1

I have three groups and aproximately 500 independent variables. I wanted to know if I had to do any multiple comparison correction (e.g. Bonferroni or Benjamini-Hochberg) on the p-values obtained with my KW tests, or directly on the post-hoc results (e.g. Dunn´s post hoc after KW). I have read in this forum that KW controls for an overall false-positive rate but I wanted to be sure. Thank you.

user109001
  • 11
  • 1
  • It's not clear what you are presuming here and there are no sources: "I have read" doesn't identify what you looked at, so please give a specific URL. There is, naturally, no sense in which any Kruskal-Wallis test knows about any other you are conducting. How many tests are you conducting? – Nick Cox Mar 18 '16 at 09:49
  • This would be the specific URL: http://stats.stackexchange.com/questions/133444/bonferroni-correction-on-multiple-kruskal-wallis-tests. I am conducting as many KW tests as independent variables (500). – user109001 Mar 18 '16 at 10:00
  • 1
    That raises standard issues with multiple tests then. – Nick Cox Mar 18 '16 at 10:37

0 Answers0