Thank you so much for your response! I checked the frequency chart but couldn’t find any abnormal from it. Then I tried to reduce the number of covariates and found when the main covariate was rejected, warnings disappeared. However, the covariate has been proved to be associated with the outcome in multivariate logistic analysis. And I don’t think it’s a good way to get rid of that covariate. ... View more

Hello, I'm handling the same problem recently. Since the outcome variable is not rare, the log link was suggested in SAS help.  I've read these solutions you suggested, but as you can see, I've no reason to reduce those covariates, and there is no interaction option initially, so what else can I do to treat this problem, or if I can ignore the warnings?   code is as follows: proc casusalmed data=data all; class X(ref=first) M(ref=last) Y(ref=first) C(ref=first); model Y=X M /link=log; mediator M=X; covar c; run;   ... View more