On 2 Mai, 17:14, Ryan <Ryan.Andrew.Bl...@gmail.com> wrote: > Hi everybody, > > I have a quick question about interpretation of results from a GLIMMIX > ESTIMATE statement. > > Here's the code: > > ------------------------------------------- > > proc glimmix data=mydata; > class condition time ; > model y = condition|time|covariate / s dist=binary link=logit; > > Estimate 'condition 1 versus condition 2' condition 1 -1 0 0 0 > condition*covariate 10 -10 0 0 0 / exp; > > random intercept / subject=person; > run; > > ----------------------------------------- > > *Will the estimate statement tell me the odds of condition one on "y" > relative to the odds of condition two on "y," after setting the > covariate = 10? > > I ask because this estimate statement is based on a lower order term > (condition*covariate) when there is a higher order term in the model > (condition*time*covariate). Please let me know if this is acceptable. > BTW, it is important to have the higher order term in there as that > answers a separate, but equally important research question. > > TIA

Dear Ryan, writing correct ESTIMATE statements is a tricky thing and I am far from being an expert. But if you want to make statements about effects at certain fixed continuous covariate levels I feel that you should use the AT= option.

Hope that helps, Oliver