77 1.76 1.75 1.74 1.73 1.70 1.67 1.64 1.61 1.57 1.53
25 2.92 2.53 2.32 2.18 2.09 2.02 1.97 1.93 1.89 1.87 1.84 1.82 1.80 1.79 1.77 1.76 1.75 1.74 1.73 1.72 1.69 1.66 1.63 1.59 1.56 1.52
26 2.91 2.52 2.31 2.17 2.08 2.01 1.96 1.92 1.88 1.86 1.83 1.81 1.79 1.77 1.76 1.75 1.73 1.72 1.71 1.71 1.68 1.65 1.61 1.58 1.54 1.50
27 2.90 2.51 2.30 2.17 2.07 2.00 1.95 1.91 1.87 1.85 1.82 1.80 1.78 1.76 1.75 1.74 1.72 1.71 1.70 1.70 1.67 1.64 1.60 1.57 1.53 1.49
28 2.89 2.50 2.29 2.16 2.06 2.00 1.94 1.90 1.87 1.84 1.81 1.79 1.77 1.75 1.74 1.73 1.71 1.70 1.69 1.69 1.66 1.63 1.59 1.56 1.52 1.48
29 2.89 2.50 2.28 2.15 2.06 1.99 1.93 1.89 1.86 1.83 1.80 1.78 1.76 1.75 1.73 1.72 1.71 1.69 1.68 1.68 1.65 1.62 1.58 1.55 1.51 1.47
30 2.88 2.49 2.28 2.14 2.05 1.98 1.93 1.88 1.85 1.82 1.79 1.77 1.75 1.74 1.72 1.71 1.70 1.69 1.68 1.67 1.64 1.61 1.57 1.54 1.50 1.46
40 2.84 2.44 2.23 2.09 2.00 1.93 1.87 1.83 1.79 1.76 1.74 1.71 1.70 1.68 1.66 1.65 1.64 1.62 1.61 1.61 1.57 1.54 1.51 1.47 1.42 1.38
60 2.79 2.39 2.18 2.04 1.95 1.87 1.82 1.77 1.74 1.71 1.68 1.66 1.64 1.62 1.60 1.59 1.58 1.56 1.55 1.54 1.51 1.48 1.44 1.40 1.35 1.29
120 2.75 2.35 2.13 1.99 1.90 1.82 1.77 1.72 1.68 1.65 1.63 1.60 1.58 1.56 1.55 1.53 1.52 1.50 1.49 1.48 1.45 1.41 1.37 1.32 1.26 1.19
∞ 2.71 2.30 2.08 1.94 1.85 1.77 1.72 1.67 1.63 1.60 1.57 1.55 1.52 1.51 1.49 1.47 1.46 1.44 1.43 1.42 1.38 1.34 1.30 1.24 1.17 1.00
 F values for α = 0.05
Degrees of
Freedom for Degrees of freedom for numerator(ν1)
Denominator
(ν2) 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 24 30 40 60 120 ∞
1 161 199 216 225 230 234 237 239 240 242 243 244 245 245 246 246 247 247 248 248 249 250 251 252 253 254
2 18.51 19.00 19.16 19.25 19.30 19.33 19.35 19.37 19.39 19.40 19.40 19.41 19.42 19.43 19.43 19.43 19.44 19.44 19.44 19.45 19.45 19.46 19.47 19.48 19.49 19.50
3 10.13 9.55 9.28 9.12 9.01 8.94 8.89 8.85 8.81 8.79 8.76 8.74 8.73 8.72 8.70 8.69 8.68 8.67 8.67 8.66 8.64 8.62 8.59 8.57 8.55 8.53
4 7.71 6.94 6.59 6.39 6.26 6.16 6.09 6.04 6.00 5.96 5.94 5.91 5.89 5.87 5.86 5.84 5.83 5.82 5.81 5.80 5.77 5.75 5.72 5.69 5.66 5.63
5 6.61 5.79 5.41 5.19 5.05 4.95 4.88 4.82 4.77 4.74 4.70 4.68 4.66 4.64 4.62 4.60 4.59 4.58 4.57 4.56 4.53 4.50 4.46 4.43 4.40 4.37
6 5.99 5.14 4.76 4.53 4.39 4.28 4.21 4.15 4.10 4.06 4.03 4.00 3.98 3.96 3.94 3.92 3.91 3.90 3.88 3.87 3.84 3.81 3.77 3.74 3.70 3.67
7 5.59 4.74 4.35 4.12 3.97 3.87 3.79 3.73 3.68 3.64 3.60 3.57 3.55 3.53 3.51 3.49 3.48 3.47 3.46 3.44 3.41 3.38 3.34 3.30 3.27 3.23
8 5.32 4.46 4.07 3.84 3.69 3.58 3.50 3.44 3.39 3.35 3.31 3.28 3.26 3.24 3.22 3.20 3.19 3.17 3.16 3.15 3.12 3.08 3.04 3.01 2.97 2.93
9 5.12 4.26 3.86 3.63 3.48 3.37 3.29 3.23 3.18 3.14 3.10 3.07 3.05 3.03 3.01 2.99 2.97 2.96 2.95 2.94 2.90 2.86 2.83 2.79 2.75 2.71
10 4.96 4.10 3.71 3.48 3.33 3.22 3.14 3.07 3.02 2.98 2.94 2.91 2.89 2.86 2.85 2.83 2.81 2.80 2.79 2.77 2.74 2.70 2.66 2.62 2.58 2.54
11 4.84 3.98 3.59 3.36 3.20 3.09 3.01 2.95 2.90 2.85 2.82 2.79 2.76 2.74 2.72 2.70 2.69 2.67 2.66 2.65 2.61 2.57 2.53 2.49 2.45 2.40
12 4.75 3.89 3.49 3.26 3.11 3.00 2.91 2.85 2.80 2.75 2.72 2.69 2.66 2.64 2.62 2.60 2.58 2.57 2.56 2.54 2.51 2.47 2.43 2.38 2.34 2.30
13 4.67 3.81 3.41 3.18 3.03 2.92 2.83 2.77 2.71 2.67 2.63 2.60 2.58 2.55 2.53 2.51 2.50 2.48 2.47 2.46 2.42 2.38 2.34 2.30 2.25 2.21
14 4.60 3.74 3.34 3.11 2.96 2.85 2.76 2.70 2.65 2.60 2.57 2.53 2.51 2.48 2.46 2.44 2.43 2.41 2.40 2.39 2.35 2.31 2.27 2.22 2.18 2.13
15 4.54 3.68 3.29 3.06 2.90 2.79 2.71 2.64 2.59 2.54 2.51 2.48 2.45 2.42 2.40 2.38 2.37 2.35 2.34 2.33 2.29 2.25 2.20 2.16 2.11 2.07
16 4.49 3.63 3.24 3.01 2.85 2.74 2.66 2.59 2.54 2.49 2.46 2.42 2.40 2.37 2.35 2.33 2.32 2.30 2.29 2.28 2.24 2.19 2.15 2.11 2.06 2.01
17 4.45 3.59 3.20 2.96 2.81 2.70 2.61 2.55 2.49 2.45 2.41 2.38 2.35 2.33 2.31 2.29 2.27 2.26 2.24 2.23 2.19 2.15 2.10 2.06 2.01 1.96
18 4.41 3.55 3.16 2.93 2.77 2.66 2.58 2.51 2.46 2.41 2.37 2.34 2.31 2.29 2.27 2.25 2.23 2.22 2.20 2.19 2.15 2.11 2.06 2.02 1.97 1.92
19 4.38 3.52 3.13 2.90 2.74 2.63 2.54 2.48 2.42 2.38 2.34 2.31 2.28 2.26 2.23 2.21 2.20 2.18 2.17 2.16 2.11 2.07 2.03 1.98 1.93 1.88
20 4.35 3.49 3.10 2.87 2.71 2.60 2.51 2.45 2.39 2.35 2.31 2.28 2.25 2.22 2.20 2.18 2.17 2.15 2.14 2.12 2.08 2.04 1.99 1.95 1.90 1.84
21 4.32 3.47 3.07 2.84 2.68 2.57 2.49 2.42 2.37 2.32 2.28 2.25 2.22 2.20 2.18 2.16 2.14 2.12 2.11 2.10 2.05 2.01 1.96 1.92 1.87 1.81
22 4.30 3.44 3.05 2.82 2.66 2.55 2.46 2.40 2.34 2.30 2.26 2.23 2.20 2.17 2.15 2.13 2.11 2.10 2.08 2.07 2.03 1.98 1.94 1.89 1.84 1.78
23 4.28 3.42 3.03 2.80 2.64 2.53 2.44 2.37 2.32 2.27 2.24 2.20 2.18 2.15 2.13 2.11 2.09 2.08 2.06 2.05 2.01 1.96 1.91 1.86 1.81 1.76
24 4.26 3.40 3.01 2.78 2.62 2.51 2.42 2.36 2.30 2.25 2.22 2.18 2.15 2.13 2.11 2.09 2.07 2.05 2.04 2.03 1.98 1.94 1.89 1.84 1.79 1.73
25 4.24 3.39 2.99 2.76 2.60 2.49 2.40 2.34 2.28 2.24 2.20 2.16 2.14 2.11 2.09 2.07 2.05 2.04 2.02 2.01 1.96 1.92 1.87 1.82 1.77 1.71
26 4.23 3.37 2.98 2.74 2.59 2.47 2.39 2.32 2.27 2.22 2.18 2.15 2.12 2.09 2.07 2.05 2.03 2.02 2.00 1.99 1.95 1.90 1.85 1.80 1.75 1.69
27 4.21 3.35 2.96 2.73 2.57 2.46 2.37 2.31 2.25 2.20 2.17 2.13 2.10 2.08 2.06 2.04 2.02 2.00 1.99 1.97 1.93 1.88 1.84 1.79 1.73 1.67
28 4.20 3.34 2.95 2.71 2.56 2.45 2.36 2.29 2.24 2.19 2.15 2.12 2.09 2.06 2.04 2.02 2.00 1.99 1.97 1.96 1.91 1.87 1.82 1.77 1.71 1.65
29 4.18 3.33 2.93 2.70 2.55 2.43 2.35 2.28 2.22 2.18 2.14 2.10 2.08 2.05 2.03 2.01 1.99 1.97 1.96 1.94 1.90 1.85 1.81 1.75 1.70 1.64
30 4.17 3.32 2.92 2.69 2.53 2.42 2.33 2.27 2.21 2.16 2.13 2.09 2.06 2.04 2.01 1.99 1.98 1.96 1.95 1.93 1.89 1.84 1.79 1.74 1.68 1.62
40 4.08 3.23 2.84 2.61 2.45 2.34 2.25 2.18 2.12 2.08 2.04 2.00 1.97 1.95 1.92 1.90 1.89 1.87 1.85 1.84 1.79 1.74 1.69 1.64 1.58 1.51
60 4.00 3.15 2.76 2.53 2.37 2.25 2.17 2.10 2.04 1.99 1.95 1.92 1.89 1.86 1.84 1.82 1.80 1.78 1.76 1.75 1.70 1.65 1.59 1.53 1.47 1.39
120 3.92 3.07 2.68 2.45 2.29 2.18 2.09 2.02 1.96 1.91 1.87 1.83 1.80 1.78 1.75 1.73 1.71 1.69 1.67 1.66 1.61 1.55 1.50 1.43 1.35 1.25
∞ 3.84 3.00 2.60 2.37 2.21 2.10 2.01 1.94 1.88 1.83 1.79 1.75 1.72 1.69 1.67 1.64 1.62 1.60 1.59 1.57 1.52 1.46 1.39 1.32 1.22 1.00

Appendix
95
 Sampling Distributions
96

F values for α = 0.025

Degrees of
Freedom Degrees of freedom for numerator(ν1)
for
Denominator
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 24 30 40 60 120 ∞
(ν2)

