I don't think there are any, but you could try doing non-parametric tests on the pairwise permutations, and then do a Bonferroni correction. For example, if the Kruskall Wallis test is comparing A v B v C, you could do separate Mann Whitney U tests on each of A v B, A v C and B v C. You could then allow for the fact that you did 3 tests by making your alpha 0.05/3=0.0167, though this is perhaps overly conservative. You could then find out which of the pairwise comparisons were responsible for the overall difference between the groups that the Kruskall wallis test detected. Of course, you would NOT do the pairwise tests unless the overall Kruskall wallis was significant (the same approach as in the parametric ANOVA).

I don't think there are any, but you could try doing non-parametric tests on the pairwise permutations, and then do a Bonferroni correction. For example, if the Kruskall Wallis test is comparing A v B v C, you could do separate Mann Whitney U tests on each of A v B, A v C and B v C. You could then allow for the fact that you did 3 tests by making your alpha 0.05/3=0.0167, though this is perhaps overly conservative. You could then find out which of the pairwise comparisons were responsible for the overall difference between the groups that the Kruskall wallis test detected. Of course, you would NOT do the pairwise tests unless the overall Kruskall wallis was significant (the same approach as in the parametric ANOVA).

I agree with Mark Perry. The only thing is, you need to be careful about how many comparisons you do because when you have too many Mann Withneys your alpha will eventually become so small that hardly anything will turn up significant.

Mark Perry and Raluca Caplescu,

thank you for your valuable suggestions.

I agree with Mark Perry, I would say a Mann Whitney so that you can compare two means as you go along individually. This is what i am using for my dissertation :)

You could use Multiple comparison test for non-parametric data http://aquaticr.wordpress.com/2012/12/18/multiple-comparison-test-for-non-parametric-data/ and/or read more here http://stats.stackexchange.com/questions/17342/is-there-a-nonparametric-equivalent-of-tukey-hsd

Tom