I got significant result after Kruskal-wallis, but when I applied two post hoc tests (Dunn test and Krystalmc), I obtained non significant results! How to explain that?
Asked
Active
Viewed 16 times
0

houneida
- 161
- 1
- 2
- 8
-
What did the many answers to more or less the same question on this site tell you? – Alexis Dec 24 '21 at 21:35
-
See [this Q&A](https://stats.stackexchange.com/questions/515524/kruskal-wallis-post-hoc-analysis/515658#515658) and perhaps others linked in the margin as "Related." Bottom line is that _post hoc_ tests use different criteria than the K-W test, so K-W may 'say' there are differences that the _ad hoc_ tests do not find. This happens most often when K-W is only barely significant and/or there are many levels of the factor to compare _ad hoc._ – BruceET Dec 25 '21 at 01:00
-
See https://stats.stackexchange.com/questions/83030/can-anova-be-significant-when-none-of-the-pairwise-t-tests-is for answers. The question is not specific to Kruskal-Wallis. – BigBendRegion Dec 25 '21 at 19:55
1 Answers
0
There is no relationship between kruskal wallis and any post hoc test it's just a good practice to apply post hoc test if kruskal wallis is significant then there is not anything behind the math which a significant kruskal wallis guarantees a significant post hoc.

Davi Américo
- 737
- 1
- 11