1 648 800 864 900 922 937 948 957 963 969 973 977 980 983 985 987 989 990 992 993 997 1001 1006 1010 1014 1018
2 38.51 39.00 39.17 39.25 39.30 39.33 39.36 39.37 39.39 39.40 39.41 39.41 39.42 39.43 39.43 39.44 39.44 39.44 39.45 39.45 39.46 39.47 39.47 39.48 39.49 39.50
3 17.44 16.04 15.44 15.10 14.88 14.73 14.62 14.54 14.47 14.42 14.37 14.34 14.30 14.28 14.25 14.23 14.21 14.20 14.18 14.17 14.12 14.08 14.04 13.99 13.95 13.90
4 12.22 10.65 9.98 9.60 9.36 9.20 9.07 8.98 8.90 8.84 8.79 8.75 8.72 8.68 8.66 8.63 8.61 8.59 8.58 8.56 8.51 8.46 8.41 8.36 8.31 8.26
5 10.01 8.43 7.76 7.39 7.15 6.98 6.85 6.76 6.68 6.62 6.57 6.52 6.49 6.46 6.43 6.40 6.38 6.36 6.34 6.33 6.28 6.23 6.18 6.12 6.07 6.02
6 8.81 7.26 6.60 6.23 5.99 5.82 5.70 5.60 5.52 5.46 5.41 5.37 5.33 5.30 5.27 5.24 5.22 5.20 5.18 5.17 5.12 5.07 5.01 4.96 4.90 4.85
7 8.07 6.54 5.89 5.52 5.29 5.12 4.99 4.90 4.82 4.76 4.71 4.67 4.63 4.60 4.57 4.54 4.52 4.50 4.48 4.47 4.42 4.36 4.31 4.25 4.20 4.14
8 7.57 6.06 5.42 5.05 4.82 4.65 4.53 4.43 4.36 4.30 4.24 4.20 4.16 4.13 4.10 4.08 4.05 4.03 4.02 4.00 3.95 3.89 3.84 3.78 3.73 3.67
9 7.21 5.71 5.08 4.72 4.48 4.32 4.20 4.10 4.03 3.96 3.91 3.87 3.83 3.80 3.77 3.74 3.72 3.70 3.68 3.67 3.61 3.56 3.51 3.45 3.39 3.33
10 6.94 5.46 4.83 4.47 4.24 4.07 3.95 3.85 3.78 3.72 3.66 3.62 3.58 3.55 3.52 3.50 3.47 3.45 3.44 3.42 3.37 3.31 3.26 3.20 3.14 3.08
11 6.72 5.26 4.63 4.28 4.04 3.88 3.76 3.66 3.59 3.53 3.47 3.43 3.39 3.36 3.33 3.30 3.28 3.26 3.24 3.23 3.17 3.12 3.06 3.00 2.94 2.88
12 6.55 5.10 4.47 4.12 3.89 3.73 3.61 3.51 3.44 3.37 3.32 3.28 3.24 3.21 3.18 3.15 3.13 3.11 3.09 3.07 3.02 2.96 2.91 2.85 2.79 2.73
13 6.41 4.97 4.35 4.00 3.77 3.60 3.48 3.39 3.31 3.25 3.20 3.15 3.12 3.08 3.05 3.03 3.00 2.98 2.96 2.95 2.89 2.84 2.78 2.72 2.66 2.60
14 6.30 4.86 4.24 3.89 3.66 3.50 3.38 3.29 3.21 3.15 3.09 3.05 3.01 2.98 2.95 2.92 2.90 2.88 2.86 2.84 2.79 2.73 2.67 2.61 2.55 2.49
15 6.20 4.77 4.15 3.80 3.58 3.41 3.29 3.20 3.12 3.06 3.01 2.96 2.92 2.89 2.86 2.84 2.81 2.79 2.77 2.76 2.70 2.64 2.59 2.52 2.46 2.40
16 6.12 4.69 4.08 3.73 3.50 3.34 3.22 3.12 3.05 2.99 2.93 2.89 2.85 2.82 2.79 2.76 2.74 2.72 2.70 2.68 2.63 2.57 2.51 2.45 2.38 2.32
17 6.04 4.62 4.01 3.66 3.44 3.28 3.16 3.06 2.98 2.92 2.87 2.82 2.79 2.75 2.72 2.70 2.67 2.65 2.63 2.62 2.56 2.50 2.44 2.38 2.32 2.25
18 5.98 4.56 3.95 3.61 3.38 3.22 3.10 3.01 2.93 2.87 2.81 2.77 2.73 2.70 2.67 2.64 2.62 2.60 2.58 2.56 2.50 2.45 2.38 2.32 2.26 2.19
19 5.92 4.51 3.90 3.56 3.33 3.17 3.05 2.96 2.88 2.82 2.76 2.72 2.68 2.65 2.62 2.59 2.57 2.55 2.53 2.51 2.45 2.39 2.33 2.27 2.20 2.13
20 5.87 4.46 3.86 3.51 3.29 3.13 3.01 2.91 2.84 2.77 2.72 2.68 2.64 2.60 2.57 2.55 2.52 2.50 2.48 2.46 2.41 2.35 2.29 2.22 2.16 2.09
21 5.83 4.42 3.82 3.48 3.25 3.09 2.97 2.87 2.80 2.73 2.68 2.64 2.60 2.56 2.53 2.51 2.48 2.46 2.44 2.42 2.37 2.31 2.25 2.18 2.11 2.04
22 5.79 4.38 3.78 3.44 3.22 3.05 2.93 2.84 2.76 2.70 2.65 2.60 2.56 2.53 2.50 2.47 2.45 2.43 2.41 2.39 2.33 2.27 2.21 2.15 2.08 2.00
23 5.75 4.35 3.75 3.41 3.18 3.02 2.90 2.81 2.73 2.67 2.62 2.57 2.53 2.50 2.47 2.44 2.42 2.39 2.37 2.36 2.30 2.24 2.18 2.11 2.04 1.97
24 5.72 4.32 3.72 3.38 3.15 2.99 2.87 2.78 2.70 2.64 2.59 2.54 2.50 2.47 2.44 2.41 2.39 2.36 2.35 2.33 2.27 2.21 2.15 2.08 2.01 1.94
25 5.69 4.29 3.69 3.35 3.13 2.97 2.85 2.75 2.68 2.61 2.56 2.51 2.48 2.44 2.41 2.38 2.36 2.34 2.32 2.30 2.24 2.18 2.12 2.05 1.98 1.91
26 5.66 4.27 3.67 3.33 3.10 2.94 2.82 2.73 2.65 2.59 2.54 2.49 2.45 2.42 2.39 2.36 2.34 2.31 2.29 2.28 2.22 2.16 2.09 2.03 1.95 1.88
27 5.63 4.24 3.65 3.31 3.08 2.92 2.80 2.71 2.63 2.57 2.51 2.47 2.43 2.39 2.36 2.34 2.31 2.29 2.27 2.25 2.19 2.13 2.07 2.00 1.93 1.85
28 5.61 4.22 3.63 3.29 3.06 2.90 2.78 2.69 2.61 2.55 2.49 2.45 2.41 2.37 2.34 2.32 2.29 2.27 2.25 2.23 2.17 2.11 2.05 1.98 1.91 1.83
29 5.59 4.20 3.61 3.27 3.04 2.88 2.76 2.67 2.59 2.53 2.46 2.48 2.43 2.39 2.36 2.30 2.27 2.25 2.23 2.21 2.15 2.09 2.03 1.96 1.89 1.81
30 5.57 4.18 3.59 3.25 3.03 2.87 2.75 2.65 2.57 2.51 2.33 2.41 2.37 2.34 2.31 2.28 2.26 2.23 2.21 2.20 2.14 2.07 2.01 1.94 1.87 1.79
40 5.42 4.05 3.46 3.13 2.90 2.74 2.62 2.53 2.45 2.39 2.26 2.29 2.25 2.21 2.18 2.15 2.13 2.11 2.09 2.07 2.01 1.94 1.88 1.80 1.72 1.64
60 5.29 3.93 3.34 3.01 2.79 2.63 2.51 2.41 2.33 2.27 2.22 2.17 2.13 2.09 2.06 2.03 2.01 1.98 1.96 1.94 1.88 1.82 1.74 1.67 1.58 1.48
120 5.15 3.80 3.23 2.89 2.67 2.52 2.39 2.30 2.22 2.16 2.10 2.05 2.01 1.98 1.94 1.92 1.89 1.87 1.84 1.82 1.76 1.69 1.61 1.53 1.43 1.31
∞ 5.02 3.69 3.12 2.79 2.57 2.41 2.29 2.19 2.11 2.05 1.99 1.94 1.90 1.87 1.83 1.80 1.78 1.75 1.73 1.71 1.64 1.57 1.48 1.39 1.27 1.00
 F values for α = 0.01
Degrees of
Freedom Degrees of freedom for numerator (ν1)
for
Denominator
(ν2)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 24 30 40 60 120 ∞
1 4063 4992 5404 5637 5760 5890 5890 6025 6025 6025 6025 6167 6167 6167 6167 6167 6167 6167 6167 6167 6235 6261 6287 6313 6339 6366
2 98.50 99.00 99.15 99.27 99.30 99.34 99.34 99.38 99.38 99.38 99.42 99.42 99.42 99.42 99.42 99.42 99.46 99.46 99.46 99.46 99.46 99.47 99.47 99.48 99.49 99.50
3 34.11 30.82 29.46 28.71 28.24 27.91 27.67 27.49 27.34 27.23 27.13 27.05 26.98 26.92 26.87 26.83 26.79 26.75 26.72 26.69 26.60 26.51 26.41 26.32 26.22 26.13
4 21.20 18.00 16.69 15.98 15.52 15.21 14.98 14.80 14.66 14.55 14.45 14.37 14.31 14.25 14.20 14.15 14.11 14.08 14.05 14.02 13.93 13.84 13.75 13.65 13.56 13.46
5 16.26 13.27 12.06 11.39 10.97 10.67 10.46 10.29 10.16 10.05 9.96 9.89 9.83 9.77 9.72 9.68 9.64 9.61 9.58 9.55 9.47 9.38 9.29 9.20 9.11 9.02
6 13.75 10.92 9.78 9.15 8.75 8.47 8.26 8.10 7.98 7.87 7.79 7.72 7.66 7.60 7.56 7.52 7.48 7.45 7.42 7.40 7.31 7.23 7.14 7.06 6.97 6.88
7 12.25 9.55 8.45 7.85 7.46 7.19 6.99 6.84 6.72 6.62 6.54 6.47 6.41 6.36 6.31 6.28 6.24 6.21 6.18 6.16 6.07 5.99 5.91 5.82 5.74 5.65
8 11.26 8.65 7.59 7.01 6.63 6.37 6.18 6.03 5.91 5.81 5.73 5.67 5.61 5.56 5.52 5.48 5.44 5.41 5.38 5.36 5.28 5.20 5.12 5.03 4.95 4.86
9 10.56 8.02 6.99 6.42 6.06 5.80 5.61 5.47 5.35 5.26 5.18 5.11 5.05 5.01 4.96 4.92 4.89 4.86 4.83 4.81 4.73 4.65 4.57 4.48 4.40 4.31
10 10.04 7.56 6.55 5.99 5.64 5.39 5.20 5.06 4.94 4.85 4.77 4.71 4.65 4.60 4.56 4.52 4.49 4.46 4.43 4.41 4.33 4.25 4.17 4.08 4.00 3.91
11 9.65 7.21 6.22 5.67 5.32 5.07 4.89 4.74 4.63 4.54 4.46 4.40 4.34 4.29 4.25 4.21 4.18 4.15 4.12 4.10 4.02 3.94 3.86 3.78 3.69 3.60
12 9.33 6.93 5.95 5.41 5.06 4.82 4.64 4.50 4.39 4.30 4.22 4.16 4.10 4.05 4.01 3.97 3.94 3.91 3.88 3.86 3.78 3.70 3.62 3.54 3.45 3.36
13 9.07 6.70 5.74 5.21 4.86 4.62 4.44 4.30 4.19 4.10 4.02 3.96 3.91 3.86 3.82 3.78 3.75 3.72 3.69 3.66 3.59 3.51 3.43 3.34 3.26 3.17
14 8.86 6.51 5.56 5.04 4.70 4.46 4.28 4.14 4.03 3.94 3.86 3.80 3.75 3.70 3.66 3.62 3.59 3.56 3.53 3.51 3.43 3.35 3.27 3.18 3.09 3.00
15 8.68 6.36 5.42 4.89 4.56 4.32 4.14 4.00 3.89 3.80 3.73 3.67 3.61 3.56 3.52 3.49 3.45 3.42 3.40 3.37 3.29 3.21 3.13 3.05 2.96 2.87
16 8.53 6.23 5.29 4.77 4.44 4.20 4.03 3.89 3.78 3.69 3.62 3.55 3.50 3.45 3.41 3.37 3.34 3.31 3.28 3.26 3.18 3.10 3.02 2.93 2.85 2.75
17 8.40 6.11 5.19 4.67 4.34 4.10 3.93 3.79 3.68 3.59 3.52 3.46 3.40 3.35 3.31 3.27 3.24 3.21 3.19 3.16 3.08 3.00 2.92 2.84 2.75 2.65
18 8.29 6.01 5.09 4.58 4.25 4.01 3.84 3.71 3.60 3.51 3.43 3.37 3.32 3.27 3.23 3.19 3.16 3.13 3.10 3.08 3.00 2.92 2.84 2.75 2.66 2.57
19 8.19 5.93 5.01 4.50 4.17 3.94 3.77 3.63 3.52 3.43 3.36 3.30 3.24 3.19 3.15 3.12 3.08 3.05 3.03 3.00 2.93 2.84 2.76 2.67 2.58 2.49
20 8.10 5.85 4.94 4.43 4.10 3.87 3.70 3.56 3.46 3.37 3.29 3.23 3.18 3.13 3.09 3.05 3.02 2.99 2.96 2.94 2.86 2.78 2.70 2.61 2.52 2.42
21 8.02 5.78 4.87 4.37 4.04 3.81 3.64 3.51 3.40 3.31 3.24 3.17 3.12 3.07 3.03 2.99 2.96 2.93 2.90 2.88 2.80 2.72 2.64 2.55 2.46 2.36
22 7.95 5.72 4.82 4.31 3.99 3.76 3.59 3.45 3.35 3.26 3.18 3.12 3.07 3.02 2.98 2.94 2.91 2.88 2.85 2.83 2.75 2.67 2.58 2.50 2.40 2.31
23 7.88 5.66 4.76 4.26 3.94 3.71 3.54 3.41 3.30 3.21 3.14 3.07 3.02 2.97 2.93 2.89 2.86 2.83 2.80 2.78 2.70 2.62 2.54 2.45 2.35 2.26
24 7.82 5.61 4.72 4.22 3.90 3.67 3.50 3.36 3.26 3.17 3.09 3.03 2.98 2.93 2.89 2.85 2.82 2.79 2.76 2.74 2.66 2.58 2.49 2.40 2.31 2.21
25 7.77 5.57 4.68 4.18 3.85 3.63 3.46 3.32 3.22 3.13 3.06 2.99 2.94 2.89 2.85 2.81 2.78 2.75 2.72 2.70 2.62 2.54 2.45 2.36 2.27 2.17
26 7.72 5.53 4.64 4.14 3.82 3.59 3.42 3.29 3.18 3.09 3.02 2.96 2.90 2.86 2.82 2.78 2.75 2.72 2.69 2.66 2.59 2.50 2.42 2.33 2.23 2.13
27 7.68 5.49 4.60 4.11 3.78 3.56 3.39 3.26 3.15 3.06 2.99 2.93 2.87 2.82 2.78 2.75 2.71 2.68 2.66 2.63 2.55 2.47 2.38 2.29 2.20 2.10
28 7.64 5.45 4.57 4.07 3.75 3.53 3.36 3.23 3.12 3.03 2.96 2.90 2.84 2.79 2.75 2.72 2.68 2.65 2.63 2.60 2.52 2.44 2.35 2.26 2.17 2.06
29 7.60 5.42 4.54 4.04 3.73 3.50 3.33 3.20 3.09 3.00 2.93 2.87 2.81 2.77 2.73 2.69 2.66 2.63 2.60 2.57 2.50 2.41 2.33 2.23 2.14 2.03
30 7.56 5.39 4.51 4.02 3.70 3.47 3.30 3.17 3.07 2.98 2.91 2.84 2.79 2.74 2.70 2.66 2.63 2.60 2.57 2.55 2.47 2.39 2.30 2.21 2.11 2.01
40 7.31 5.18 4.31 3.83 3.51 3.29 3.12 2.99 2.89 2.80 2.73 2.66 2.61 2.56 2.52 2.48 2.45 2.42 2.39 2.37 2.29 2.20 2.11 2.02 1.92 1.81
60 7.08 4.98 4.13 3.65 3.34 3.12 2.95 2.82 2.72 2.63 2.56 2.50 2.44 2.39 2.35 2.31 2.28 2.25 2.22 2.20 2.12 2.03 1.94 1.84 1.73 1.60
120 6.85 4.79 3.95 3.48 3.17 2.96 2.79 2.66 2.56 2.47 2.40 2.34 2.28 2.23 2.19 2.15 2.12 2.09 2.06 2.03 1.95 1.86 1.76 1.66 1.53 1.38
∞ 6.64 4.61 3.78 3.32 3.02 2.80 2.64 2.51 2.41 2.32 2.25 2.19 2.13 2.08 2.04 2.00 1.97 1.94 1.91 1.88 1.79 1.70 1.59 1.47 1.33 1.00

