daisywang21@HOTMAIL.COM wrote back: > >Hi Peter, > >Sorry for missing the codes. By using the analog of education example, >here is the code I used. > >PROC MIXED DATA=work.test METHOD=ML IC NOCLPRINT scoring=5 maxfunc=500 >convh=1E-6 lognote; > CLASS school student outcome; > MODEL score = _var_ / s ddfm = bw OUTP=dept.OUT&num. ; > RANDOM INTERCEPT / SUBJECT=school GROUP=outcome /*s*/; > REPEATED outcome / SUBJECT=student(school) TYPE=CS; > *ods output CovParms=cp; >RUN;QUIT; > >The predictors (_var_ ) are about 10-15 variables plus their interaction >with outcome. There are about 50 outcomes for each of students nested >under school. So you can imagine how big the model becomes with the >interaction. I know it sounds crazy, but that's how it was designed. SAS >gave me all the estimated parameters, but not the prediction (error >happens to outp=). I'm wondering if there is anyway to get the prediction >as well. Thank you.

This looks like a survey sample to me. Is that the case? Do you have the survey design and survey weights?

If so, then PROC MIXED may not be the right tool to use here. You may need to start with something like PROC SURVEYREG instead.

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

_________________________________________________________________ Don’t miss your chance to WIN $10,000 and other great prizes from Microsoft Office Live http://clk.atdmt.com/MRT/go/aub0540003042mrt/direct/01/