Prof.

Nandini Rai
KJSCE
 So far we have dealt with the problems of
estimation of a parameter or testing an
hypothesis about a parameter.
 Such tests which deals with the parameter of
the population are called parametric tests.
 On the other hand tests which do not deal
with the parameter are called non
parametric tests.
 One such test, which we are going to study
is 𝜒 2 - test (pronounced as ‘ki’ square test –
‘ki’ as in kite)
Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 2
 Suppose we are given a die and we want to
know whether it is biased or unbiased.
 OR suppose in a cholera epidemic we
inoculate a group and we want to know
whether inoculation is effective in
preventing the attack of cholera.
 In such situations Chi square test is used to
test the hypothesis
 e.g the die is not biased or the inoculation is
not effective.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 3

𝜒 is calculated on the assumption of status
2

quo i,e there is no change.

 To test such a hypothesis we toss the die for
say, 138 times and observe how many times
we get 1, 2, 3, 4, 5, 6.
 We can calculate expected frequencies of 1,
2, 3, 4, 5, 6 in 138 tosses. Then we find the
value of Chi – square from the following.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 4

 Thestatistic 𝜒 2 first used by Karl Pearson is
defined by
2 𝑂−𝐸 2
𝜒 =
𝐸
 Where,O = observed frequency, E = expected
frequency.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 5

We calculate expected frequencies on certain
assumption such as
(i) the coin or a die is unbiased,
(ii) there is no association between the attributes.
(iii) the accident occur evenly on all days.
(iv) digits occur evenly on all pages,
(v) the events occur in the given ratio,
(vi) the events occur according to the given
distribution (Binomial, Poisson, Normal).
This is called testing goodness of fit.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 6

We now compare the calculated value of 𝜒 2
with the table value for the given degrees of
freedom and at a specified level of
significance.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 7

If the calculated value of 𝝌𝟐 is greater than
the table value we conclude that the
difference between the observed values and
expected frequencies is significant and the
hypothesis is rejected.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 8

If on the other hand the calculated value of
𝝌𝟐 is less than the table value, we conclude
that the difference between the observed
values and the expected frequencies is not
significant and the hypothesis is accepted.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 9

Note:
The value of 𝜒 2 will be zero if the observed
and expected frequencies coincide.
The value of 𝜒 2 is always positive.
As observed frequencies depart from expected
frequencies 𝜒 2 goes on increasing.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 10

 N, the total number of observations must be
sufficiently large. Preferably N should be
greater than 50.
 Frequency of every cell must be greater than
10. If any frequency is less than 10 it is
combined with Neighboring frequencies so
that the combined frequency is greater than
10 and the degrees of freedom are reduced
accordingly.
 The number of classes n must not be too
small not too large. Preferably we should
have 4 ≤ 𝑛 ≤ 16.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 11

 It may be noted that the 𝜒 2 test depends
upon
 The observed frequencies
 The expected frequencies,
 The number of observations only.
 It does not make any assumption regarding
the nature of parent population.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 12

 1. To determine association between two or
more attributes:
 𝜒 2 - test is widely used to test whether there is
association between two or more attributes.
 For example, 𝜒 - test can be used to determine
2
whether there is association between the colour
of mother’s eye and daughter’s eye, between
inoculation and prevention of a diseases.
 In such cases we proceed on the null hypothesis
that there is no association between the
attributes. If the calculated value of 𝜒 2 at a
certain level of significance is less than the table
value, the hypothesis is accepted otherwise
rejected.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 13

 In the same way 𝜒 2 test is also used to test if
a characteristic is dependent upon another
characteristic.
 For example, 𝜒 2 - test can be used to test
whether the performance of workers in a
factory is dependent on the training or to
test whether performance of students in a
particular subject is dependent on the
performance in another subject.
 Using 𝜒 2 - distribution in this way to test the
independence of one attribute on another is
called test of independence

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 14

𝜒
2
- test is very commonly known as 𝜒 2 - test
of goodness of fit because it enables us to
ascertain how well the theoretical
distributions such as Binomial, Poisson or
Normal fit the observed frequencies.
 In such cases we proceed on the null
hypothesis that the theory supports the
observations i.e. the fit is good.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 15