Appendix
97
APPENDIX Appendix
r
nx
Table I: Cumulative Binomial Probabilities,  p  1  p
x 0
x
;r  0,1,2,...,n  1

n r p = 0.5 n r p = 0.5 n r p = 0.5 n r p = 0.5

1 0 0.5000 3 0.1719 11 0.9935 3 0.0038
1 1.0000 4 0.3770 12 0.9991 4 0.0154
2 0 0.2500 5 0.6230 13 0.9999 5 0.0481
1 0.7500 7 0.9453 14 1.0000 6 0.1189
2 1.0000 8 0.9893 15 0 0.0000 7 0.2403
3 0 0.1250 9 0.9990 2 0.0037 8 0.4073
1 0.5000 10 1.0000 3 0.0176 9 0.5927
2 0.8750 11 0 0.0005 4 0.0592 10 0.7597
3 1.0000 1 0.0059 5 0.1509 11 0.8811
4 0 0.0625 2 0.0327 6 0.3036 12 0.9519
1 0.3125 3 0.1133 7 0.5000 13 0.9846
2 0.6875 4 0.2744 8 0.6964 14 0.9962
3 0.9375 5 0.5000 9 0.8491 15 0.9993
4 1.0000 6 0.7256 10 0.9408 16 0.9999
5 0 0.0312 7 0.8867 11 0.9824 17 1.0000
1 0.1875 8 0.9673 12 0.9963 18 1.0000
2 0.5000 9 0.9941 13 0.9995 19 0 0.0000
3 0.8125 10 0.9995 14 1.0000 1 0.0000
4 0.9688 11 1.0000 15 1.0000 2 0.0004
5 1.0000 12 0 0.0002 16 0 0.0000 3 0.0022
6 0 0.0156 1 0.0032 1 0.0003 4 0.0096
1 0.1096 2 0.0193 2 0.0021 5 0.0318
2 0.3438 3 0.0730 3 0.0106 6 0.0835
3 0.6562 4 0.1938 4 0.0384 7 0.1796
4 0.8606 5 0.3872 5 0.1051 8 0.3238
5 0.9844 6 0.6128 6 0.2272 9 0.5000
6 1.0000 7 0.8062 7 0.4018 10 0.6762
7 0 0.0078 8 0.9270 8 0.5982 11 0.8204
1 0.0625 9 0.9807 9 0.7728 12 0.9165
2 0.2266 10 0.9968 10 0.8949 13 0.9682
3 0.5000 11 0.9998 11 0.9616 14 0.9904
4 0.7734 12 1.0000 12 0.9894 15 0.9978
5 0.9375 13 0 0.0001 13 0.9979 16 0.9996
6 0.9922 1 0.0017 14 0.9997 17 1.0000
7 1.0000 2 0.0112 15 1.0000 18 1.0000
8 0 0.0039 3 0.0461 16 1.0000 19 1.0000
1 0.0352 4 0.1334 17 0 0.0000 20 0 0.0000
2 0.1445 5 0.2905 1 0.0001 1 0.0000
3 0.3633 6 0.5000 2 0.0012 2 0.0002
4 0.6367 7 0.7095 3 0.0064 3 0.0013
5 0.8555 8 0.8666 4 0.0245 4 0.0059
6 0.9648 9 0.9539 5 0.0717 5 0.0207
7 0.9961 10 0.9888 6 0.1662 6 0.0577
8 1.0000 11 0.9983 7 0.3145 7 0.1316
9 0 0.0020 12 0.9999 8 0.5000 8 0.2517
1 0.0195 13 1.0000 9 0.6855 9 0.4119
2 0.0898 14 0 0.0001 10 0.8338 10 0.5881
3 0.2539 1 0.0009 11 0.9283 11 0.7483
4 0.5000 2 0.0065 12 0.9755 12 0.8684
5 0.7461 3 0.0287 13 0.9936 13 0.9423
6 0.9102 4 0.0898 14 0.9989 14 0.9793
7 0.9805 5 0.2120 15 0.9999 15 0.9941
8 0.9980 6 0.3953 16 1.0000 16 0.9987
9 1.0000 7 0.6047 17 1.0000 17 0.9998
10 0 0.0010 8 0.7880 18 0 0.0000 18 1.0000
1 0.0107 9 0.9102 1 0.0001 19 1.0000
2 0.0547 10 0.9713 2 0.0007 20 1.0000

109
Non-Parametric Tests

TABLE II: Critical Values for Wilcoxon Test

One-Tailed Test:
α= 0.005 0.01 0.025 0.05 0.10
Two-Tailed Test:
α= 0.01 0.02 0.05 0.10 0.20
n=3 0 0 0 0 0
4 0 0 0 0 1
5 0 0 0 1 3
6 0 0 1 3 4
7 0 1 3 4 6
8 1 2 4 6 9
9 2 4 6 9 11
10 4 6 9 11 15
11 6 8 11 14 18
12 8 10 14 18 22
13 10 13 18 22 27
14 13 16 22 26 32
15 16 20 26 31 37
16 20 24 30 36 43
17 24 28 35 42 49
18 28 33 41 48 56
19 33 38 47 54 63
20 38 44 53 61 70
21 44 50 59 68 78
22 49 56 67 76 87
23 55 63 74 84 95
24 62 70 82 92 105
25 69 77 90 101 114

110
 Table III: Critical Values of Runs Test Appendix

Part-I (Lower Critical Values)

(α = 0.025)

n2
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
n1
2 2 2 2 2 2 2 2 2 2
3 2 2 2 2 2 2 2 2 2 3 3 3 3 3 3
4 2 2 2 3 3 3 3 3 3 3 3 4 4 4 4 4
5 2 2 3 3 3 3 3 4 4 4 4 4 4 4 5 5 5
6 2 2 3 3 3 3 4 4 4 4 5 5 5 5 5 5 6 6
7 2 2 3 3 3 4 4 5 5 5 5 5 6 6 6 6 6 6
8 2 3 3 3 4 4 5 5 5 6 6 6 6 6 7 7 7 7
9 2 3 3 4 4 5 5 5 6 6 6 7 7 7 7 8 8 8
10 2 3 3 4 5 5 5 6 6 7 7 7 7 8 8 8 8 9
11 2 3 4 4 5 5 6 6 7 7 7 8 8 8 9 9 9 9
12 2 2 3 4 4 5 6 6 7 7 7 8 8 8 9 9 9 10 10
13 2 2 3 4 5 5 6 6 7 7 8 8 9 9 9 10 10 10 10
14 2 2 3 4 5 5 6 7 7 8 8 9 9 9 10 10 10 1 11
15 2 3 3 4 5 6 6 7 7 8 8 9 9 10 10 11 11 11 12
16 2 3 4 4 5 6 6 7 8 8 9 9 12 10 11 1 11 12 12
17 2 3 4 4 5 6 7 7 8 9 9 10 10 11 11 1 12 12 13
18 2 3 4 5 5 6 7 8 8 9 9 10 10 11 11 12 12 13 13
19 2 3 4 5 6 6 7 8 8 9 10 10 11 11 12 12 13 13 13
20 2 3 4 5 6 6 7 8 9 9 10 10 11 12 12 13 13 13 14

Part-II (Upper Critical Values)

