Tests involving Small Sample n < 30

Some times it is not practical or even possible to obtain large samples, cost, available
time and other factors may require us to work with small samples for our purposes we
will say a sample of size less than 30 is a small sample

Assumption for t-test

1. The parent population from which sample is drawn is normal

2. The sample is random (small n<30)
3. Population standard deviation is unkown.

t-test for Single mean

Suppose we want to test if a random sample x1 , x 2 ,...x n of size n is drawn from a
normal population with specified mean  The null hypothesis is written
Ho: μ =  0
H1: μ <  0 (one tail test, left tail)

The t- statistic is given by

x
t (t has student t distribution with n-1df)
Sx / n

d.f.: (no of independent observations in the sample) -

(no of population parameter which must be estimated from sample observations)

n–k here k=1

Example: A machine is producing washers having a thickness of ‘0.5 mm’ To determine

whether the machine is in proper working order, a sample of 10 washers is chosen for
which the mean thickness is 0.53 m.m. and the standard deviation is .03 mm. Test the
hypothesis that the machine is in proper working order using a level of significance (a) .
05 (b) 0.01

Ho: μ = 0.50
H1: μ  0.5 (Two tail test)

The t- statistic is given by

x 0.53  .5
t =  3.16
Sx / n .03 / 10
Example 2:- A test of breaking strength of 6 ropes manufactured by a company showed a
mean breaking strength of 7750 N and a standard deviation of 145 N where as
manufacturer claimed a mean breaking strength of 8000 N. can we support the
manufacturer’s claim at a level of significance 5%

Ho: μ = 8000N
H1: μ < 8000N (left tail test)

The t- statistic is given by

x 7750  .8000

t =  3.86
Sx / n 145 / 6
Two Sample Tests of Means (n1+n2 < 30)

Two sample tests are frequently used to compare two methods of instruction, two brands,
two cities and other similar things. Let the two populations are f ( 1 ,  12 ) and
f (  2 ,  22 ) , here 1 and  2 are the means of the population,  12 and  22 (both
unknown) are the variance of the population.

The null hypothesis can be that two population have equal means
H 0 : 1   2

The alternatives can be one of the following

H 1 : 1   2 or 1   2 or 1   2

The test statistic is given by

x y
t
(n1  1) S  (n 2  1) S 22 1
2
1
1
(  )
n1  n 2  2 n1 n2

The statistic t has a t-distribution with n1+n2-2 d.f., if it can be assumed that both
populations are approximately normal.

Example 1
The I.Q of 16 students from one area of city showed a mean of 107 with a standard
deviation of 10, while the I.Q.’s of 14 students from another area of the city showed a
mean of 112 with a standard deviation of 8. Is there a significant difference between the
I.Q of the groups? Use (a) .01 and (b) .05 level of significance

H 0 : 1   2 , there is no difference between the groups

H 1 : 1   2 , there is significant difference between the groups

Under Ho, the test statistic is

x y 112  107
t  1.45
(n1  1) S  (n 2  1) S 1
2 2
1 = 15(100)  13(64) 1 1
1
(  )
2
(  )
n1  n 2  2 n1 n2 16  14  2 16 14
Accept H0 Reject H0

Example 2:- At an agricultural station it was desired to test the effect of a given fertilizer
on wheat production. To accomplish this, 24 plots of land having equal areas were
chosen, half of there were treated with the fertilizer and other half were untreated.
Otherwise conditions were same. The mean yield of wheat on the untreated plots was 4.8
kgs. With a standard deviation of 4 kgs, while the mean yield on the treated plots was 5.1
kg. with a standard deviation of 3.6 kg. can we conclude that there is significant
improvement in wheat production because of the fertilizer if the significance level of (a)
1 % and (b) 5 % is used.

H 0 : 1   2 , there is no difference between the groups

H 1 : 1   2 , there is significant difference between the groups

Under Ho, the test statistic is

x y 5.1  4.8
t  1.8.5
(n1  1) S  (n 2  1) S 1
2 2
1 = 11(16)  11(3.6) 2 1 1
1 2
(  ) (  )
n1  n 2  2 n1 n2 12  12  2 12 12

Paired t-test for difference of means

(Dependent samples, sample observations are paired)
Let us now consider the care when
(i) sample sizes are equal n1 = n 2 =n (say) and
(ii) the two samples are not independent but the sample observations are paired
together i.e. the pair of observations ( xi , y i ) corresponds the same i th sample
unit. i  1,2,..., n

The problem is to test of the sample means differ significantly.

Many statistical problems use paired data samples to draw conclusions about the
difference between two populations, data pairs occur may naturally in before and after
situations, where the same object or item is measured both before and after a treatment.

Example 1:- The diastolic blood pressure of 12 were taken before and after administering
a certain drug

Blood pressure 94 80 84 86 88 78 91 92 81 90 100 90 before giving drug

Blood pressure after giving 89 78 76 87 91 78 93 91 76 90 96 90 to drug

Can we say that the drug has reduced the blood pressure

In this problem we want to test the claim that drug has reduced the Blood pressure. This
means difference between d i = A-B should tend to be positive and the population mean
difference values  d should also be positive. The null hypothesis that there is no
difference among the pairs i.e.  d = 0 i.e.

H 0 : d  0
H : d  0

To test this hypothesis we use the following test statistic

d  d
t The statistic t has student t-distribution
Sd / n

Patient No. B.P. before B.P. after administering D= A – B D2

drug A the B drug
1 94 89 5 25
2 80 78 2 4
3 84 76 8 64
4 86 87 -1 1
5 88 91 -3 9
6 78 78 0 0
7 91 93 -2 4
8 92 91 1 1
9 81 76 5 25
10 90 90 0 0
11 100 96 4 16
12 96 90 6 36

S d  3.09 d  2.583

d  d 2.583  0
t =  2.89
Sd / n 3.09 / 12

Reject H 0 and conclude that the drug does reduce the blood pressure

Example 2:- An IQ test was administered to 5 persons before and after they were trained.
The results are given below:-

Candidates I II III IV V
IQ. Before training 110 120 123 132 125
IQ. After training 120 118 125 136 121

Test Whether any change in I.Q,. after the training programme.

H 0 : d  0
H : d  0

Candidates IQ. Before IQ after training d = B – A D2

training A B
I 110 120 10 100
II 120 118 -2 4
III 123 125 2 4
IV 132 136 4 16
V 125 121 -4 16

Sd  30 d 2

d  d 20
t =  0.816
Sd / n 5 6/ 5

We accept H 0 and conclude that there is no effect of training.