 For example, suppose we toss 3 fair coins
200 times and observe the frequencies of
0,1,2,3 heads. We can also calculate the
expected frequencies by using Binomial
Distribution.
 𝜒 2 -test can be used to ascertain whether
Binomial Distribution fits well. If the
calculated value of 𝜒 2 at a certain level of
significance is less than the table value, the
fit is supposed to be good otherwise the fit is
supposed to be poor.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 16

 To test the discrepancies between
observed frequencies and expected
frequencies
 𝜒 2 -test can also be used to ascertain
whether the difference between observed
frequencies and the expected frequencies
is purely due to chance or whether due to
inadequacy in the theory applied.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 17

 To test equality of several proportions:
 𝜒 - test can also be used to test whether the
2

proportions 𝑝1 , 𝑝2 , 𝑝3 , 𝑝4 in different
populations are equal
 i.e. 𝜒 2 test can also be used to test the null
hypothesis that 𝑝1 = 𝑝2 = 𝑝3 = 𝑝4 .

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 18

 Totest the hypothesis about 𝝈𝟐 : 𝜒 2 is also
used to test the population variance.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 19

 The following table gives the number of
accidents in a city during a week. Find
whether the accidents are uniformly
distributed over a week.

Sun Wed Tota

Day: Mon Tue. Thu. Fri. Sat.
. . l
No. of
13 15 9 11 12 10 14 84
Accidents:

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 20

 The null hypothesis 𝐻0 : Accidents are equally
distributed over all days of a week
 Alternative Hypothesis 𝐻𝑎 : Accidents do not
occur equally
 Calculation of test statistic: If the accidents
occur equally on all days of week, there will
be 84 7 = 12 accidents per day

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 21

 𝑂−𝐸 2
2
∴ 𝜒 =
𝐸
13−12 2 15−12 2 9−12 2 11−12 2
= + + + +
12 12 12 12
12−12 2 10−12 2 14−12 2
+ +
12 12 12
= 2.33

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 22

 Level of significance: 𝛼 = 0.05,
 Degrees of freedom = 𝑛 − 1 = 7 − 1 = 6
 Critical value: For 6 degrees of freedom at
5% level of significance table value of 𝜒 2 is
12.59
 Decision: Since the calculated value of 𝜒 2 is
less than the table value.
 The hypothesis is accepted
 ∴ The accidents occur equally on all working
days

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 23

A die was thrown 132 times and the following
frequencies were observed.
Test the hypothesis that the die is unbiased.

No.
1 2 3 4 5 6 Total
obtained
Frequency 15 20 25 15 29 28 132

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 24

 Null hypothesis 𝐻0 : The die is unbiased
 Alternative Hypothesis 𝐻𝑎 : The die is not
unbiased
 Calculation of test statistic: On the
hypothesis that the die is unbiased we should
expect the frequency of each number to be
132 6 = 22
𝑂−𝐸 2
 𝜒2 =
𝐸
= 8.91

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 25

 Level of significance: 𝛼 = 0.05
 Number of degree of freedom = 𝑛 − 1 = 5
 Critical value: For 5 dof. at 5% level of
significance the table value of 𝜒 2 is
 11.07
 Decision: Since the calculated value of
𝜒 2 = 8.91 is less than the table value of
𝜒 2 = 11.07,
 The hypothesis is accepted
 ∴ The die is unbiased

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 26

 300 digits were chosen at random from a
table of random numbers. The frequency of
digits was as follows. Using 𝜒 2 - test examine
the hypothesis that the digits were
distributed in equal numbers in the table.

Digit 0 1 2 3 4 5 6 7 8 9 Tot
al
Freque 28 29 33 31 26 35 32 30 31 25 300
ncy

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 27

 Null hypothesis 𝐻0 : The digits are distributed
equally
 Alternative Hypothesis 𝐻𝑎 : The digits are not
distributed equally
 Calculation of test statistic:
 On the basis of the hypothesis, the frequency
of each digit = total 10 = 30
𝑂−𝐸 2
∴ 𝜒2=
𝐸
= 2.87

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 28

 Level of significance: 𝛼 = 0.05
 degrees of freedom = 9
 Critical value: For 9 d.f. at 5% level of
significance, the table value of 𝜒 2 is
 16.92
 Decision: Since the calculated value of
𝜒 2 = 2.87 is less than the table value of 𝜒 2 ,
 the hypothesis is accepted
 ∴ Digits are equally distributed in the table.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 29

 Theory predicts that the proportion of beans
in the four groups 𝐴, 𝐵, 𝐶, 𝐷 should be
9: 3: 3: 1. In an experiment among 1600 beans
the numbers in the four groups were
882, 313, 287 and 118. Does the Experimental
results support the theory?

