```Date: Fri, 7 Apr 2000 23:35:26 GMT Reply-To: Dale McLerran Sender: "SAS(r) Discussion" From: Dale McLerran Organization: Fred Hutchinson Cancer Research Center Subject: Re: GLM: Simple effects for within subjs factors [posted and mailed] jweedon@earthlink.net (Jay Weedon) wrote in : >Hi folks, > >Is there any easy way to get SAS to do post-hoc tests on >within-subjects effects, using PROC GLM or otherwise? All the post-hoc >options in GLM seem to focus only on between-subjects effects. > >Example: A mixed model ANOVA has 2 groups (A1, A2) and 3 repeats >(B1,B2,B3) per subject. I want to do LSD pairwise comparisons among >B1,B2,B3 pooled across groups, and simple effects of B at A1, B at A2. > >TIA, >Jay Weedon. > Jay, How about using PROC MIXED as follows: proc mixed; class A B SUBJECT; model resp = A|B / s; repeated B / subject=SUBJECT type=UN; lsmeans B / diff; lsmeans A*B / slice=A diff; I specified an unstructured covariance matrix for the repeated measures. You may choose a different covariance structure, but unstructured should be the same as doing a repeated measures analysis using GLM. The structure of the data must be different for a repeated measures analysis using PROC MIXED than it is for PROC GLM. You need one record per time period by subject. That is, you would have 3*Ns observations, where Ns is the number of subjects in your dataset. The first record for subject 1 has response at period B1 and variable B=1, the second record for subject 1 has response at period B2 and variable B=2, etc. In this framework, there is no difference in how we analyze A and B, except that we have specified that B is the variable which identifies the order of repetitions. I believe that this is easier to work with than in the GLM framework. Dale -------------------------------------- Dale McLerran Fred Hutchinson Cancer Research Center Seattle, WA 98109 mailto:dmclerra@fhcrc.org ph: (206) 667-2926 fax: (206) 667-5977 -------------------------------------- ```

Back to: Top of message | Previous page | Main SAS-L page