In: Statistics and Probability
Statistics exercise
Kruskal-Wallis H/One-way Between Groups (BG) ANOVA
Suppose you would like to know if political party affiliation (democrat, republican, libertarian) has an effect on how people feel about the war in Afghanistan. You ask people to rate on a scale from 1-100 how much they support the war effort. Use the data below to see if the groups differed.
Democrats |
Republicans |
Libertarians |
35 |
65 |
15 |
15 |
60 |
25 |
50 |
75 |
35 |
40 |
60 |
40 |
30 |
90 |
35 |
45 |
75 |
25 |
25 |
65 |
45 |
40 |
70 |
35 |
Let us denote population means of the three groups by
We test the hypothesis that mean scale of all groups is same.
Null hypothesis:
against alternative hypothesis that group means are different from each other.
Let us test this hypothesis at 5% level of significance and the test is one tailed.
Group 1: (Democrat)
Democrat | deviation | squared deviation | |
35 | 0 | 0 | |
15 | -20 | 400 | |
50 | 15 | 225 | |
40 | 5 | 25 | |
30 | -5 | 25 | |
45 | 10 | 100 | |
25 | -10 | 100 | |
40 | 5 | 25 | |
Mean | 35 | ss | 900 |
Variance of group 1:
Group 2: (Repluclicans)
Replublicans | deviation | squared deviation | |
65 | 30 | 900 | |
60 | 25 | 625 | |
75 | 40 | 1600 | |
60 | 25 | 625 | |
90 | 55 | 3025 | |
75 | 40 | 1600 | |
65 | 30 | 900 | |
70 | 35 | 1225 | |
Mean | 70 | ss | 10500 |
Variance of group 2:
Group 3: (Libertarians)
Libertarians | deviation | squared deviation | |
15 | -20 | 400 | |
25 | -10 | 100 | |
35 | 0 | 0 | |
40 | 5 | 25 | |
35 | 0 | 0 | |
25 | -10 | 100 | |
45 | 10 | 100 | |
35 | 0 | 0 | |
Mean | 31.875 | ss | 725 |
Variance of group 3:
Number of observations in each group is 8. i.e. n = 8
Total number of groups is 3. i.e. k = 8
Total number of observations are 24. i.e. N = 24
Therefore,
Find grand mean:
Group mean | Grand mean | deviation | squared deviation |
35 | 44.5833 | -9.5833 | 91.83963889 |
70 | 44.5833 | 25.4167 | 646.0086389 |
31.875 | 44.5833 | -12.7083 | 161.5008889 |
SS | 899.3491667 |
Test statistic:
df of test statistic is (k-1, N-k)
Thus here df is (2, 21)
p-value for is 0.0075.
p-value < 0.05
Hence we reject null hypothesis.
Thus we conclude that mean scale level for all three gropus is not equal. That means political party affiliation (democrat, republican, libertarian) likely to have an effect on how people feel about the war in Afghanistan.
APA writeup:
df = (2, 21), p-value = 0.0075