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 30

 Null hypothesis 𝐻0 : The proportion of the
beans in the four groups 𝐴, 𝐵, 𝐶, 𝐷 is the
given proportion 9: 3: 3: 1
 Alternative Hypothesis 𝐻𝑎 : The proportion is
not as given above
 Calculation of test statistic: On the basis of
this hypothesis, since the sum is 9 + 3 + 3 +
1 = 16, the number of beans in the four
groups will be

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 31

 9
𝐴 = × 1600 = 900,
16
3
𝐵 = × 1600 = 300,
16
3
𝐶 = × 1600 = 300,
16
1
𝐷 = × 1600 = 100
16
2 𝑂−𝐸 2
∴ 𝜒 = =
𝐸
 4.72

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 32

 Level of significance: 𝛼 = 0.05
 Number of degree of freedom = 3
 Critical value: For 3 degree of freedom at
5% level of significance, the table value of
𝜒 2 is
 7.81
 Decision: Since the calculated value of
𝜒 2 = 4.72 is less than the table value of
𝜒 2 = 7.81,
 the null hypothesis is accepted
 ∴ The proportion 9: 3: 3: 1 is correct

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 33

 Investigatethe association between the
darkness of eye colour in father and son from
the following data.

Colour of father’s eyes

Dark Not Dark Total
Dark 48 90 138
Colour
Not
of son’s 80 782 862
Dark
eyes
Total 128 872 1000
Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 34
 Null hypothesis 𝐻0 : There is no association
between the darkness of eye colour in father
and son
 Alternative Hypothesis 𝐻𝑎 : There is an
association

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 35

 Calculationof test statistic: On the basis of
this hypothesis the expected frequency of
𝐴×𝐵
dark eyed sons with dark eyed fathers =
𝑁
 where, 𝐴 = number of dark eyed
fathers (total of first column)
 𝐵 = number of dark eyed sons
(total of first row)
 𝑁 = total number of observations
128×138
∴ Expected frequency = = 18
1000

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 36

 Having obtained the expected frequency in
the first cell, since the totals remain the
same, the figures in other cells can be easily
obtained as,

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 37

 Colour of father’s eyes

Dark Not Dark Total

Dark 18 120 138

Colour of
Not Dark 110 752 862
son’s eyes
Total 128 872 1000

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 38

 Calculation of 𝑶 − 𝑬 𝟐
𝑬

𝑂 𝐸 𝑂−𝐸 2 𝑂−𝐸 2 𝐸
48 18
80 110
90 120
782 752
Total

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 39

 2
𝜒 =
 66.88
 Level of significance: 𝛼 = 0.05
 Degrees of freedom = 𝑟 − 1 𝑐 − 1 = 1
 Critical value: For 1 d.f. at 5% level of
significant the table value of 𝜒 2 is
 3.84

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 40

 Decision: Since the calculated value of
𝜒 = 66.88 is much greater than the table
2

value of 𝜒 2 = 3.84,
 the hypothesis is rejected
 ∴ There is an association between darkness
of eye colour of fathers and sons

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 41

 The figures given below are (a) the observed
frequencies of a distribution, (b) the frequencies
of the normal distribution, having the same
mean, standard deviation and the total
frequency as in (a)
(a)1, 12, 66, 220, 495, 792, 924, 792, 495, 220, 66, 12, 1
(b)2, 15, 66, 210, 484, 799, 943, 799, 484, 210, 66, 15, 2

Apply 𝜒 2 test of goodness of fit.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 42

  Sincethe frequencies at the beginning and
end are less than 10, we group them and
then apply the 𝜒 2 test

