topic Re: Kruskal Wallis in SAS Procedures

Kruskal Wallis

deleted_user — Fri, 27 Jun 2008 17:58:01 GMT

Does anyone know how to do a Kruskal Wallis test in SAS with TWO independent variables instead of just one?

Re: Kruskal Wallis

Olivier — Fri, 27 Jun 2008 20:52:37 GMT

I haven't checked what test KW is exactly, but from what I remember, it is like an analysis of variance on ranks, isn't it ?
If so, just compute ranks with PROC RANK and then proceed with PROC GLM.
Am I correct ?

Re: Kruskal Wallis

deleted_user — Fri, 27 Jun 2008 21:04:11 GMT

Thanks. Yes, it is based on rank as far as I can tell. It seems to be a nonparametric equivalent of ANOVA and you can use proc npar1way, but now I want to add a second independent variable. It doesn't seem to allow this and the only thing I could find when I googled it was some talk of macros that created a multivariate KW, but couldn't actually get to the macro itself.

Re: Kruskal Wallis

LAP — Mon, 30 Jun 2008 21:46:41 GMT

You can use proc npar1way to do a Kruskal-Wallis (Wilcoxen)

From Documentation....
The Nonparametric One-Way ANOVA task enables you to perform nonparametric tests for location and scale when you have a continuous dependent variable and a single independent classification variable. You can perform a nonparametric one-way ANOVA using Wilcoxon (Kruskal-Wallis), median, Van der Waerden,.....

Re: Kruskal Wallis

Doc_Duke — Tue, 01 Jul 2008 01:51:25 GMT

Just rank the dependent variable and then run the two way ANOVA via GLM.

Generally, you can do any rank-based analysis using the normal theory equivalent procedure in SAS. However, you are relying on the Central Limit Theorem to get the p-values, so you may need substantial sample sizes for this to work. [That's what Wilcoxon was doing when he developed the Rank-Sum test; he was looking for an easier way to do a t-test (think hand calculation!).]