What (if any) is the appropriate post-hoc test for a Kruskal-Wallis
test in SPSS?  Given an ordinal variable, and a categorical variable,
I want to see if there is a significant difference between the groups
determined by the categorical variable.  However, if there is a
difference, I would like specific information regarding the
between-group differences.  Is there a way to do this without
performing multiple Kruskal-Wallis tests?  If not, what is the
appropriate correction for performing multiple tests.  Thank you for
you help in solving this problem.

---

Clarification of Question by mgdwyer-ga on 27 Sep 2004 20:12 PDT

Sorry, I mistakenly typed Kruskal-Wallis a second time there.  What I
meant to say was: is there a way to do this without performing
multiple MANN-WHITNEY tests.

Hi mgdwyer,

Here's some initial info for you; a researcher may form this into a
formal answer for you.

Usual practice would be to followup your significant K-W test with a
series of Mann-Whitney tests. This feature isn't built into SPSS but a
script has been written to run all pairwise Mann-Whitney tests (see
'Multiple Mann-Whitney tests.SPS' at the great SPSS resource site
http://pages.infinit.net/rlevesqu/SampleSyntax.htm#T-Test).

You will have to apply your own multiple testing correction manually.
An easy method to implement is the Holm method - based on the simple
Bonferroni method but more powerful. You can see details of the Holm
method in this article, available through Ovid: Aickin, M & Gensler,
H. Adjusting for multiple testing when reporting research results: the
Bonferroni vs Holm methods. American Journal of Public Health. 1996;
86(5):726-728.

Hope this helps!
S.