Hi @AmyR , I think you are correct about the 1 (I am too used to treating time as a categorical variable so that I can fit a covariance structure to it), but I believe you still need the ResilTot term in your estimate statement, as the parameter estimate for time*ResilTot is an estimate of the deviation in slope from the ResilTot estimate.  And you should have 3 separate estimate statements, because as @jiltao pointed out, your current syntax provides a joint test, rather than individual tests. ... View more

As a work-around, you might consider reading mcmc_est_parms into the WORK library, and then hardcoding from TYPE= to TYPE=EST.  Without seeing that particular dataset, I can't say for sure what might be happening that would make the out= option lead to the behavior you mention.   SteveDenham ... View more

I don't believe your ESTIMATE statement is complete as written. There is no value to plug in for the main effect of ResilTot.  Please try:   estimate 'low resilience at time 1' time_rev 1 ResilTot 94.50 time_rev*ResilTot 94.50, 'moderate resilience at time 1' time_rev 1 ResilTot 135 time_rev*ResilTot 135, 'high resilience at time 1' time_rev 1 ResilTot 162.5 time_rev*ResilTot 162.5/e; I think the lack of significance for the conditional comparisons is coming from the fact that they are only being estimated at time_rev=1.   In this design, I would be tempted very strongly to consider time_rev as a categorical variable rather than as a continuous variable.  Then you could test whether the slope of adjusting for ResilTot is the same at all time points (See SAS for Mixed Models, in the chapter on analysis of covariance).   SteveDenham   ... View more

To add on to @PaigeMiller 's reply, the main effects being significant isn't really important as the interaction is significant.  For a design with two factors A and B , the main effect of A tests whether the levels of A differ, averaged over all levels of B, and similarly for B, but since you have a significant interaction, averaging over the levels misses the key fact that the differences between levels of A depend on which level of B you consider and again similarly for B.   SteveDenham ... View more

All of the factors you list seem (at least to me) to be fixed effects. I know you said your data was "clustered" and I can see that driving to GLIMMIX, but if you are in need of penalized estimation that won't help.  What about considering the clusters as a fixed effect or as strata, and getting narrow-inference space results.from LOGISTIC? I suppose this is a relatively gentle way of saying GLIMMIX really doesn't have an analog to the Firth algorithm that I am aware of, and that LOGISTIC is likely the best approach (or PHREG, as it looks like there are some examples in the documentation that might be of interest)..   SteveDenham ... View more

From your initial post, it appears to me that you want all possible comparisons within the various factors.  Given that, consider this approach:   proc mixed data=work; class id time group race ; model outcome= time group race time*group race*group time*group*race; Repeated time / Subject=ID type=cs; lsmeans group time*group race*group  time*group*race /cl diff ; The additions are in bold.  One thing to consider is that you now have a lot of comparisons, and without some adjustment for this multiplicity, many of the "significant" comparisons are liable to be false positives.  To handle this, read up on the adjust= option in the PROC MIXED documentation.   SteveDenham   ... View more

Could you show the model you are currently using?  I believe all of your questions can be resolved either by 'slicing' the LSMEANS or by writing appropriate LSMESTIMATE statements.  It just would help to know what you are fitting before trying to write any code.   SteveDenham ... View more