Inference for Two-way Table
Chi-square test can be used to test
the null hypothesis that there are no differences among the proportion
of successes in different population ( treatment ).
In the case of Exercise
9.2 ( BPS, Chapter 9, page 475 ), a survey of students' smoking
habit was taken in eight Arizona high schools. Data are arranged
in a 3 x 2 table. The 2 columns are "Smoker" and "Nonsmoker"
student groups respectively. The 3 rows defines the smoking habit
of their parents, which are "Both parents smoke", "One
parent smokes" and "Neither parent smokes" groups.
The column only can contain integer values. We want to address
the question that the proportion of smoking students in these
three kinds of families are equal or not. The alternative hypothesis
for this case is that not all three proportions are equal.
Select
Stat->Tables->Chi-Square Test from the menu, then
just enter the Columns containing the table in the dialog
box before clicking OK.
The outcome
is shown on the right,which providing the Chi-square statistic,
the number of degrees of freedom and the P-value. Df is equal
to (row-1) x (column-1). In this example, the P-value is approximately
zero, which gives us strong confidence to reject null hypothesis
that there is no differences between the proportion of smoking
students with different parents smoking habits.
Another use of Chi-square
test is to test whether there is a statistically significant
relationship between two categorical variable. Exercise 9.13 (BPS
Chapter 9, page 491) describes a study of the relationship between
genders and the interests in different business programs. A survey
was given among 722 members of the senior class in the College
of Business Administration at the University of Illinois. One
question asked which major within the business program the student
had chosen. The data from students who responded is shown. Row
1, 2, 3 and 4 show the number of students interested in "Accounting",
"Administration", "Economics", and "Finance"
respectively.
Select Stat->Tables->Chi-Square
Test from the menu to test the null hypothesis : there is
no relationship between gender and program interests.
The P-value
is 0.013, which is quite small. So we have strong evidence that
interests in different programs is related to gender.
go back to index