My dataset has 995 adolescents who completed questionnaires over T=2 timepoints. I am focusing on examining the effect of several predictors (family structure and relationships) on a binary outcome (adherence or non-adherence to a medication regimen). I was using mi estimate: xtlogit, fe (equivalent to clogit) in Stata 14.2 to estimate conditional fixed effects logistic regression to understand if a change in family circumstances is linked to a change in the odds of the medication adherence outcome.

For presenting the results, in addition to odds ratios from the regression models, I would like to include average semi elasiticity using aextlogit command developed by Joao Santos Silva because, as I understand, it is not recommended to use margins with clogit, and odds ratios are not always easy to communicate to those who are not researchers.

1. One of the predictors for which I would like to estimate the average semi elasticity is a continuous indicator of child-parent communication. I was wondering if Professor Silva or others on the forum may be able to advise whether it is possible with aextlogit to look at semi elasticity at different levels of the predictor, as one does with margins? (e.g. to estimate the probability of the outcome for those with best, or worst level of communication, the predictor?)

2. Is it possible to use aextlogit with mi multiple imputation suite in Stata? (I have about 12% missing data on one of the predictors).

I am new to the Stata forum, please let me know if it is more helpful to email this question - I saw a few threads on aextlogit, and so thought it may be helpful to have a discussion for other users trying to get at predicted probability from a fixed effects regression. Thank you!