Paul.R.Swank@UTH.TMC.EDU replied: > >If the covariate is time invariant then simply adsd it to the right side of >the model statement. > >Proc glm; > class cat; > model a1-a5 = cat cov / nouni solution; > repeated y 5 / nom summary; >run; > >If is is a time varying covariate thenyou should move to proc mixed. Of >course you could do the one above in proc mixed as well. > >Also, if you wish to check for homogeneiry of regression slopes, you need >to include the interaction of the classification variable and the covariate >(cat*cov). > >Paul R. Swank, Ph.D. >Professor, Developmental Pediatrics >Director of Research, > > >University of Texas Health Science Center at Houston

I would add to Paul's sage advice by urging the original poster to move to PROC MIXED no matter what.

The sphericity assumption needed by PROC GLM to do repeated measures is usually not supported by the data.

The format required by PROC GLM for repeated measures is not natural for a SAS data set.

PROC MIXED gives far greater flexibility in the possible covariance structures and the ways of comparing covariance structures.

PROC MIXED handles missing values in a repeated measures analysis much more gracefully.

Do I need to go on?

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

_________________________________________________________________ Stay in touch with old friends and meet new ones with Windows Live Spaces http://clk.atdmt.com/MSN/go/msnnkwsp0070000001msn/direct/01/?href=http://spaces.live.com/spacesapi.aspx?wx_action=create&wx_url=/friends.aspx&mkt=en-us