B0002

Chi-Square Test

Data that are categorical,

classificatory, count, frequency,
enumerative or nominal in nature like
gender, marital status, educational
attainment, course or academic
program, etc. can be treated using
Chi-Square test.

Chi-square test can either be used in

testing the following:

a. Goodness-of-Fit
b. Independence

Chi-Square Test * Property of STI

Page 1 of 17
 B0002

Chi-Square Goodness-of-Fit

 used to determine whether the

observed frequencies in a one-way
classification table are in complete
agreement with the expected or
hypothesized frequencies

Ho: The data follow a specified

distribution.
(The observed frequencies are in
agreement with the expected
frequencies).

Ha: The data do not follow a

specified distribution.
(The observed frequencies are not
in agreement with the expected
frequencies).

Chi-Square Test * Property of STI

Page 2 of 17
 B0002

Chi-Square Goodness-of-Fit
k k

O   E
i i n
Notations: i 1 i 1

Let Oi = observed frequency in the ith

cell or class
Ei = expected frequency under Ho
in the ith cell or class
k = number of cells or classes,
and
= total sample size.

Assumptions:
1. There are k mutually exclusive classes;
2. Each observation falls in one and only one
class, and
3. No more than 20% of the classes have
expected frequency is less than 1.

*One may have to collapse classes in some

meaningful manner to attain (3).

Chi-Square Test * Property of STI

Page 3 of 17
 B0002

Chi-Square Goodness-of-Fit

Test statistic:

 2c  
k
Oi  Ei 2
i 1 Ei

which follows a Chi-square distribution

with (k-1) degrees of freedom under Ho.
Under the null hypothesis, Ei = nPi where Pi
is the hypothesized probability (under Ho)
that an observation will fall in the ith
class.

Decision Rule:

Reject Ho if  c 
2 2
 , k 1, otherwise,
fail to reject Ho.

Chi-Square Test * Property of STI

Page 4 of 17
 B0002

Chi-Square Goodness-of-Fit

Example:

The latest TV Ad of Coca Cola Softdrinks

was rated by 200 viewers (with 5 –
strongly disliked to 1 – strongly liked). The
data are as follows:

Rating 1 2 3 4 5
Frequency 30 75 20 45 30

At α = 0.01, test the claim that the

proportions of viewers are equal in each
rating scale

Chi-Square Test * Property of STI

Page 5 of 17
 B0002

Chi-Square Goodness-of-Fit

Solution:

Hypotheses
Ho: The proportions of viewers are equal
in each rating scale.
Ha: The proportions of viewers are not
equal in each rating scale.
 2 test
Test Statistic

k
 2c  
Oi  Ei 2
i 1 Ei

Level of Significance
1%

Chi-Square Test * Property of STI

Page 6 of 17
 B0002

Chi-Square Goodness-of-Fit

Decision Rule

Reject Ho if  c    ,k 1   0.01, 4  13.277 ,

2 2 2

otherwise, fail to reject Ho.

Computation
n = 200
1
P1  P2    P5 
5
1
Ei  200 Pi  200   40
5

Oi 30 75 20 45 30

Ei 40 40 40 40 40

Chi-Square Test * Property of STI

Page 7 of 17
 B0002

Chi-Square Goodness-of-Fit

Computation:

 2

30  40 75  40 20  40 45  40 30  40
2

2

2

2

2

c
40 40 40 40 40
c2  46.25

Decision
Since 46.25 is greater than the critical
value (  02.01, 4  13.277 ), reject Ho.

Conclusion
The proportions of viewers are not equal
in each rating scale at 1% level of
significance.

Chi-Square Test * Property of STI

Page 8 of 17
 B0002

Test of Independence

 tests whether two random

variables are independent from
each other

The data from this study are arranged in

a two-way table generally known as an r
x c contingency table. The data are
classified into r classes of the row factor
A into c classes of the column factor B
represented as follows:

Chi-Square Test * Property of STI

Page 9 of 17
 B0002

Test of Independence

where
Oij = observed frequency in the ith row
and the jth column
Ri = total observed frequency for the ith
row;
Cj = total observed frequency fot the jth
column; and
n = total sample size

Chi-Square Test * Property of STI

Page 10 of 17
 B0002

Test of Independence

Hypotheses

Ho: The row factor is independent of the

column factor

Ha: There is an association between the

row and column factors.

If the null hypothesis is true, i.e., the row

factor is independent of the column
factor, then

Ri * C j i = 1, 2, 3,…,r and
Eij 
n j = 1, 2, …,c.

where

Eij = expected frequency in the the ith

row and the jth column

Chi-Square Test * Property of STI

Page 11 of 17
 B0002

Test of Independence

Test statistic

r
 c2  
c O
ij  Eij 
2

i 1 j 1 Eij

which follows a chi-square distribution

with (r-1)(c-1) degrees of freedom under
Ho.

Decision Rule
Reject Ho at the α level of significance if
 2c   2,r 1c 1 , otherwise, fail to
reject Ho.

Remark: The necessary assumptions for

the test of independence are similar to
the assumptions on the goodness of fit
test in the previous section.

Chi-Square Test * Property of STI

Page 12 of 17
 B0002

Test of Independence

Example:

Rural areas are known to be low-noise

environments in contrast with urban
areas, where people are constantly
exposed to noise such as traffic noise,
industrial noise and loud music. In relation
to this, a study was conducted on the
relation of hearing loss on the type of
environment. A sample of 100 middle-
aged residents were taken from each of
the two areas and tested. The following
results were obtained:

Test at α = 0.01 whether the type of

environment affects the degree of hearing
loss.

Chi-Square Test * Property of STI

Page 13 of 17
 B0002

Test of Independence

Solution:

Hypotheses

Ho: The type of environment does not

affect the degree of hearing loss.
Ha: The type of environment affects the
degree of hearing loss.

Test Statistic

 2 test
r
 c2  
c O
ij  Eij 
2

i 1 j 1 Eij

Level of Significance
1%

Chi-Square Test * Property of STI

Page 14 of 17
 B0002

Test of Independence

Decision Rule

Reject Ho if  c    ,r 1c 1  02.01, 2  9.21 ,

2 2

otherwise, fail to reject Ho.

Computation
The observed and expected frequencies
are presented as follows:

Chi-Square Test * Property of STI

Page 15 of 17
 B0002

Test of Independence

Computations:

100 * 65 100 * 65
E11   32.5 E21   32.5
200 200

100 * 75 100 * 75
E12   37.5 E22   37.5
200 200

100 * 60 100 * 60
E13   30.0 E23   30.0
200 200

2 
40  32.5 2

45  37.5 2

45  30
2

c
32.5 37.5 30
 c2  21.46

Chi-Square Test * Property of STI

Page 16 of 17
 B0002

Test of Independence

Decision

Since 21.46 is greater than the

critical value ( 0.01, 2  9.21 ), reject
 2

Ho.

Conclusion

The type of environment affects

the degree of hearing loss at 1%
level of significance.

Chi-Square Test * Property of STI

Page 17 of 17