Post hoc comparisons for non-parametric stats


I have what I'm sure is a fairly ignorant question for those who know about stat but I would appreciate any help.

I am trying to differentiate 3 groups by means of olfactory function amongst other cognitive measures to distinguish Alzheimer's disease.

I have conducted a series of Kruskal-Wallis anovas and am wondering what is the optimal method for assessing differences in the means? Currently I have conducted Mann-Whitney comparisons and intend to correct with bonferroni. Is this ok? I keep seeing reference to a 'Dunn' test for non-parametric tests and am wondering if this is the same thing as a bonferroni/dunn or different. I have an example of a test (LSD) which looks at the least significant difference in the mean ranks between groups to assess significance-is this the same thing?

My knowledge of SPSS (version 11) is limited to say the least and I rely on a good interface as am unfamiliar with syntax etc..

Many thanks,



TS Contributor
Since you have only 3 groups, doing post-hoc Mann-Whitney Tests are fine.

If your original hypotheses (prior to data collection) included simple comparisons, then there's no reason to do a Bonferroni correction (even if you didn't plan for these, I wouldn't worry too much about an escalating Type I error rate for only 3 comparisons).

Nonparametric post-hoc tests have been discussed on this site before, and it was actually a central part of my masters thesis.
Thanks for the help. I saw your previous link just now and those other tests such as Dunn etc were what I was wondering about. My n's are small so if you say Man-Whitney is ok, I'll stick with that. Thanks again.