(α = 0.025)

n2
2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
n1
2
3
4 9 9
5 9 10 10 11 11
6 9 10 11 12 12 13 13 13 13
7 11 12 13 13 14 14 14 14 15 15 15
8 11 12 13 14 14 15 15 16 16 16 16 17 17 17 17 17
9 13 14 14 15 16 16 16 17 17 18 18 18 18 18 18
10 13 14 15 16 16 17 17 18 18 18 19 19 19 20 20
11 13 14 15 16 17 17 18 19 19 19 20 20 20 21 21
12 13 14 16 16 17 18 19 19 20 20 21 21 21 22 22
13 15 16 17 18 19 19 20 20 21 21 22 22 23 23
14 15 16 17 18 19 20 20 21 22 22 23 23 23 24
15 15 16 18 18 19 20 21 22 22 23 23 24 24 25
16 17 18 19 20 21 21 22 23 23 24 25 25 25
17 17 18 19 20 21 22 23 23 24 25 25 26 26
18 17 18 19 20 21 22 23 24 25 25 26 26 27
19 17 18 20 21 22 23 23 24 25 26 26 27 27
20 17 18 20 21 22 23 24 25 25 26 27 27 27

111
Non-Parametric Tests Table IV: Critical Values of Kolmogorov-Smirnov One-Sample Test

One-Tailed Test:
α= 0.10 0.05 0.025 0.01 0.005 α= 0.10 0.05 0.025 0.01 0.005
Two-Tailed Test:
α= 0.20 0.10 0.05 0.02 0.01 α= 0.20 0.10 0.05 0.02 0.01
n = 1 0.900 0.950 0.975 0.990 0.995 n = 21 0.226 0.259 0.287 0.321 0.344
2 0.684 0.776 0.842 0.900 0.929 22 0.221 0.253 0.281 0.314 0.337
3 0.565 0.636 0.708 0.785 0.829 23 0.216 0.247 0.275 0.307 0.330
4 0.493 0.565 0.624 0.689 0.734 24 0.212 0.242 0.269 0.301 0.323
5 0.447 0.509 0.563 0.627 0.669 25 0.208 0.238 0.264 0.295 0.317
6 0.410 0.468 0.519 0.577 0.617 26 0.204 0.233 0.259 0.290 0.311
7 0.381 0.436 0.483 0.538 0.576 27 0.200 0.229 0.254 0.284 0.305
8 0.358 0.410 0.454 0.507 0.542 28 0.197 0.225 0.250 0.279 0.300
9 0.339 0.387 0.430 0.480 0.513 29 1.193 0.221 0.246 0.275 0.295
10 0.323 0.369 0.409 0.457 0.489 30 0.190 0.218 0.242 0.270 0.290
11 0.308 0.352 0.391 0.437 0.468 31 0.187 0.214 0.238 0.266 0.285
12 0.296 0.338 0.375 0.419 0.449 32 0.184 0.211 0.234 0.262 0.281
13 0.285 0.325 0.361 0.404 0.432 33 0.182 0.208 0.231 0.258 0.277
14 0.275 0.314 0.349 0.390 0.418 34 0.179 0.205 0.227 0.254 0.273
15 0.266 0.304 0.338 0.377 0.404 35 0.177 0.202 0.224 0.251 0.269
16 0.258 0.295 0.327 0.366 0.392 36 0.174 0.199 0.221 0.247 0.265
17 0.250 0.286 0.318 0.355 0.381 37 0.172 0.196 0.218 0.244 0.262
18 0.244 0.279 0.309 0.346 0.371 38 0.170 0.194 0.215 0.241 0.258
19 0.237 0.271 0.301 0.337 0.361 39 0.168 0.191 0.213 0.238 0.255
20 0.232 0.265 0.294 0.329 0.352 40 0.165 0.189 0.210 0.235 0.252
Approximation
1.07 1.22 1.36 1.52 1.63
for n > 40
n n n n n

112
 Table V: Critical Values of Kolmogorov-Smirnov Test for Two Samples of Appendix
Equal Size

One-Tailed Test:
α= 0.10 0.05 0.025 0.01 0.005 α= 0.10 0.05 0.025 0.01 0.005
Two-Tailed Test:
α= 0.20 0.10 0.05 0.02 0.01 α= 0.20 0.10 0.05 0.02 0.01
n = 3 2/3 2/3 n = 20 6/20 7/20 8/20 9/20 10/20
4 3/4 3/4 3/4 21 6/21 7/21 8/21 9/21 10/21
5 3/5 3/5 4/5 4/5 4/5 22 7/22 8/22 8/22 10/22 10/22
6 3/6 4/6 4/6 5/6 5/6 23 7/23 8/23 9/23 10/23 10/23
7 4/7 4/7 5/7 5/7 5/7 24 7/24 8/24 9/24 10/24 11/24
8 4/8 4/8 5/8 5/8 6/8 25 7/25 8/25 9/25 10/25 11/25
9 4/9 5/9 5/9 6/9 6/9 26 7/26 8/26 9/26 10/26 11/26
10 4/10 5/10 6/10 6/10 7/10 27 7/27 8/27 9/27 11/27 11/27
11 5/11 5/11 6/11 7/11 7/11 28 8/28 9/28 10/28 11/28 12/28
12 5/12 5/12 6/12 7/12 7/12 29 8/29 9/29 10/29 11/29 12/29
13 5/13 6/13 6/13 7/13 8/13 30 8/30 9/30 10/30 11/30 12/30
14 5/14 6/14 7/14 7/14 8/14 31 8/31 9/31 10/31 11/31 12/31
15 5/15 6/15 7/15 8/15 8/15 32 8/32 9/32 10/32 12/32 12/32
16 6/16 6/16 7/16 8/16 9/16 34 8/34 10/34 11/34 12/34 13/34
17 6/17 7/17 7/17 8/17 9/17 36 9/36 10/36 11/36 12/36 13/36
18 6/18 7/18 8/18 9/18 9/18 38 9/38 10/38 11/38 13/38 14/38
19 6/19 7/19 8/19 9/19 9/19 40 9/40 10/40 12/40 13/40 14/40

Approximation
1.52 1.73 1.92 2.15 2.303
for n > 40
n n n n n

113
Non-Parametric Tests Table VI: Critical Values of K-S Test for Two Samples of Unequal Size
One -Sided Test: α = 0.10 0.05 0.25 0.01 0.005
Two -Sided Test: α = 0.20 0.1 0.05 0.25 0.01
n1 = 1 n2 = 9 17/18
10 9/10
n1 = 2 n2 = 3 5/6
4 3/4
5 4/5 4/5
6 5/6 5/6
7 5/7 6/7
8 3/4 7/8 7/8
9 7/9 8/9 8/9
10 7/10 4/5 9/10
n1 = 3 n2 = 4 3/4 3/4
5 2/3 4/5 4/5
6 2/3 2/3 5/6
7 2/3 5/7 6/7 6/7
8 5/8 3/4 3/4 7/8
9 2/3 2/3 7/9 8/9 8/9
10 3/5 7/10 4/5 9/10 9/10
12 7/12 2/3 3/4 5/6 11/12
n1 = 4 n2 = 5 3/5 3/4 4/5 4/5
6 7/12 2/3 3/4 5/6 5/6
7 17/28 5/7 3/4 6/7 6/7
8 5/8 5/8 3/4 7/8 7/8
9 5/9 2/3 3/4 7/9 8/9
10 11/20 13/20 7/10 4/5 4/5
12 7/12 2/3 2/3 3/4 5/6
16 9/16 5/8 11/16 3/4 13/16
n1 = 5 n2 = 6 3/5 2/3 2/3 5/6 5/6
7 4/7 23/35 5/7 29/35 6/7
8 11/20 5/8 47/40 4/5 4/5
9 5/9 3/5 31/45 7/9 4/5
10 1/2 3/5 7/10 7/10 4/5
15 8/15 3/5 2/3 11/15 11/15
20 1/2 11/20 3/5 7/10 3/4
n1 = 6 n2 = 7 23/42 4/7 29/42 5/7 5/6
8 1/2 7/12 2/3 3/4 3/4
9 1/2 5/9 2/3 13/18 7/9
10 1/2 17/30 19/30 7/10 11/15
12 1/2 7/12 7/12 2/3 3/4
18 4/9 5/9 11/18 2/3 13/18
24 11/24 1/2 7/12 5/8 2/3
n1 = 7 n2 = 8 27/56 33/56 5/8 41/56 3/4
9 31/63 5/9 40/63 5/7 47/63
10 33/70 39/70 43/70 7/10 5/7
14 3/7 1/2 4/7 9/14 5/7
28 3/7 13/28 15/28 17/28 9/14
n1 = 8 n2 = 9 4/9 13/24 5/8 2/3 3/4
10 19/40 21/40 23/40 27/40 7/10
12 11/24 1/2 7/12 5/8 2/3
16 7/16 1/2 9/16 5/8 5/8
32 13/32 7/16 1/2 9/16 19/32
n1 = 9 n2 = 10 7/15 1/2 26/45 2/3 31/45
12 4/9 1/2 5/9 11/18 2/3
15 19/45 22/45 8/15 3/5 29/45
18 7/18 4/9 1/2 5/9 11/18
36 13/36 5/12 17/36 19/36 5/9
n1 = 10 n2 = 15 2/5 7/15 1/2 17/30 19/30
20 2/5 9/20 1/2 11/20 3/5
40 7/20 2/5 9/20 1/2
n1 = 12 n2 = 15 23/60 9/20 1/2 11/20 7/12
16 3/8 7/16 23/48 13/24 7/12
18 13/36 5/12 17/36 19/36 5/9
20 11/30 5/12 7/15 31/60 17/30
n1 = 14 n2 = 20 7/20 2/5 13/30 29/60 31/60
n1 = 16 n2 = 20 27/80 33/80 17/40 19/40 41/80
Large-sample 1.07
n1  n 2
1.07
n1  n 2
1.36
n1  n 2
1.52
n1  n 2
1.63
n1  n 2
n1 n 2 n1 n 2 n 1n 2 n1 n 2 n1 n 2
approximation

114
 Table-VII: Critical Values of Mann-Whiney U Test (Lower Critical Values) Appendix

n1 α n2 = 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
0.001 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3
0.005 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4
0.01 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 5 5
2
0.025 3 3 3 3 3 3 4 4 4 5 5 5 5 5 5 6 6 6 6
0.05 3 3 4 4 4 4 5 5 5 5 6 6 7 7 7 7 8 8 8
0.1 3 4 6 5 5 5 6 6 7 7 8 8 8 9 9 10 10 11 11
0.001 6 6 6 6 6 6 6 6 6 6 6 6 6 6 6 7 7 7 7
0.005 6 6 6 6 6 6 6 7 7 7 8 8 8 9 9 9 9 10 10
0.01 6 6 6 6 6 7 7 8 8 8 9 9 9 10 10 11 11 11 12
3
0.025 6 6 7 7 8 8 9 9 10 10 11 11 12 12 13 13 14 14 15
0.05 6 7 7 8 9 9 10 11 11 12 12 13 14 14 15 16 16 17 18
0.1 7 8 8 9 10 11 12 12 13 14 15 16 17 17 18 19 20 21 22
0.001 10 10 10 10 10 10 10 10 11 11 11 12 12 12 13 13 14 14 14
0.005 10 10 10 10 11 11 12 12 13 13 14 14 15 16 16 17 17 18 19
0.01 10 10 10 11 12 12 13 14 14 15 16 16 17 18 18 19 20 20 21
4
0.025 10 10 11 12 13 14 15 15 16 17 18 19 20 21 22 22 23 24 25
0.05 10 11 12 13 14 15 16 17 18 19 20 21 22 23 25 26 27 28 29
0.1 11 12 14 15 16 17 18 20 21 22 23 24 26 27 28 29 31 32 33
0.001 15 15 15 15 15 15 16 17 17 18 18 19 19 20 21 21 22 23 23
0.005 15 15 15 16 17 17 18 19 20 21 22 23 23 24 25 26 27 28 29
0.01 15 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
5
0.025 15 16 17 18 19 21 22 23 24 25 27 28 29 30 31 33 34 35 36
0.05 16 17 18 20 21 22 24 25 27 28 29 31 32 34 35 36 38 39 41
0.1 17 18 20 21 23 24 26 28 29 31 33 34 36 38 39 41 43 44 46
0.001 21 21 21 21 21 21 23 24 25 26 26 27 28 29 30 31 32 33 34
0.005 21 21 22 23 24 25 26 27 28 29 31 32 33 34 35 37 38 39 40
0.01 21 21 23 24 25 26 28 29 30 31 33 34 35 37 38 40 41 42 44
6
0.025 21 23 24 25 27 28 30 32 33 35 36 38 39 41 43 44 46 47 49
0.05 22 24 25 27 29 30 32 34 36 38 39 41 43 45 47 48 50 52 54
0.1 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 56 58 60
0.001 28 28 28 28 29 30 31 32 34 35 36 37 38 39 40 42 43 44 45
0.005 28 28 29 30 32 33 35 36 38 39 41 42 44 45 47 48 50 51 53
0.01 28 29 30 32 33 35 36 38 40 41 43 45 46 48 50 52 53 55 57
7
0.025 28 30 32 34 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63
0.05 29 31 33 35 37 40 42 44 46 48 50 53 55 57 59 62 64 66 68
0.1 30 33 35 37 40 42 45 47 50 52 55 57 60 62 65 67 70 72 75
0.001 36 36 36 37 38 39 41 42 43 45 46 48 49 51 52 54 55 57 58
0.005 36 36 38 39 41 43 44 46 48 50 52 54 55 57 59 61 63 65 67
0.01 36 37 39 41 43 44 46 48 50 52 54 56 59 61 63 65 67 69 71
8
0.025 37 39 41 43 45 47 50 52 54 56 59 61 63 66 68 71 73 75 78
0.05 38 40 42 45 47 50 52 55 57 60 63 65 68 70 73 76 78 81 84
0.1 39 42 44 47 50 53 53 59 61 64 67 70 73 76 79 82 85 88 91
0.001 45 45 45 47 48 49 51 53 54 56 58 60 61 63 65 67 69 71 72
0.005 45 46 47 49 51 53 55 57 59 62 64 66 68 70 73 75 77 79 82
0.01 45 47 49 51 53 55 57 60 62 64 67 69 72 74 77 79 82 84 86
9
0.025 46 48 50 53 56 58 61 63 66 69 72 74 77 80 83 85 88 91 94
0.05 47 50 52 55 58 61 64 67 70 73 76 79 82 85 88 91 94 97 100
0.1 48 51 55 58 61 64 68 71 74 77 81 84 87 91 94 98 101 104 108
0.001 55 55 56 57 59 61 62 64 66 68 70 73 75 77 79 81 83 85 88
0.005 55 56 58 60 62 65 67 69 72 74 77 80 82 85 87 90 93 95 98
0.01 55 57 59 62 64 67 69 72 75 78 80 83 86 89 92 94 97 100 103
10
0.025 56 59 61 64 67 70 73 76 79 82 85 89 92 95 98 101 104 108 111
0.05 57 60 63 67 70 73 76 80 83 87 90 93 97 100 104 107 111 114 118
0.1 59 62 66 69 73 77 80 84 88 92 95 99 103 107 110 114 118 122 126

