Mi estimate logit

[pgr451]

I want to run gender specific logit models under the the mi estimation.
When not running regular binary regression models, is used to do the following

sort by sex
by sex: logit var1 var2 var3

Is there anyway I can run the models this way using the MI estimate command?

[Aspen Chen]

Please see the FAQ on the advice for using full names.

The by option cannot be combined with -mi estimate-, but you can always do the sample with an if condition (or a for loop if there are multiple categories). There are two caveats, though.

a) Stata returns an error when an if condition produces varying estimation samples across imputations, or when the condition is based on any imputed or passive variables. You can still override the error with the esampvaryok option, but think twice about what it means. See page 22 of the - mi estimate- manual.
b) Conceptually, your model assumes there might be an interaction effect with sex. So consider using the by option in -mi impute- to impute separately by sex. See subsection titlted "imputing on subsamples" in the -mi impute- manual.