Date:         Wed, 26 Nov 2003 10:00:37 -0800
Reply-To:     jas <jwortley@CHASS.UTORONTO.CA>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         jas <jwortley@CHASS.UTORONTO.CA>
Organization: http://groups.google.com
Subject:      Skewed residuals in proc mixed
Content-Type: text/plain; charset=ISO-8859-1

Hello,

I'm hoping that someone can help me with the following problem. I'm
conducting analyses for my Ph.D. dissertation and have run into an
issue that my advisors haven't been able to help me with.

I'm using census and population health data (N=38,000) and have a very
skewed continuous dependent variable (i,e., distress). In using proc
mixed in SAS, tests show that the residuals are, unfortunately, also
very skewed.  Several transformations did not significantly improve
the appearance of the residuals (although the log-transformation looks
the most promising). Tests for normality (i.e., the
Kolmogorov-Smirnov, Cramer-von Mises and the Anderson-Darling) show
significant p-values but I'm not sure how applicable they are to such
a large sample.  Can I apply the BIC criteria to these tests, and if
so, how?

Your advice on this matter would be greatly appreciated.

Thanks,

Jas.