115
Non-Parametric Tests

n1 α n2 = 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
0.001 66 66 67 69 71 73 75 77 79 82 84 87 89 91 94 96 99 101 104
0.005 66 67 69 72 74 77 80 83 85 88 91 94 97 100 103 106 109 112 115
0.01 66 68 71 74 76 79 82 85 89 92 95 98 101 104 108 111 114 117 120
11
0.025 67 70 73 76 80 83 86 90 93 97 100 104 107 111 114 118 122 125 129
0.05 68 72 75 79 83 86 90 94 98 101 105 109 113 117 121 124 128 132 136
0.1 70 74 78 82 86 90 94 98 103 107 111 115 119 124 128 132 136 140 145
0.001 78 78 79 81 83 86 88 91 93 96 98 102 104 106 110 113 116 118 121
0.005 78 80 82 85 88 91 94 97 100 103 106 110 113 116 120 123 126 130 133
0.01 78 81 84 87 90 93 96 100 103 107 110 114 117 121 125 128 132 135 139
12
0.025 80 83 86 90 93 97 101 105 108 112 116 120 124 128 132 136 140 144 148
0.05 81 84 88 92 96 100 105 109 111 117 121 126 130 134 139 143 147 151 156
0.1 83 87 91 96 100 105 109 114 118 123 128 132 137 142 146 151 156 160 165
0.001 91 91 93 95 97 100 103 106 109 112 115 118 121 124 127 130 134 137 140
0.005 91 93 95 99 102 105 109 112 116 119 123 126 130 134 137 141 145 149 152
0.01 92 94 97 101 104 108 112 115 119 123 127 131 135 139 143 147 151 155 159
13
0.025 93 96 100 104 108 112 116 120 125 129 133 137 142 146 151 155 159 164 168
0.05 94 98 102 107 111 116 120 125 129 134 139 143 148 153 157 162 167 172 176
0.1 96 101 105 110 115 120 125 130 135 140 145 150 155 160 166 171 176 181 186
0.001 105 105 107 109 112 115 118 121 125 128 131 135 138 142 145 149 152 156 160
0.005 105 107 110 113 117 121 124 128 132 136 140 144 148 152 156 160 164 169 173
0.01 106 108 112 116 119 123 128 132 136 140 144 149 153 157 162 166 171 175 179
14
0.025 107 111 115 119 123 128 132 137 142 146 151 156 161 165 170 175 180 184 189
0.05 109 113 117 122 127 132 137 142 147 152 157 162 167 172 177 183 188 193 198
0.1 110 116 121 126 131 137 142 147 153 158 164 169 175 180 186 191 197 203 208
0.001 120 120 122 125 128 133 135 138 142 145 149 153 157 161 164 168 172 176 180
0.005 120 123 126 129 133 137 141 145 150 154 158 163 167 172 176 181 185 190 194
0.01 121 124 128 132 136 140 145 149 154 158 163 168 172 177 182 187 191 196 201
15
0.025 122 126 131 135 140 145 150 155 160 165 170 175 180 185 191 196 201 206 211
0.05 124 128 133 139 144 149 154 160 165 171 176 182 187 193 198 204 209 215 221
0.1 126 131 137 143 148 154 160 166 172 178 184 189 195 201 207 213 219 225 231
0.001 136 136 139 142 145 148 152 156 160 164 168 172 176 180 185 189 193 197 202
0.005 136 139 142 146 150 155 159 164 168 173 178 182 187 192 197 202 207 211 216
0.01 137 140 144 149 153 158 163 168 173 178 183 188 193 198 203 208 213 219 224
16
0.025 138 143 148 152 158 163 168 174 179 184 190 196 201 207 212 218 223 229 235
0.05 140 145 151 156 162 167 173 179 185 191 197 202 208 214 220 226 232 238 244
0.1 142 148 154 160 166 173 179 185 191 198 204 211 217 223 230 236 243 249 256
0.001 153 154 156 159 163 167 171 175 179 183 188 192 197 201 206 211 215 220 224
0.005 153 156 158 164 169 173 178 183 188 193 198 203 208 214 219 224 229 235 240
0.01 154 158 160 167 172 177 182 187 192 198 203 209 214 220 225 231 236 242 247
17
0.025 156 160 163 171 176 182 188 193 199 205 211 217 223 229 235 241 247 253 259
0.05 157 163 166 174 180 187 193 199 205 211 218 224 231 237 243 250 256 263 269
0.1 160 166 172 179 185 192 199 206 212 219 226 233 239 246 253 260 267 274 281
0.001 171 172 174 178 182 186 190 195 199 204 209 214 218 223 228 233 238 243 248
0.005 171 174 176 183 188 193 198 203 209 214 219 225 230 236 242 247 253 259 264
0.01 172 176 179 186 191 196 202 208 213 219 225 231 237 242 248 254 260 266 272
18
0.025 174 179 181 190 196 202 208 214 220 227 233 239 246 252 258 265 271 278 284
0.05 176 181 185 194 200 207 213 220 227 233 240 247 254 260 267 274 281 288 295
0.1 178 185 191 199 206 213 220 227 234 241 249 256 263 270 278 285 292 300 307
0.001 190 191 194 198 202 206 211 216 220 225 231 236 241 246 251 257 262 268 243
0.005 191 194 195 203 208 213 219 224 230 236 242 248 254 260 265 272 278 284 290
0.01 192 195 198 206 211 217 223 229 235 241 247 254 260 266 273 279 285 292 298
19
0.025 193 198 201 210 216 223 229 236 243 249 256 263 269 276 283 290 297 304 310
0.05 195 201 205 214 221 228 235 242 249 256 263 271 278 285 292 300 307 314 321
0.1 198 205 211 219 227 234 242 249 257 264 272 280 288 295 303 311 319 326 334
0.001 210 211 214 218 223 227 232 237 243 248 253 259 265 270 276 281 287 293 299
0.005 211 214 219 224 229 235 241 247 253 259 265 271 278 284 290 297 303 310 316
0.01 212 216 221 227 233 239 245 251 258 264 271 278 284 291 298 304 311 318 325
20
0.025 213 219 225 231 238 245 2512 259 266 273 280 287 294 301 309 316 323 330 338
0.05 215 222 229 236 243 250 258 265 273 280 288 295 303 311 318 326 334 341 349
0.1 218 226 233 241 249 257 265 273 281 289 297 305 313 321 330 338 346 354 362
116
 (Upper Critical Values) Appendix

n1 α n2 = 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
0.001 10 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43
0.005 10 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 40 42
0.01 10 9 11 13 15 17 19 21 23 25 27 28 30 32 34 36 38 39 41
2
0.025 10 9 11 13 15 17 18 20 22 23 25 27 29 31 33 34 36 38 40
0.05 10 9 10 12 14 16 17 19 21 23 24 26 27 29 31 33 34 36 38
0.1 10 8 8 11 13 15 16 18 19 21 22 24 26 27 29 30 32 33 35
0.001 18 15 18 21 24 27 30 33 36 39 42 45 48 51 54 56 59 62 65
0.005 18 15 18 21 24 27 30 32 35 38 40 43 46 48 51 54 57 59 62
0.01 18 15 18 21 24 26 29 31 34 37 39 42 45 47 50 52 55 58 60
3
0.025 18 15 17 20 22 25 27 30 32 35 37 40 42 45 47 50 52 55 57
0.05 18 14 17 19 21 24 26 28 31 33 36 38 40 43 45 47 50 52 54
0.1 18 13 16 18 20 22 24 27 29 31 33 35 37 40 42 44 46 48 50
0.001 28 22 26 30 34 38 42 46 49 53 57 60 64 68 71 75 78 82 86
0.005 28 22 26 30 33 37 40 44 47 51 54 58 61 64 68 71 75 78 81
0.01 28 22 26 29 32 36 39 42 46 49 52 56 59 62 66 69 72 76 79
4
0.025 28 22 25 28 31 34 37 41 44 47 50 53 56 59 62 66 69 72 75
0.05 28 21 24 27 30 33 36 39 42 45 48 51 54 57 59 62 65 68 71
0.1 28 20 22 25 28 31 34 36 39 42 45 48 50 53 56 59 61 64 67
0.001 40 30 35 40 45 50 54 58 63 67 72 76 81 85 89 94 98 102 107
0.005 40 30 35 39 43 48 52 56 60 64 68 72 77 81 85 89 93 97 101
0.01 40 30 34 38 42 46 50 54 58 62 66 70 74 78 82 86 90 94 98
5
0.025 40 29 33 37 41 44 48 52 56 60 63 67 71 75 79 82 86 90 94
0.05 40 28 32 35 39 43 46 50 53 57 61 64 68 71 75 79 82 86 89
0.1 40 27 30 34 37 41 44 47 51 54 57 61 64 67 71 74 77 81 84
0.001 54 39 45 51 57 63 67 72 77 82 88 93 98 103 108 113 118 123 128
0.005 54 39 44 49 54 59 64 69 74 79 83 88 93 98 103 107 112 117 122
0.01 54 39 43 48 53 58 62 67 72 77 81 86 91 95 100 104 109 114 118
6
0.025 54 37 42 47 51 56 60 64 69 73 78 82 87 91 95 100 104 109 113
0.05 54 36 41 45 49 54 58 62 66 70 75 79 83 87 91 96 100 104 108
0.1 54 35 39 43 47 51 55 59 63 67 71 75 79 83 87 91 94 98 102
0.001 70 49 56 63 69 75 81 87 92 98 104 110 116 122 128 133 139 145 151
0.005 70 49 55 61 66 72 77 83 88 94 99 105 110 116 121 127 132 138 143
0.01 70 48 54 59 65 70 76 81 86 92 97 102 108 113 118 123 129 134 139
7
0.025 70 47 52 57 63 68 73 78 83 88 93 98 103 108 113 118 123 128 133
0.05 70 46 51 56 61 65 70 75 80 85 90 94 99 104 109 113 118 123 128
0.1 70 44 49 54 58 63 67 72 76 81 85 90 94 99 103 108 112 117 121
0.001 88 60 68 75 82 89 95 102 109 115 122 128 135 141 148 154 161 167 174
0.005 88 60 66 73 79 85 92 98 104 110 116 122 129 135 141 147 153 159 165
0.01 88 59 65 71 77 84 90 96 102 108 114 120 125 131 137 143 149 155 161
8
0.025 88 57 63 69 75 81 86 92 98 104 109 115 121 126 132 137 143 149 154
0.05 88 56 62 67 73 78 84 89 95 100 105 111 116 122 127 132 138 143 148
0.1 88 54 60 65 70 75 83 85 91 96 101 106 111 116 121 126 131 136 141
0.001 108 72 81 88 96 104 111 118 126 133 140 147 155 162 169 176 183 190 198
0.005 108 71 79 86 93 100 107 114 121 127 134 141 148 155 161 168 175 182 188
0.01 108 70 77 84 91 98 105 111 118 125 131 138 144 151 157 164 170 177 184
9
0.025 108 69 76 82 88 95 101 108 114 120 126 133 139 145 151 158 164 170 176
0.05 108 67 74 80 86 92 98 104 110 116 122 128 134 140 146 152 158 164 170
0.1 108 66 71 77 83 89 94 100 106 112 117 123 129 134 140 145 151 157 162
0.001 130 85 94 103 111 119 128 136 144 152 160 167 175 183 191 199 207 215 222
0.005 130 84 92 100 108 115 123 131 138 146 153 160 168 175 183 190 197 205 212
0.01 130 83 91 98 106 113 121 128 135 142 150 157 164 171 178 186 193 200 207
10
0.025 130 81 89 96 103 110 117 124 131 138 145 151 158 165 172 179 186 192 199
0.05 130 80 87 93 100 107 114 120 127 133 140 147 153 160 166 173 179 186 192
0.1 130 78 84 91 97 103 110 116 122 128 135 141 147 153 160 166 172 178 184

