SAS code for multiple comparison for KRUSKAL-WALLIS

SAS code for multiple comparison for KRUSKAL-WALLIS

Post by Mr J Simaka » Wed, 02 Aug 2000 04:00:00



Dear Colleagues,

Could anyone please supply some code for multiple comparison for
KRUSKAL-WALLIS TEST( one -way nonparametric equivalent for ANOVA).
Always say thank you.

June Simakani
Department of Statistics
University of Transkei
South Africa

 
 
 

SAS code for multiple comparison for KRUSKAL-WALLIS

Post by David L. Casse » Wed, 02 Aug 2000 04:00:00



> Could anyone please supply some code for multiple comparison for
> KRUSKAL-WALLIS TEST( one -way nonparametric equivalent for ANOVA).

You don't need to.  SAS can do this for you in PROC NPAR1WAY.

proc npar1way wilcoxon data=whatever;
  class classvar;
  var testvar;
  run;

When you use the WILCOXON option on more than two classes, you end
up with Kruskal-Wallis.  Convenient, isnt it?

Depending on your underlying distribution(s) , Kruskal-Wallis may not
be the optimal test.  But I can't evaluate your data from way over here.  :-)
Kruskal-Wallis is locally most powerful if you're testing for location shifts in
a series of logistic distributions [if that's any help].

Quote:> Always say thank you.

You're welcome.

David
--

Senior computing specialist
mathematical statistician

 
 
 

1. Macro Required > Nonparametric Tests > Post hoc comparisons for Kruskal Wallis test

Hi

I was wondering if anyone has a macro to perform a post hoc analysis (such
as Dunn's Test) for a Kruskal-Wallis Test. Surprisingly SPSS will perform
the KW test, but even in 11.5 does not provide a means of analysing which
pairs of groups differ significantly. I note the SPSS web site suggests
performing a series of Mann Whitney tests, but Dunn's Test (Neave and
Worthington, 1988, Distribution Free Statistics, Unwin Hyman: London)
provides a more acceptable resolution. I note that Minitab provides this
functionality since version 9 or 10.

If anyone has a macro/script to perform these multiple comparisons then it
would be greatly appreciated.

Yours faithfully
Peter Edwards
University of Sydney, Australia

2. newbie: BIND not able to use named.root

3. Can the Mann-Whitney U be used following a Kruskal-Wallis test?

4. Tell Command a server startup

5. Kruskal-Wallis H adjusted for ties in SPSS?

6. Switching character disks

7. ANOVA and Kruskal-Wallis

8. National Embedded NS486

9. Kruskal-Wallis Test

10. post hoc analysis following Kruskal-Wallis Test

11. How to preform Cuzick's test after a Kruskal-Wallis test?

12. Kruskal-Wallis One-Way ANOVA

13. post hoc comparisons for Kruskall-Wallis nonparametric Tests