𝑂 1 12 66 220 495 792 924 792 495 220 66 12 1

𝐸 2 15 66 210 484 799 943 799 484 210 66 15 2

0.4 0.2 0.0 0.3 0.0 0.2 0.4
𝑂−𝐸 2 𝐸 0.94 0.0 0 0.94
8 5 6 8 6 5 8

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 43

 Null hypothesis 𝐻0 : The fit is good
 Alternative Hypothesis 𝐻𝑎 : The fit is not good
 Calculation of test statistic: 𝜒 2 = 3.84
 Level of significance: 𝛼 = 0.05

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 44

 Number of degree of freedom: There are
originally 13 classes. Since they are reduced
to 11 by grouping twice, the degrees of
freedom is reduced by 2. Further, since the
mean, the standard deviation and the total
frequency of original data are used, there
constrains are introduced, reducing the
degree of freedom by 3. For calculating the
mean and the standard deviation three sums
𝑓𝑖 𝑓𝑖 𝑥𝑖 and 𝑓𝑖 𝑥𝑖 2 are required. Hence,
the degree of freedom is reduced by 3. Thus,
the d.f. = 13 − 2 − 3 = 8
Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 45
 Critical value: For 8 d.f. at 5% level of
significance, the table value of 𝜒 2 = 15.51
 Decision: Since the calculated value of
𝜒 2 = 3.84 is less than the table value of
𝜒 2 = 15.51,
 the hypothesis is accepted
 ∴ The fit is good

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 46

 In a 2 × 2 table the degrees of freedom is
(2 – 1) (2 − 1) = 1. If any of the cell
frequency is less than 5, we have to use
pooling method. But this will result in 𝜒 2
with zero degrees of freedom. This is
meaningless.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 47

 In this case Yates in 1934 suggested to use
𝑂−𝐸 −0.5 2
𝜒2 =
𝐸
 He showed that by taking 𝜒 2 as defined
above, 𝜒 2 approximation is improved

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 48

 Two batches of 12 animals each are given test of
inoculation. One batch was inoculated and the
other was not. The number of dead and surviving
animals are given in the following table for both
cases. Can the inoculation be regarded as
effective against the disease at 5% level of
significance (Make Yates correction)

Dead Surviving Total

Inoculated 2 10 12
Non- Inoculated 8 4 12
Total 10 14 24
Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 49
 Null hypothesis 𝐻0 : There is no association
between inoculation and death
 Alternative Hypothesis 𝐻𝑎 : There is association
between inoculation and death
 Calculation of test statistic: On the basis of this
𝐴×𝐵
hypothesis the number in the first cell =
𝑁

 where, 𝐴 = total in the first column, 𝐵 = total in

the first row, 𝑁 = Total no of observations
10×12
 ∴ The number in the first cell = =5
24
 apply Yates correction and prepare the table
Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 50
 Calculation of 𝜒 2

𝑂−𝐸 −0.5 2
𝑂 𝐸 𝑂 − 𝐸 − 0.5
𝐸
10×12
2 =5
24
10 12 − 5 = 7
8 10 − 5 = 5
4 12 − 5 = 7

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 51

 𝜒 2 = 4.29
 Level of significance: 𝛼 = 0.05
 Degrees of freedom = 𝑟 − 1 𝑐 − 1 = 1
 Critical value: For 1 degree of freedom at 5%
level of significance, the table value of 𝜒 2 =
 3.81
 Decision: Since the calculated value of 𝜒 2 = 4.29
is greater than the table value of 𝜒 2 = 3.81,
 the hypothesis is rejected.
 There is association between inoculation and
death i.e. inoculation is effective against the
disease

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 52

 Thefollowing mistakes per page were
observed in a book . Fit a Poisson distribution
and test the goodness of fit .

No. of mistakes per

0 1 2 3 4
page

No. of pages 211 90 19 5 0

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 53

 Null hypothesis 𝐻0 : The mistakes follow
Poisson’s distribution. The fit is good
 Alternative Hypothesis 𝐻𝑎 : The mistakes do
not follow Poisson’s distribution
 Calculation of test statistic: The expected
frequencies by Poisson’s distribution are
given by
𝑒 −𝑚 𝑚𝑥
Expected frequency = 𝑁𝑝 = 𝑁 ×
𝑥!

