RE: st: RE: How to perform a non parametric manova

["Lachenbruch, Peter" <Peter.Lachenbruch@oregonstate.edu>]

There is an old book by Puri and Sen (1971) "Nonparametric Multivariate Analysis" published by Wiley (not exactly sure of the title).  
Two other approaches:
1.	use rank transforms on the data (replace the observations by their ranks) and then do a MANOVA.
2.	Adopt a permutation approach:  find the statistic of interest from MANOVA and do a permutation test on it.  1000 permutations is a good start, for the final paper, I'd go up a bit to maybe 10000 permutations.

Tony

Peter A. Lachenbruch
Department of Public Health
Oregon State University
Corvallis, OR 97330
Phone: 541-737-3832
FAX: 541-737-4001


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Nick Cox
Sent: Wednesday, May 26, 2010 5:31 AM
To: statalist@hsphsun2.harvard.edu
Subject: RE: st: RE: How to perform a non parametric manova

Thanks for that reference. This kind of question presumably arises out of wanting to have it both ways, to do a MANOVA while worrying about whether assumptions are satisfied; no criticism there, as to some degree most statistical science is under the same tension. 

I suspect you would have to go outside Stata to do this. Many ecologists these days have moved towards R. Alternatively, transformation of the data before MANOVA may give some guidance what is fragile and what is robust, in one sense of that over-used word. 

Nick 
n.j.cox@durham.ac.uk 

Steve Samuels

"Does such a thing even exist?"

Apparently, yes. A google search of "nonparametric manova" turns up a
permutation test: Austral Ecology (2001) 26, 32-46.  A new method for
non-parametric multivariate analysis of variance, by  Marti J.
Anderson

The test isn't implemented in Stata. And, "nonparametric" doesn't mean
"robust". To quote the paper (p. 37): "Like its univariate
counterpart, which is sensitive to heterogeneity of variances, this
test and its predecessors that use permutations.... will also be
sensitive to differences in the dispersions of points, even if the
locations do not differ."

On Wed, May 26, 2010 at 7:42 AM, Nick Cox <n.j.cox@durham.ac.uk>
wrote: > Does such a thing even exist? For example, even
Kruskal-Wallis is a very > limited parody of -anova-. (No scope for
handling interactions so far as > I know.) >
>
> amatoallah ouchen >
> Does anyone have an idea about how to perform a non parametric manova? > an equivalent of the kruskal wallis test for anova? >
> > * > *   For searches and help try: > *   http://www.stata.com/help.cgi?search > *   http://www.stata.com/support/statalist/faq > *   http://www.ats.ucla.edu/stat/stata/ >

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/