I tried to use multiple imputation to deal with missing values in a panel data and, then, do logistic regression. This dataset has two waves with 9431 observations in each. The interval between two waves is one year. When I use the code "xtlogit", Stata reminded me " 6413 groups (10009 obs) dropped because of all positive or all negative outcomes". I know it because the dv did not vary between two waves. My question is should I accept this result or use random effect instead?

Additionally, when I applied multiple imputation, Stata generated a variable named "_mi_miss". This variable indicated that 4777 of the observations were marked as "incomplete" I checked the imputed variables whose "_mi_miss==1". None of them has a missing value. Why I got that variable? Does it affect my following analysis using "xtreg"?

mi set wide
mi register imputed $milist
mi impute mvn $milist, add(50) rseed(472195)
Thank you!