Due to sparse data and separation issue, I am running Firth or exact logistic regression instead of regular logistic regression. The statistics of interest is risk difference instead of odds ratio. I noticed that I can run "%nlmeans" with output from proc logistic firth but there is warning message "WARNING: The final Hessian matrix is not positive definite, and therefore the estimated covariance matrix is not full rank and may be unreliable. The variance of some parameter estimates is zero or some parameters are linearly related to other parameters."  Are the output still valid? If not, is there any alternative? How about obtaining risk difference from exact logistic regression? ... View more

I am running MMRM on log10 transformed outcome. There is some medical reasoning for log10 transformation instead of natural log. I would like to model using SAS PROC MIXED. To get geometric mean and geometric ratio, I would need to reverse the transformation. I understand that direct reverse transformation gives median instead of the expected value (Solved: Non-normal data; PROC GLIMMIX - SAS Support Communities). How should I go about this? I would like to get GMR of treatment versus control at a specific time point.   proc mixed data=final2; class id trt avisit; model log10aval=trt avisit log10base trtp*avisit / S; repeated avisit / subject=id type= UN R; LSMEANS trt*avisit / PDIFF CL; estimate "Visit 9 - Week 52" trt 1 -1 trt*avisit 0 0 0 0 1 0 0 0 0 -1 / CL; run;     ... View more