I don't know if the problem has been fixed since Michael Friendly posted the following on York University's site but, apparently, there was a problem/bug with regard to orthoginal polinomial contrasts with unequal Ns.

His macro, http://euclid.psych.yorku.ca/ftp/sas/macros/orpoly.sas , includes the comment:

" =Bugs:

For some strange reason, PROC GLM considers the contrasts non-estimable. It does the proper tests when the sample sizes are equal, but not when they differ."

Anyone know anything about this?

Art -------- On Thu, 12 Apr 2007 18:49:53 -0400, Jeffrey Bewley <jeffreybewley@GMAIL.COM> wrote:

>I was in a meeting today where a colleague, who is strongly anti-SAS, >suggested that one of the reasons that we should not use SAS is because >there is a bug in the way that SAS handles contrasts. Could someone address >what bug he may be referring to, if it still exists, and why it is or isn't >a concern? > >Thank you, >Jeffrey Bewley > >------------------------------------ >Department of Animal Sciences >Jeffrey Bewley >Graduate Research Assistant >jeffreybewley@gmail.com