Once in a statistics class I've seen the suggestion that one non-parametric approach for ANOVA would be to perform the ANOVA analysis on the ranks of the original data (basically as an alternative to Kruskal-Wallis). In an answer to a different question (Rank transformed 2-way ANOVA), @FrankHarrel suggests that "A two-way ANOVA on ranks is not based on strong statistical principles."

So, more generally, is it advisable/sane to do ANOVA on ranks? And why?

  • 2
    $\begingroup$ One way ANOVA on ranks (as long as you correctly compute significance via the permutation distribution of the test statistic under the null) is the Kruskal-Wallis. However, in the case of two- (and higher-) way ANOVA, things become somewhat more complicated. See here. The whole article and its references have both some good news and some bad news. $\endgroup$
    – Glen_b
    Sep 16 '14 at 3:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.