Kruskall-Wallis: what can we do if not significant?

#1
Hi all,

I have a general question regarding Kruskall-Wallis test:

When Kruskall-Wallis test DOES NOT yield significant results, can we still run Mann-Whitney tests with specific pairs of the conditions? If those pairs are significant, can we consider those results?

Thanks a lot in advance,
Iolanda
 

Karabiner

TS Contributor
#2
Why should one perform a K-W test if they ignore the results anyway. Moreover, it is not
clear whether all possible pairs are tested or whether pre-specified specific hypotheses
are tested (i.e. a subset of pairs), and whether there is a correction for multiple testing.

As is often the case, a "general" question does not make too much sense. Maybe you
describe your actual study instead (objective, measures taken, sample size).

Regards

K.