I have 4 groups that I compare on an interval dependent variable.
I coded these groups 1, 2, 3, 5(=control). There normally is a group 4 but none of the values are included in the analysis as I am using a filter.
Kruskal Wallis indicates there is a significant difference between groups.
However, when I run separate Mann Whitney U tests in order to compare which group differs with respect to the control group(5), then none of the differences are significant. Anyone has an explanation for this?