11

I have read the Wikipedia pages for Friedman's and Kruskal-Wallis' test, but I am not sure which one to use. Are there differences in the assumptions?

beza1e1
  • 229
  • 1
  • 2
  • 6
  • 3
    It would be great if you could say a little more about your data/design so we can really help you to decide which test is the best for your specific problem. – Henrik Jun 17 '11 at 13:30

1 Answers1

15

Kruskal-Wallis' test is a non parametric one way anova. While Friedman's test can be thought of as a (non parametric) repeated measure one way anova.

If you don't understand the difference, I compiled a list of tutorials I found about doing repeated measure anova with R, you can find them here...

Tal Galili
  • 19,935
  • 32
  • 133
  • 195
  • 2
    So the difference is "repeated measure", which means what assumption? – beza1e1 Jun 17 '11 at 17:01
  • 3
    Yep. "repeated measure" means that you have (well) repeated measure on the same "subject". For example - comparing the grades of (3 different groups) students on 3 different tests will be one way anova. But if we had for each student his score for each of the same test, then we could compare between the test using the repeated measures taken for each student. – Tal Galili Jun 17 '11 at 17:12
  • 2
    In other words: "repeated measures" = within-subject experiment design, whereas the basic one-way ANOVA is "between subjects". – antoine Sep 16 '15 at 19:09