🚀 go-pugleaf

     I suspect that someone has developed some obscure technique to do
multiple comparisons after a significant Kruskal-Wallis -- new multiple
comparison techniques seem to pop up with such great frequency that I
generally tend to ignore them.  So what help can I be -- well, if you find
nothing else appealing, try this (intuitively appealing to me) approach --
count the number of comparisons you wish to make and adjust the per comparison
alpha by dividing the maximum acceptable familywise alpha by that number --
that is, do a Bonferroni test.  For a test statistic, use either the
Wilcoxon rank sum test (Mann-Whithey U) or the Kruskal-Wallis again.  Of
course, you will have to obtain exact p values for your comparisons to
apply the Bonferroni test.