22
Oct
Mann Whitney U Test Sample Size. We would expect this and consequently suggest a suitable adjustment to the sample size calculation if a Mann-Whitney U test or other non-parametric hypothesis test were used 9 10. It is important to remember the null hypothesis for this test and to differentiate it from the nulls for the t-test and the median test. First only n2 vs. 08 Results alpha 2-sided 00500000 Power 08000000 Estimated relative effect p 03491124 N total sample size needed 1113719511 t n1 N 05000000 n1 in Group 1 556859755 n2 in Group 2 556859755 N rounded 1120000000 n1 rounded 560000000 n2 rounded 560000000.
The sample sizes are like n113k n23k n3200 n430 But first I ran with the test with sample sizes above I had different patterns than when I increased the sample size. Note however that your statistical power ie the ability to detect a difference that really is there will diminish as the group sizes become more unequal. C 05 Target Unlike t-test that compares the means the Mann-Whitney U test compares a randomly selected value from group1 to a randomly selected value from group2. Kendalls tau -b is sometimes used and varies from approximately 1 to 1. The minimum acceptable probability of preventing type I error. Here the sample size is large so the Z approximation p-value of 0017 should be used.
When normality holds the MannWhitney U test has an asymptotic efficiency of 3 π or about 095 when compared to the t-test.
It is important to remember the null hypothesis for this test and to differentiate it from the nulls for the t-test and the median test. Sample size is 40 or less an Exact test is performed automatically and another row appears in the output entitled Exact Sig. C 05 Target Unlike t-test that compares the means the Mann-Whitney U test compares a randomly selected value from group1 to a randomly selected value from group2. C -05 when U μ. If you expect a small difference between groups then you likely need a larger sample. Here the sample size is large so the Z approximation p-value of 0017 should be used.