117
Non-Parametric Tests
n1 α n2= 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
0.001 154 99 109 118 127 136 145 154 163 171 180 188 197 206 214 223 231 240 248
0.005 154 98 107 115 124 132 140 148 157 165 173 181 189 197 205 213 221 229 237
0.01 154 97 105 113 122 130 138 146 153 161 169 177 185 193 200 208 216 224 232
11
0.025 154 95 103 111 118 126 134 141 149 156 164 171 179 186 194 201 208 216 223
0.05 154 93 101 108 115 123 130 137 144 152 159 166 173 180 187 195 202 209 216
0.1 154 91 98 105 112 119 126 133 139 146 153 160 167 173 180 187 194 201 207
0.001 180 114 125 135 145 154 164 173 183 192 202 210 220 230 238 247 256 266 275
0.005 180 112 122 131 140 149 158 167 176 185 194 202 211 220 228 237 246 254 263
0.01 180 111 120 129 138 147 156 164 173 181 190 198 207 215 223 232 240 249 257
12
0.025 180 109 118 126 135 143 151 159 168 176 184 192 200 208 216 224 232 240 248
0.05 180 108 116 124 132 140 147 155 165 171 179 186 194 202 209 217 225 233 240
0.1 180 105 113 120 128 135 143 150 158 165 172 180 187 194 202 209 216 224 231
0.001 208 130 141 152 163 173 183 193 203 213 223 233 243 253 263 273 282 292 302
0.005 208 128 139 148 158 168 177 187 196 206 215 225 234 243 253 262 271 280 290
0.01 208 127 137 146 156 165 174 184 193 202 211 220 229 238 247 256 265 274 283
13
0.025 208 125 134 143 152 161 170 179 187 196 205 214 222 231 239 248 257 265 274
0.05 208 123 132 140 149 157 166 174 183 191 199 208 216 224 233 241 249 257 266
0.1 208 120 129 137 145 153 161 169 177 185 193 201 209 217 224 232 240 248 256
0.001 238 147 159 171 182 193 204 215 225 236 247 257 268 278 289 299 310 320 330
0.005 238 145 156 167 177 187 198 208 218 228 238 248 258 268 278 288 298 307 317
0.01 238 144 154 164 175 185 194 204 214 224 234 243 253 263 272 282 291 301 311
14
0.025 238 141 151 161 171 180 190 199 208 218 227 236 245 255 264 273 282 292 301
0.05 238 139 149 158 167 176 185 194 203 212 221 230 239 248 257 265 274 283 292
0.1 238 136 145 154 163 171 180 189 197 206 214 223 231 240 248 257 265 273 282
0.001 270 165 178 190 202 212 225 237 248 260 271 282 293 304 316 327 338 349 360
0.005 270 162 174 186 197 208 219 230 240 251 262 272 283 293 304 314 325 335 346
0.01 270 161 172 183 194 205 215 226 236 247 257 267 278 288 298 308 319 329 339
15
0.025 270 159 169 180 190 200 210 220 230 240 250 260 270 280 289 299 309 319 329
0.05 270 157 167 176 186 196 206 215 225 234 244 253 263 272 282 291 301 310 319
0.1 270 154 163 172 182 191 200 209 218 227 236 246 255 264 273 282 291 300 309
0.001 304 184 197 210 223 236 248 260 272 284 296 308 320 332 343 355 367 379 390
0.005 304 181 194 206 218 229 241 252 264 275 286 298 309 320 331 342 353 365 376
0.01 304 180 192 203 215 226 237 248 259 270 281 292 303 314 325 336 347 357 368
16
0.025 304 177 188 200 210 221 232 242 253 264 274 284 295 305 316 326 337 347 357
0.05 304 175 185 196 206 217 227 237 247 257 267 278 288 298 308 318 328 338 348
0.1 304 172 182 192 202 211 221 231 241 250 260 269 279 289 298 308 317 327 336
0.001 340 203 218 232 245 258 271 284 297 310 322 335 347 360 372 384 397 409 422
0.005 340 201 216 227 239 252 264 276 288 300 312 324 336 347 359 371 383 394 406
0.01 340 199 214 224 236 248 260 272 284 295 307 318 330 341 353 364 376 387 399
17
0.025 340 197 211 220 232 243 254 266 277 288 299 310 321 332 343 354 365 376 387
0.05 340 194 208 217 228 238 249 260 271 282 292 303 313 324 335 345 356 366 377
0.1 340 191 202 212 223 233 243 253 264 274 284 294 305 315 325 335 345 355 365
0.001 378 224 240 254 268 282 296 309 323 336 349 362 376 389 402 415 428 441 454
0.005 378 222 238 249 262 275 288 301 313 326 339 351 364 376 388 401 413 425 438
0.01 378 220 235 246 259 272 284 296 309 321 333 345 357 370 382 394 406 418 430
18
0.025 378 217 233 242 254 266 278 290 302 313 325 337 348 360 372 383 395 406 418
0.05 378 215 229 238 250 261 273 284 295 307 318 329 340 352 363 374 385 396 407
0.1 378 211 223 233 244 255 266 277 288 299 309 320 331 342 352 363 374 384 395
0.001 418 246 262 277 292 307 321 335 350 364 377 391 405 419 433 446 460 473 517
0.005 418 243 261 272 286 300 313 327 340 353 366 379 392 405 419 431 444 457 470
0.01 418 242 258 269 283 296 309 322 335 348 361 373 386 399 411 424 437 449 462
19
0.025 418 239 255 265 278 290 303 315 327 340 352 364 377 389 401 413 425 437 450
0.05 418 236 251 261 273 285 297 309 321 333 345 356 368 380 392 403 415 427 439
0.1 418 232 245 256 267 279 290 302 313 325 336 347 358 370 381 392 403 415 426
0.001 460 269 286 302 317 333 348 363 377 392 407 421 435 450 464 479 493 507 521
0.005 460 266 281 296 311 325 339 353 367 381 395 409 422 436 450 463 477 490 504
0.01 460 264 279 293 307 321 332 349 362 376 389 402 416 429 442 456 469 482 495
20
0.025 460 261 275 289 302 315 330 341 354 367 380 393 406 419 431 444 457 470 482
0.05 460 258 271 284 297 310 322 335 347 360 372 385 397 409 422 434 446 459 471
0.1 460 254 267 279 291 303 315 327 339 351 363 375 387 399 410 422 434 446 458

118
 Table VIII: Critical Values of Kruskal-Wallis Test Appendix

n1 n2 n3 α = 0.1 0.05 0.025 0.01

2 2 2 4.571 − − −
3 2 1 4.286 − − −
3 2 2 4.500 4.714 − −
3 3 1 4.571 5.143 − −
3 3 2 4.556 5.361 5.556 −
3 3 3 4.622 5.600 5.956 7.200
4 2 1 4.500 − − −
4 2 2 4.458 5.333 5.500 −
4 3 1 4.056 5.208 5.833 −
4 3 2 4.511 5.444 6.000 6.444
4 3 3 4.709 5.791 6.155 6.745
4 4 1 4.167 4.967 6.167 6.667
4 4 2 4.555 5.455 6.327 7.036
4 4 3 4.545 5.598 6.394 7.144
4 4 4 4.654 5.692 6.615 7.654
5 2 1 4.200 5.000 − −
5 2 2 4.373 5.160 6.000 6.533
5 3 1 4.018 4.960 6.044 −
5 3 2 4.651 5.251 6.004 6.909
5 3 3 4.533 5.648 6.315 7.079
5 4 1 3.987 4.985 5.858 6.955
5 4 2 4.541 5.273 6.068 7.205
5 4 3 4.549 5.656 6.410 7.445
5 4 4 4.668 5.657 6.673 7.760
5 5 1 4.109 5.127 6.000 7.309
5 5 2 4.623 5.338 6.346 7.338
5 5 3 4.545 5.705 6.549 7.578
5 5 4 4.523 5.666 6.760 7.823
5 5 5 4.560 5.780 6.740 8.000

Notes: 1. The critical values in above table are approximated values.

2. When this table is not applicable then we use chi-square table for critical values of Kruskal-Wallis test
which is given in the Appendix at the end of the block 1 of this course.

119
Non-Parametric Tests Table IX: Critical Values of Freidman Test

k=3 k=4 k=5 k=6

n
α = 5% α = 1% α = 5% α = 1% α = 5% α = 1% α = 5% α = 1%
2 − − − 6.00 − − 9.14 9.71
3 − 6.00 7.40 9.00 7.60 8.00 9.86 11.76
4 6.50 8.00 7.80 9.60 8.53 10.13 10.29 12.71
5 6.40 8.40 7.80 9.96 8.80 11.20 10.49 13.23
6 7.00 9.00 7.60 10.20 8.96 11.68 10.57 13.62
7 7.14 8.86 7.80 10.54 9.07 11.87 10.67 13.86
8 6.25 9.00 7.65 10.50 9.14 12.11 10.71 14.00
9 6.22 9.56 7.67 10.73 9.20 13.20 10.78 14.14
10 6.20 9.60 7.68 10.68 9.24 12.44 10.80 14.23
11 6.55 9.46 7.69 10.75 9.28 12.48 10.84 14.32
12 6.50 9.50 7.70 10.80 9.31 12.58 10.86 14.38
13 6.62 9.39 7.80 10.85 9.33 12.60 10.89 14.45
14 6.14 9.14 7.71 10.89 9.35 12.68 10.90 14.49
15 6.40 8.93 7.72 10.92 9.37 12.74 10.92 14.54
16 6.50 9.38 7.80 10.95 9.39 12.80 10.96 14.57
17 6.12 9.29 7.80 10.05 9.40 12.85 10.95 14.61
18 6.33 9.00 7.73 10.93 9.42 12.89 10.95 14.63
19 6.42 9.58 7.86 11.02 9.43 12.88 11.00 14.67
20 6.30 9.30 7.80 11.80 9.40 12.92 11.00 14.66

Note: 1. When this table is not applicable then we use chi-square table for critical values of Freidman test
which is given in the Appendix at the end of the block 1 of this course.