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 54

where, 𝑚 = mean of the distribution,
𝑥 =random variable, 𝑁 =number of
observations
𝑓𝑥
 Here, 𝑚= = 0.44
𝑓
𝑥 = 0, 1, 2, 3, 4; 𝑁 = 325
325×𝑒 −0.44 0.44 𝑥
 Exp. Freq =
𝑥!

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 55

No. of mistakes
0 1 2 3 4
per page

20.2
No. of pages 209.31 92.1 2.97 0.36
6

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 56

 Calculation of 𝑂 − 𝐸 2
𝐸

No of 2 2
𝑂 𝐸 𝑂−𝐸 𝑂−𝐸 𝐸
mistakes
0 211 209.31 2.87 0.014
1 90 92.10 4.41 0.048
2 19 20.26
3 5 2.97 0.17 0.007
4 0 0.36

0.069
Total
Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 57
 𝑂−𝐸 2
∴ 𝜒2 = = 0.069
𝐸
 Level of significance: 𝛼 = 0.05
 Degrees of freedom 5 − 4 = 1
 (The number of degrees of freedom is 1 for
each class. There are 5 classes originally.
Hence, the degrees of freedom originally is
5. But we reduced the classes by two, thus
reduced the degree of freedom by 2. Further,
while calculating the parameter 𝑚, we used
two sums 𝑓𝑖 and 𝑓𝑖 𝑥𝑖 , thus, reducing the
degree of freedom again by 2)

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 58

 Critical value: For 1 degree of freedom at
5% level of significance the table value of 𝜒 2
is 3.84
 Decision: Since the calculated value of
𝜒 2 = 0.069 is less than the table value of
𝜒 2 = 3.84, we accept the hypothesis
 ∴ The mistakes follow Poisson’s distribution.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 59

 Samples of three shipments 𝐴, 𝐵, 𝐶 of
defective items gave the following results.
Test whether the proportion of defective
items is same in the three shipments at 0.05
level of significance

Shipment 𝐴 Shipment 𝐵 Shipment 𝐶 Total

Defective 5 8 9 22
Non-defective 35 42 51 128
Total 40 50 60 150

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 60

 Null hypothesis 𝐻0 : 𝑝1 = 𝑝2 = 𝑝3
 Alternative Hypothesis 𝐻𝑎 : 𝑝1 ≠ 𝑝2 ≠ 𝑝3

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 61

 Calculation of test statistic: If the
proportion of defective item is same in the
𝟒𝟎×𝟐𝟐
three shipment then there will be 𝟏𝟓𝟎 =
𝟓. 𝟖𝟔 defective items in shipment 𝐴 and the
remaining 40 − 5.86 = 34.14 non defective.
50×22
 Inthe same ways defectives in 𝐵 = =
150
7.33 and remaining 50 − 7.33 = 42.67 non
defective.
60×22
 Defective in 𝐶 = = 8.8 and remaining
150
60 − 8.8 = 51.2 non defective

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 62

 Calculation of 𝜒 2
𝑂 𝐸 𝑂−𝐸 2 𝑂−𝐸 2 𝐸
5 5.86
35 34.14
8 7.33
42 42.67
9 8.8
51 51.2

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 63

 Calculation of 𝜒 2
𝑂 𝐸 𝑂−𝐸 2 𝑂−𝐸 2 𝐸
5 5.86 0.7396 0.1262
35 34.14 0.7396 0.0217
8 7.33 0.4489 0.0612
42 42.67 0.4489 0.0105
9 8.8 0.04 0.0045
51 51.2 0.04 0.0008
0.2249

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 64

 2
𝜒 = 0.2249
 Level of significance: 𝛼 = 0.05
 Degrees of freedom: = 𝑟 − 1 𝑐 − 1 = 2
 Critical value: For 2 degrees of freedom at
5% level of significance, the table value of
2
 𝜒 = 5.991

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 65

 Decision: Since the calculated value of
𝜒 = 0.2249 is less than the table value
2

𝜒 2 = 5.991,
 the hypothesis is accepted
 ∴ Proportion of defective items is same in all
shipments.

Prof. Nandini Rai χ^2- test Wednesday, March 20, 2024 66