Date:         Thu, 8 Jun 2006 22:31:50 -0700
Reply-To:     David L Cassell <davidlcassell@MSN.COM>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         David L Cassell <davidlcassell@MSN.COM>
Subject:      Re: ANOVA with very different cell variances and non-normal data
In-Reply-To:  <200606081335.k58DFJ3d012729@mailgw.cc.uga.edu>
Content-Type: text/plain; format=flowed

smarkandsusan@AOL.COM wrote:
>Hi:
>    I'm told that you really shouldn't use ANOVA under these conditions.
>Somebody suggested a Scheirer-RAy-Hare test might be appropriate but
>I've never heard of it, does
>anybody know where I might find additional info or have other
>suggestions?
>
>Thanks,
>Mark

You're right.  You should *not* go with ordinary ANOVA in your case.
Good work on spotting that, by the way.  Lots of people don't.

You have several options, depending on your data, and your data sources,
and your meta-data.

If your data come from a survey sample, then you ought to be using
PROC SURVEYREG to do the anova.

If your data are a reasonable ANOVA model except for some outliers and/or
leverage points, then I would suggest that you look into PROC ROBUSTREG.

If there is a logical, scientifically meaningful transformation of the data
that would stabilize the variances, then PROC TRANSREG would be
a reasonable choice.

There is Analysis of Means in the SAS/QC module.

And there is Kruskal-Wallis in PROC NPAR1WAY, if you are willing to
test that the medians are equal, instead of the means.

HTH,
David
--
David L. Cassell
mathematical statistician
Design Pathways
3115 NW Norwood Pl.
Corvallis OR 97330

_________________________________________________________________
Don’t just search. Find. Check out the new MSN Search!
http://search.msn.click-url.com/go/onm00200636ave/direct/01/