120
 Table X: Poisson Probability Appendix
For a given value of λ entry indicates the probability of obtaining a specified value of X
X λ = 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
0 0.9048 0.8187 0.7408 0.6703 0.6065 0.5488 0.4966 0.4493 0.4066 0.3679
1 0.0905 0.1637 0.2222 0.2681 0.3033 0.3293 0.3476 0.3595 0.3659 0.3679
2 0.0045 0.0164 0.0333 0.0536 0.0758 0.0988 0.1217 0.1438 0.1647 0.1839
3 0.0002 0.0011 0.0033 0.0072 0.0126 0.0198 0.0284 0.0383 0.0494 0.0613
4 0 0.0001 0.0003 0.0007 0.0016 0.003 0.005 0.0077 0.0111 0.0153
5 0 0 0 0.0001 0.0002 0.0004 0.0007 0.0012 0.002 0.0031
6 0 0 0 0 0 0 0.0001 0.0002 0.0003 0.0005
7 0 0 0 0 0 0 0 0 0 0.0001
X λ = 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0
0 0.3329 0.3012 0.2725 0.2466 0.2231 0.2019 0.1827 0.1653 0.1496 0.1353
1 0.3662 0.3614 0.3543 0.3452 0.3347 0.323 0.3106 0.2975 0.2842 0.2707
2 0.2014 0.2169 0.2303 0.2417 0.251 0.2584 0.264 0.2678 0.27 0.2707
3 0.0738 0.0867 0.0998 0.1128 0.1255 0.1378 0.1496 0.1607 0.171 0.1804
4 0.0203 0.026 0.0324 0.0395 0.0471 0.0551 0.0636 0.0723 0.0812 0.0902
5 0.0045 0.0062 0.0084 0.0111 0.0141 0.0176 0.0216 0.026 0.0309 0.0361
6 0.0008 0.0012 0.0018 0.0026 0.0035 0.0047 0.0061 0.0078 0.0098 0.012
7 0.0001 0.0002 0.0003 0.0005 0.0008 0.0011 0.0015 0.002 0.0027 0.0034
8 0 0 0.0001 0.0001 0.0001 0.0002 0.0003 0.0005 0.0006 0.0009
9 0 0 0 0 0 0 0.0001 0.0001 0.0001 0.0002
X λ = 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3.0
0 0.1225 0.1108 0.1003 0.0907 0.0821 0.0743 0.0672 0.0608 0.055 0.0498
1 0.2572 0.2438 0.2306 0.2177 0.2052 0.1931 0.1815 0.1703 0.1596 0.1494
2 0.27 0.2681 0.2652 0.2613 0.2565 0.251 0.245 0.2384 0.2314 0.224
3 0.189 0.1966 0.2033 0.209 0.2138 0.2176 0.2205 0.2225 0.2237 0.224
4 0.0992 0.1082 0.1169 0.1254 0.1336 0.1414 0.1488 0.1557 0.1622 0.168
5 0.0417 0.0476 0.0538 0.0602 0.0668 0.0735 0.0804 0.0872 0.094 0.1008
6 0.0146 0.0174 0.0206 0.0241 0.0278 0.0319 0.0362 0.0407 0.0455 0.0504
7 0.0044 0.0055 0.0068 0.0083 0.0099 0.0118 0.0139 0.0163 0.0188 0.0216
8 0.0011 0.0015 0.0019 0.0025 0.0031 0.0038 0.0047 0.0057 0.0068 0.0081
9 0.0003 0.0004 0.0005 0.0007 0.0009 0.0011 0.0014 0.0018 0.0022 0.0027
10 0.0001 0.0001 0.0001 0.0002 0.0002 0.0003 0.0004 0.0005 0.0006 0.0008
11 0 0 0 0 0 0.0001 0.0001 0.0001 0.0002 0.0002
12 0 0 0 0 0 0 0 0 0 0.0001
X λ = 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 4.0
0 0.045 0.0408 0.0369 0.0334 0.0302 0.0273 0.0247 0.0224 0.0202 0.0183
1 0.1397 0.1304 0.1217 0.1135 0.1057 0.0984 0.0915 0.085 0.0789 0.0733
2 0.2165 0.2087 0.2008 0.1929 0.185 0.1771 0.1692 0.1615 0.1539 0.1465
3 0.2237 0.2226 0.2209 0.2186 0.2158 0.2125 0.2087 0.2046 0.2001 0.1954
4 0.1734 0.1781 0.1823 0.1858 0.1888 0.1912 0.1931 0.1944 0.1951 0.1954
5 0.1075 0.114 0.1203 0.1264 0.1322 0.1377 0.1429 0.1477 0.1522 0.1563
6 0.0555 0.0608 0.0662 0.0716 0.0771 0.0826 0.0881 0.0936 0.0989 0.1042
7 0.0246 0.0278 0.0312 0.0348 0.0385 0.0425 0.0466 0.0508 0.0551 0.0595
8 0.0095 0.0111 0.0129 0.0148 0.0169 0.0191 0.0215 0.0241 0.0269 0.0298
9 0.0033 0.004 0.0047 0.0056 0.0066 0.0076 0.0089 0.0102 0.0116 0.0132
10 0.001 0.0013 0.0016 0.0019 0.0023 0.0028 0.0033 0.0039 0.0045 0.0053
11 0.0003 0.0004 0.0005 0.0006 0.0007 0.0009 0.0011 0.0013 0.0016 0.0019
12 0.0001 0.0001 0.0001 0.0002 0.0002 0.0003 0.0003 0.0004 0.0005 0.0006
13 0 0 0 0 0.0001 0.0001 0.0001 0.0001 0.0002 0.0002
14 0 0 0 0 0 0 0 0 0 0.0001
X λ = 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 5.0
0 0.0166 0.015 0.0136 0.0123 0.0111 0.0101 0.0091 0.0082 0.0074 0.0067
1 0.0679 0.063 0.0583 0.054 0.05 0.0462 0.0427 0.0395 0.0365 0.0337
2 0.1393 0.1323 0.1254 0.1188 0.1125 0.1063 0.1005 0.0948 0.0894 0.0842
3 0.1904 0.1852 0.1798 0.1743 0.1687 0.1631 0.1574 0.1517 0.146 0.1404
4 0.1951 0.1944 0.1933 0.1917 0.1898 0.1875 0.1849 0.182 0.1789 0.1755
5 0.16 0.1633 0.1662 0.1687 0.1708 0.1725 0.1738 0.1747 0.1753 0.1755
6 0.1093 0.1143 0.1191 0.1237 0.1281 0.1323 0.1362 0.1398 0.1432 0.1462
7 0.064 0.0686 0.0732 0.0778 0.0824 0.0869 0.0914 0.0959 0.1002 0.1044
8 0.0328 0.036 0.0393 0.0428 0.0463 0.05 0.0537 0.0575 0.0614 0.0653
9 0.015 0.0168 0.0188 0.0209 0.0232 0.0255 0.028 0.0307 0.0334 0.0363
10 0.0061 0.0071 0.0081 0.0092 0.0104 0.0118 0.0132 0.0147 0.0164 0.0181
11 0.0023 0.0027 0.0032 0.0037 0.0043 0.0049 0.0056 0.0064 0.0073 0.0082
12 0.0008 0.0009 0.0011 0.0014 0.0016 0.0019 0.0022 0.0026 0.003 0.0034
13 0.0002 0.0003 0.0004 0.0005 0.0006 0.0007 0.0008 0.0009 0.0011 0.0013
14 0.0001 0.0001 0.0001 0.0001 0.0002 0.0002 0.0003 0.0003 0.0004 0.0005
15 0 0 0 0 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002

121
Non-Parametric Tests
X λ = 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 5.9 6.0
0 0.0061 0.0055 0.005 0.0045 0.0041 0.0037 0.0033 0.003 0.0027 0.0025
1 0.0311 0.0287 0.0265 0.0244 0.0225 0.0207 0.0191 0.0176 0.0162 0.0149
2 0.0793 0.0746 0.0701 0.0659 0.0618 0.058 0.0544 0.0509 0.0477 0.0446
3 0.1348 0.1293 0.1239 0.1185 0.1133 0.1082 0.1033 0.0985 0.0938 0.0892
4 0.1719 0.1681 0.1641 0.16 0.1558 0.1515 0.1472 0.1428 0.1383 0.1339
5 0.1753 0.1748 0.174 0.1728 0.1714 0.1697 0.1678 0.1656 0.1632 0.1606
6 0.149 0.1515 0.1537 0.1555 0.1571 0.1584 0.1594 0.1601 0.1605 0.1606
7 0.1086 0.1125 0.1163 0.12 0.1234 0.1267 0.1298 0.1326 0.1353 0.1377
8 0.0692 0.0731 0.0771 0.081 0.0849 0.0887 0.0925 0.0962 0.0998 0.1033
9 0.0392 0.0423 0.0454 0.0486 0.0519 0.0552 0.0586 0.062 0.0654 0.0688
10 0.02 0.022 0.0241 0.0262 0.0285 0.0309 0.0334 0.0359 0.0386 0.0413
11 0.0093 0.0104 0.0116 0.0129 0.0143 0.0157 0.0173 0.019 0.0207 0.0225
12 0.0039 0.0045 0.0051 0.0058 0.0065 0.0073 0.0082 0.0092 0.0102 0.0113
13 0.0015 0.0018 0.0021 0.0024 0.0028 0.0032 0.0036 0.0041 0.0046 0.0052
14 0.0006 0.0007 0.0008 0.0009 0.0011 0.0013 0.0015 0.0017 0.0019 0.0022
15 0.0002 0.0002 0.0003 0.0003 0.0004 0.0005 0.0006 0.0007 0.0008 0.0009
16 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002 0.0002 0.0002 0.0003 0.0003
17 0 0 0 0 0 0 0.0001 0.0001 0.0001 0.0001
X λ = 6.1 6.2 6.3 6.4 6.5 6.6 6.7 6.8 6.9 7.0
0 0.0022 0.002 0.0018 0.0017 0.0015 0.0014 0.0012 0.0011 0.001 0.0009
1 0.0137 0.0126 0.0116 0.0106 0.0098 0.009 0.0082 0.0076 0.007 0.0064
2 0.0417 0.039 0.0364 0.034 0.0318 0.0296 0.0276 0.0258 0.024 0.0223
3 0.0848 0.0806 0.0765 0.0726 0.0688 0.0652 0.0167 0.0584 0.0552 0.0521
4 0.1294 0.1249 0.1205 0.1162 0.1118 0.1076 0.1034 0.0992 0.0952 0.0912
5 0.1579 0.1549 0.1519 0.1487 0.1454 0.142 0.1385 0.1349 0.1314 0.1277
6 0.1605 0.1601 0.1595 0.1586 0.1575 0.1562 0.1546 0.1529 0.1511 0.149
7 0.1399 0.1418 0.1435 0.145 0.1462 0.1472 0.148 0.1486 0.1489 0.149
8 0.1066 0.1099 0.113 0.116 0.1188 0.1215 0.124 0.1263 0.1284 0.1304
9 0.0723 0.0757 0.0791 0.0825 0.0858 0.0891 0.0923 0.0954 0.0985 0.1014
10 0.0441 0.0469 0.0498 0.0528 0.0558 0.0588 0.0618 0.0649 0.0679 0.071
11 0.0245 0.0265 0.0285 0.0307 0.033 0.0353 0.0377 0.0401 0.0426 0.0452
12 0.0124 0.0137 0.015 0.0164 0.0179 0.0194 0.021 0.0227 0.0245 0.0264
13 0.0058 0.0065 0.0073 0.0081 0.0089 0.0098 0.0108 0.0119 0.013 0.0142
14 0.0025 0.0029 0.0033 0.0037 0.0041 0.0046 0.0052 0.0058 0.0064 0.0071
15 0.001 0.0012 0.0014 0.0016 0.0018 0.002 0.0023 0.0026 0.0029 0.0033
16 0.0004 0.0005 0.0005 0.0006 0.0007 0.0008 0.001 0.0011 0.0013 0.0014
17 0.0001 0.0002 0.0002 0.0002 0.0003 0.0003 0.0004 0.0004 0.0005 0.0006
18 0 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002 0.0002 0.0002
19 0 0 0 0 0 0 0 0.0001 0.0001 0.0001
X λ = 7.1 7.2 7.3 7.4 7.5 7.6 7.7 7.8 7.9 8.0
0 0.0008 0.0007 0.0007 0.0006 0.0006 0.0005 0.0005 0.0004 0.0004 0.0003
1 0.0059 0.0054 0.0049 0.0045 0.0041 0.0038 0.0035 0.0032 0.0029 0.0027
2 0.0208 0.0194 0.018 0.0167 0.0156 0.0145 0.0134 0.0125 0.0116 0.0107
3 0.0492 0.0464 0.0438 0.0413 0.0389 0.0366 0.0345 0.0324 0.0305 0.0286
4 0.0874 0.0836 0.0799 0.0764 0.0729 0.0696 0.0663 0.0632 0.0602 0.0573
5 0.1241 0.1204 0.1167 0.113 0.1094 0.1057 0.1021 0.0986 0.0951 0.0916
6 0.1468 0.1445 0.142 0.1394 0.1367 0.1339 0.1311 0.1282 0.1252 0.1221
7 0.1489 0.1486 0.1481 0.1474 0.1465 0.1454 0.1442 0.1428 0.1413 0.1396
8 0.1321 0.1337 0.1351 0.1363 0.1373 0.1382 0.1388 0.1392 0.1395 0.1396
9 0.1042 0.107 0.1096 0.1121 0.1144 0.1167 0.1187 0.1207 0.1224 0.1241
10 0.074 0.077 0.08 0.0829 0.0858 0.0887 0.0914 0.0941 0.0967 0.0993
11 0.0478 0.0504 0.0531 0.0558 0.0585 0.0613 0.064 0.0667 0.0695 0.0722
12 0.0283 0.0303 0.0323 0.0344 0.0366 0.0388 0.0411 0.0434 0.0457 0.0481
13 0.0154 0.0168 0.0181 0.0196 0.0211 0.0227 0.0243 0.026 0.0278 0.0296
14 0.0078 0.0086 0.0095 0.0104 0.0113 0.0123 0.0134 0.0145 0.0157 0.0169
15 0.0037 0.0041 0.0046 0.0051 0.0057 0.0062 0.0069 0.0075 0.0083 0.009
16 0.0016 0.0019 0.0021 0.0024 0.0026 0.003 0.0033 0.0037 0.0041 0.0045
17 0.0007 0.0008 0.0009 0.001 0.0012 0.0013 0.0015 0.0017 0.0019 0.0021
18 0.0003 0.0003 0.0004 0.0004 0.0005 0.0006 0.0006 0.0007 0.0008 0.0009
19 0.0001 0.0001 0.0001 0.0002 0.0002 0.0002 0.0003 0.0003 0.0003 0.0004
20 0 0 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002
21 0 0 0 0 0 0 0 0 0 0.0001

