Scribd
Upload
95 views

Chi Square Notes

This document discusses the chi-square test statistic and its uses for hypothesis testing. The chi-square distribution can be used for (1) tests involving single population variance if the data is normally distributed, (2) goodness of fit tests to determine if data fits a particular distribution, and (3) tests of independence between two factors. For tests of single population variance, the null hypothesis compares the variance to a hypothesized value. Goodness of fit tests check if data fits an expected distribution. Tests of independence evaluate if two factors are independent or dependent. All three types of tests use a chi-square calculated test statistic and compare it to critical values from the chi-square distribution table based on the test type and degrees of freedom.

Uploaded by

hari yahoo
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
95 views

Chi Square Notes

This document discusses the chi-square test statistic and its uses for hypothesis testing. The chi-square distribution can be used for (1) tests involving single population variance if the data is normally distributed, (2) goodness of fit tests to determine if data fits a particular distribution, and (3) tests of independence between two factors. For tests of single population variance, the null hypothesis compares the variance to a hypothesized value. Goodness of fit tests check if data fits an expected distribution. Tests of independence evaluate if two factors are independent or dependent. All three types of tests use a chi-square calculated test statistic and compare it to critical values from the chi-square distribution table based on the test type and degrees of freedom.

Uploaded by

hari yahoo
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Chi Square Test Statistic

Gunnvant Saini
Department of Commerce and Business Management
GNDU
October 4, 2013

What is 2 distribution ?

Just like a normal distribution, 2 is also a continuous probability distribution. For the purposes of this course we
will use chi square distribution to conduct hypothesis tests for the following set of situations:
(i) Hypothesis tests involving single population variance
(ii) Goodness of fit tests
(iii) Hypothesis tests concerning independence of two factors

Tests involving single population variance

In our discussion so far we have discussed hypothesis tests involving mean and proportion as parameters. We have
also discussed the case of comparing two population variances using f test statistic. But the case of hypothesis
tests involving just a single population variance has not been discussed. The chi square test statistic can be used to
conduct hypothesis tests concerning variances under a strict assumption of Normality. The data whose variance is
being hypothesized about should be normally distributed.

2.1

Form of Hypothesis

If the data is normally distributed then one can move ahead and use chi square distribution for hypothesis testing.
The null and alternate hypothesis can take either of the three forms mentioned below:
Form 1
Ho : = o ; Ha : 6= o
Form 2
Ho : = o ; Ha : > o
Form 3
Ho : = o ; Ha : < o
The test statistic is calculated using the following relation:
2
2cal = (n1)s
2
0

2.2

Process of testing

The procedure for testing of hypothesis involves the usual mechanism of computing a test statistic and then comparing
with a standard value of test statistic at a given level of significance and corresponding to a particular value for
degrees of freedom.But the overall process depends upon the form taken by null and alternate hypothesis.
2.2.1

Form 1

The form 1 is essentially a scenario involving a two tail test. Let the significance level be and degrees of freedom
be n-1. Now in order to perform a two tail test two standard values are read off from a 2 table. One value is called
2U , the other 2L . To look for the value of 2U , we halve the significance level i.e calculate /2. Then the table
value corresponding to /2 and df is looked up from the table. In order to determine the value of 2L , we find out
a = 1 /2. Then the value corresponding to a and df is read off from the table.
The rule for rejection is, Reject Ho if:
2cal > 2U
Or
2cal < 2L
1

2.2.2

Form 2

When ever the Ho and Ha are in form 2 only the value of 2U is read off from the table. The value is determined by
reading off the value corresponding to and df. Note unlike the case of From 1 we dont halve the significance level.
The rule for rejecting Ho is:
Reject Ho if 2cal > 2U
2.2.3

Form 3

In form three, only the 2L is read off from the table corresponding to 1 and df.
The rule for rejecting Ho is :
Reject Ho if 2cal < 2L

Test involving goodness of fit

The goodness of fit tests are used to test if the data is distributed in a certain way or not. The general form of Ho
and Ha is as follows:
Ho:Data is distributed in a certain way
Ha:Data is not distributed in a certain way
The test statistic is calculated as follows:
2

e)
2cal = (fo f
fe

Commonly used form of this test is used to test the uniform distribution of data. For example an HR manager
might want to know if the number of employees absent on any day of the week are same as those absent on any
other day, in other words we wish to test if the number of absent employees is uniformly distributed over the week.
In such a situation 2 test can be used.
The degrees of freedom, df are calculated as follows
df = m 1
Here m is the number of categories.

Tests of Factor independence

Another situation in which the independence of two factors is evaluated. The null and alternate hypothesis take this
form:
Ho:Factor 1 is independent of Factor 2
Ha:Factor 1 is dependent on Factor 2
2
e)
2cal = (fo f
fe
The degrees of freedom, df are calculated as
df = (r 1)(c 1)
Here r is the number of rows and c is the number of columns.

You might also like