In: Statistics and Probability
Test the null hypothesis of independence of the two classifications A and B of the 3x2 contingency table shown below. Use alpha = 0.05.
Males | Females | |
In Favor of X | 40 | 72 |
Against X | 63 | 53 |
where X can be any policy issue. The null hypothesis is that gender (Males/Females) does not affect the views on X (in favor/against). The alternative is that there is a gender bias ( the view is not independent of gender) . Remember that the problem has a test statistic that follows the chi-square distribution.
null hypothesis : gender and view on X are independent
Alternate hypothesis : gender and view on X are not independent
degree of freedom(df) =(rows-1)*(columns-1)= | 1 | |
for 1 df and 0.05 level , critical value χ2= | 3.841 | |
Decision rule : reject Ho if value of test statistic X2>3.841 |
Applying chi square test of independence: |
Observed | males | females | Total | |
in favor | 40 | 72 | 112 | |
against | 63 | 53 | 116 | |
total | 103 | 125 | 228 | |
Expected | Ei=row total*column total/grand total | males | females | Total |
in favor | 50.596 | 61.404 | 112.00 | |
against | 52.404 | 63.596 | 116.00 | |
total | 103.00 | 125.00 | 228.00 | |
chi square χ2 | =(Oi-Ei)2/Ei | males | females | Total |
in favor | 2.219 | 1.829 | 4.0479 | |
against | 2.143 | 1.7656 | 3.9083 | |
total | 4.3619 | 3.5942 | 7.956 | |
test statistic X2 = | 7.956 |
since test statistic falls in rejection region we reject null hypothesis | |||
we have sufficient evidence to conclude that gender and view on X are not independent |