122
 Appendix

X λ = 8.1 8.2 8.3 8.4 8.5 8.6 8.7 8.8 8.9 9.0
0 0.0003 0.0003 0.0002 0.0002 0.0002 0.0002 0.0002 0.0002 0.0001 0.0001
1 0.0025 0.0023 0.0021 0.0019 0.0017 0.0016 0.0014 0.0013 0.0012 0.0011
2 0.01 0.0092 0.0086 0.0079 0.0074 0.0068 0.0063 0.0058 0.0054 0.005
3 0.0269 0.0252 0.0237 0.0222 0.0208 0.0195 0.0183 0.0171 0.016 0.015
4 0.0544 0.0517 0.0491 0.0466 0.0443 0.042 0.0398 0.0377 0.0357 0.0337
5 0.0882 0.0849 0.0816 0.0784 0.0752 0.0722 0.0692 0.0663 0.0635 0.0607
6 0.1191 0.116 0.1128 0.1097 0.1066 0.1034 0.1003 0.0972 0.0941 0.0911
7 0.1378 0.1358 0.1338 0.1317 0.1294 0.1271 0.1247 0.1222 0.1197 0.1171
8 0.1395 0.1392 0.1388 0.1382 0.1375 0.1366 0.1356 0.1344 0.1332 0.1318
9 0.1256 0.1269 0.128 0.129 0.1299 0.1306 0.1311 0.1315 0.1317 0.1318
10 0.1017 0.104 0.1063 0.1084 0.1104 0.1123 0.114 0.1157 0.1172 0.1186
11 0.0749 0.0776 0.0802 0.0828 0.0853 0.0878 0.0902 0.0925 0.0948 0.097
12 0.0505 0.053 0.0555 0.0579 0.0604 0.0629 0.0654 0.0679 0.0703 0.0728
13 0.0315 0.0334 0.0354 0.0374 0.0395 0.0416 0.0438 0.0459 0.0481 0.0504
14 0.0182 0.0196 0.021 0.0225 0.024 0.0256 0.0272 0.0289 0.0306 0.0324
15 0.0098 0.0107 0.0116 0.0126 0.0136 0.0147 0.0158 0.0169 0.0182 0.0194
16 0.005 0.0055 0.006 0.0066 0.0072 0.0079 0.0086 0.0093 0.0101 0.0109
17 0.0024 0.0026 0.0029 0.0033 0.0036 0.004 0.0044 0.0048 0.0053 0.0058
18 0.0011 0.0012 0.0014 0.0015 0.0017 0.0019 0.0021 0.0024 0.0026 0.0029
19 0.0005 0.0005 0.0006 0.0007 0.0008 0.0009 0.001 0.0011 0.0012 0.0014
20 0.0002 0.0002 0.0002 0.0003 0.0003 0.0004 0.0004 0.0005 0.0005 0.0006
21 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002 0.0002 0.0002 0.0002 0.0003
22 0 0 0 0 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001
X λ = 9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 9.9 10.0
0 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0
1 0.001 0.0009 0.0009 0.0008 0.0007 0.0007 0.0006 0.0005 0.0005 0.0005
2 0.0046 0.0043 0.004 0.0037 0.0034 0.0031 0.0029 0.0027 0.0025 0.0023
3 0.014 0.0131 0.0123 0.0115 0.0107 0.01 0.0093 0.0087 0.0081 0.0076
4 0.0319 0.0302 0.0285 0.0269 0.0254 0.024 0.0226 0.0213 0.0201 0.0189
5 0.0581 0.0555 0.053 0.0506 0.0483 0.046 0.0439 0.0418 0.0398 0.0378
6 0.0881 0.0851 0.0822 0.0793 0.0764 0.0736 0.0709 0.0682 0.0656 0.0631
7 0.1145 0.1118 0.1091 0.1064 0.1037 0.101 0.0982 0.0955 0.0928 0.0901
8 0.1302 0.1286 0.1269 0.1251 0.1232 0.1212 0.1191 0.117 0.1148 0.1126
9 0.1317 0.1315 0.1311 0.1306 0.13 0.1293 0.1284 0.1274 0.1263 0.1251
10 0.1198 0.121 0.1219 0.1228 0.1235 0.1241 0.1245 0.1249 0.125 0.1251
11 0.0991 0.1012 0.1031 0.1049 0.1067 0.1083 0.1098 0.1112 0.1125 0.1137
12 0.0752 0.0776 0.0799 0.0822 0.0844 0.0866 0.0888 0.0908 0.0928 0.0948
13 0.0526 0.0549 0.0572 0.0594 0.0617 0.064 0.0662 0.0685 0.0707 0.0729
14 0.0342 0.0361 0.038 0.0399 0.0419 0.0439 0.0459 0.0479 0.05 0.0521
15 0.0208 0.0221 0.0235 0.025 0.0265 0.0281 0.0297 0.0313 0.033 0.0347
16 0.0118 0.0127 0.0137 0.0147 0.0157 0.0168 0.018 0.0192 0.0204 0.0217
17 0.0063 0.0069 0.0075 0.0081 0.0088 0.0095 0.0103 0.0111 0.0119 0.0128
18 0.0032 0.0035 0.0039 0.0042 0.0046 0.0051 0.0055 0.006 0.0065 0.0071
19 0.0015 0.0017 0.0019 0.0021 0.0023 0.0026 0.0028 0.0031 0.0034 0.0037
20 0.0007 0.0008 0.0009 0.001 0.0011 0.0012 0.0014 0.0015 0.0017 0.0019
21 0.0003 0.0003 0.0004 0.0004 0.0005 0.0006 0.0006 0.0007 0.0008 0.0009
22 0.0001 0.0001 0.0002 0.0002 0.0002 0.0002 0.0003 0.0003 0.0004 0.0004
23 0 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002 0.0002
24 0 0 0 0 0 0 0 0.0001 0.0001 0.0001
X λ = 11 12 13 14 15 16 17 18 19 20.0
0 0 0 0 0 0 0 0 0 0 0
1 0.0002 0.0001 0 0 0 0 0 0 0 0
2 0.001 0.0004 0.0002 0.0001 0 0 0 0 0 0
3 0.0037 0.0018 0.0008 0.0004 0.0002 0.0001 0 0 0 0
4 0.0102 0.0053 0.0027 0.0013 0.0006 0.0003 0.0001 0.0001 0 0
5 0.0224 0.0127 0.007 0.0037 0.0019 0.001 0.0005 0.0002 0.0001 0.0001
6 0.0411 0.0255 0.0152 0.0087 0.0048 0.0026 0.0014 0.0007 0.0004 0.0002
7 0.0646 0.0437 0.0281 0.0174 0.0104 0.006 0.0034 0.0018 0.001 0.0005
8 0.0888 0.0655 0.0457 0.0304 0.0194 0.012 0.0072 0.0042 0.0024 0.0013
9 0.1085 0.0874 0.0661 0.0473 0.0324 0.0213 0.0135 0.0083 0.005 0.0029
10 0.1194 0.1048 0.0859 0.0663 0.0486 0.0341 0.023 0.015 0.0095 0.0058
11 0.1194 0.1144 0.1015 0.0844 0.0663 0.0496 0.0355 0.0245 0.0164 0.0106
12 0.1094 0.1144 0.1099 0.0984 0.0829 0.0661 0.0504 0.0368 0.0259 0.0176
13 0.0926 0.1056 0.1099 0.106 0.0956 0.0814 0.0658 0.0509 0.0378 0.0271
14 0.0728 0.0905 0.1021 0.106 0.1024 0.093 0.08 0.0655 0.0514 0.0387

123
Non-Parametric Tests
X λ = 11 12 13 14 15 16 17 18 19 20.0
15 0.0534 0.0724 0.0885 0.0989 0.1024 0.0992 0.0906 0.0786 0.065 0.0516
16 0.0367 0.0543 0.0719 0.0866 0.096 0.0992 0.0963 0.0884 0.0772 0.0646
17 0.0237 0.0383 0.055 0.0713 0.0847 0.0934 0.0963 0.0936 0.0863 0.076
18 0.0145 0.0256 0.0397 0.0554 0.0706 0.083 0.0909 0.0936 0.0911 0.0844
19 0.0084 0.0161 0.0272 0.0409 0.0557 0.0699 0.0814 0.0887 0.0911 0.0888
20 0.0046 0.0097 0.0177 0.0286 0.0418 0.0559 0.0692 0.0798 0.0866 0.0888
21 0.0024 0.0055 0.0109 0.0191 0.0299 0.0426 0.056 0.0684 0.0783 0.0846
22 0.0012 0.003 0.0065 0.0121 0.0204 0.031 0.0433 0.056 0.0676 0.0769
23 0.0006 0.0016 0.0037 0.0074 0.0133 0.0216 0.032 0.0438 0.0559 0.0669
24 0.0003 0.0008 0.002 0.0043 0.0083 0.0144 0.0226 0.0328 0.0442 0.0557
25 0.0001 0.0004 0.001 0.0024 0.005 0.0092 0.0154 0.0237 0.0336 0.0446
26 0 0.0002 0.0005 0.0013 0.0029 0.0057 0.0101 0.0164 0.0246 0.0343
27 0 0.0001 0.0002 0.0007 0.0016 0.0034 0.0063 0.0109 0.0173 0.0254
28 0 0 0.0001 0.0003 0.0009 0.0019 0.0038 0.007 0.0117 0.0181
29 0 0 0.0001 0.0002 0.0004 0.0011 0.0023 0.0044 0.0077 0.0125
30 0 0 0 0.0001 0.0002 0.0006 0.0013 0.0026 0.0049 0.0083
31 0 0 0 0 0.0001 0.0003 0.0007 0.0015 0.003 0.0054
32 0 0 0 0 0.0001 0.0001 0.0004 0.0009 0.0018 0.0034
33 0 0 0 0 0 0.0001 0.0002 0.0005 0.001 0.002
34 0 0 0 0 0 0 0.0001 0.0002 0.0006 0.0012
35 0 0 0 0 0 0 0 0.0001 0.0003 0.0007
36 0 0 0 0 0 0 0 0.0001 0.0002 0.0004
37 0 0 0 0 0 0 0 0 0.0001 0.0002
38 0 0 0 0 0 0 0 0 0 0.0001
39 0 0 0 0 0 0 0 0 0 0.0001

124