In: Statistics and Probability
In a study of binge drinking among undergraduates at Ohio
University, a researcher was interested in gender differences as
related to binge drinking and to drinking-related arrests. She
wanted to know two things: (a) Is there a significant relationship
between gender and binge drinking (as defined by 5 or more drinks
at one sitting), and (b) Is there a significant relationship
between gender and drinking-related arrests? A random sample of
males and females were asked about their experiences with binge
drinking and with drinking-related arrests. Test for a relationship
in the following data:
Binge
Drinking?
YES NO
Male 36 21
Female 26 45
What is the calculated chi-squared value
Solution:
Here, we have to use chi square test for independence of two categorical variables.
Null hypothesis: H0: There is no relationship in two variables.
Alternative hypothesis: Ha: There is a relationship in two variables.
We assume level of significance = α = 0.05
Test statistic formula is given as below:
Chi square = ∑[(O – E)^2/E]
Where, O is observed frequencies and E is expected frequencies.
E = row total * column total / Grand total
We are given
Number of rows = r = 2
Number of columns = c = 2
Degrees of freedom = df = (r – 1)*(c – 1) = 1*1 = 2
α = 0.05
Critical value = 3.841459
(by using Chi square table or excel)
Calculation tables for test statistic are given as below:
Observed Frequencies |
|||
Binge Drinking? |
|||
Gender |
Yes |
No |
Total |
Male |
36 |
21 |
57 |
Female |
26 |
45 |
71 |
Total |
62 |
66 |
128 |
Expected Frequencies |
|||
Binge Drinking? |
|||
Gender |
Yes |
No |
Total |
Male |
27.60938 |
29.39063 |
57 |
Female |
34.39063 |
36.60938 |
71 |
Total |
62 |
66 |
128 |
Calculations |
|
(O - E) |
|
8.390625 |
-8.39063 |
-8.39063 |
8.390625 |
(O - E)^2/E |
|
2.549952 |
2.39541 |
2.047145 |
1.923075 |
Test Statistic = Chi square = ∑[(O – E)^2/E] = 8.915582
χ2 statistic = 8.915582
P-value = 0.002827
(By using Chi square table or excel)
P-value < α = 0.05
So, we reject the null hypothesis
There is sufficient evidence to conclude that there is a relationship in two variables.