Generating Predicted Probabilities using meqrlogit

[MichelleS Wong]

I am trying to generate predicted probabilities from a logistic model (2-level model of individuals nested within sites, and model includes an interaction between two categorical variables [race and ethnicity x sex]). I would like to generate predicted probabilities and 95%CI for each of the race and ethnicity by sex categories. Unfortunately, the model did not converge using -melogit-, so I had to use -meqrlogit-. When I run the -margins- command after --meqrlogit-, I believe this does not generate predicted probabilities, but instead these are just predicted means for the fixed effects only. How do I generate the analogous predicted probabilities from the -margins- command after -melogit-?

[Clyde Schechter]

I think that post-estimation is very limited after -meqrlogit- and I don't know of any way to get around that. So let me ask you why your model failed to converge with -melogit-. Often, non-convergence is due to problems with model specification or bad data. Can you show example data, the exact -melogit- command used, and the complete output you got (including the last half-dozen or so lines of the iteration log) when you tried using -melogit-?