I want to compare the means between groups 1
vs. 2
and also 2
vs. 3
. All three groups are nonnormal, as the Shapiro-Wilk test resulted in p values of 0.03, 0.04 and <0.0001 respectively. n is 10, 13 and 38 respectively, and the Levene test for equal variances resulted in p values of 0.22 for 1
vs. 2
and 0.44 for 2
vs. 3
(so they seem to have equal variances). Both, the parametric t-test and the nonparametric Wilcoxon test resulted in insignificant differences with p values larger than 0.5. Which test is more appropriate to this situation? The groups do not follow a normal distribution, have different sample size, but seem to be homoscedastic.
Your help will be very much appreciated.
1 Answer
The Mann-Whitney U (also called Wilcoxon rank sum) test is the nonparametric "equivalent" of the two sample t-test. Sounds like you did the right test(s).