Help with mixed effects logistic regression in STATA


I'm new to using mixed effects modeling, and I'm running a mixed effects logistic regression in STATA as my data were collected at different schools. Everything looked good when I ran the model, but after using multiple imputation to handle missing data, the random effects portion of the output looked strange (i.e., much larger standard error and no confidence interval). Here is the code and portion of the output:

mi estimate: xtmelogit y i.x1 x2 x3 x4 x5 || school:
Random-effects Parameters | Estimate Std. Err. [95% Conf. Interval]
school: Identity |
sd(_cons) | .000012 916.0454 0 .

Any ideas as to why this might be happening?


I'm not entirely sure - you tend to get this kind of odd result (tiny SD of random intercept with huge standard error) when you don't have enough data points. If the model worked well without the imputations, perhaps there was a problem with your imputations, or perhaps you don't have enough of them. In general it's helpful if you show us the commands and output you used to get to this point - we might be able to detect the problem.

Incidentally if you don't plan to make the model more complex you don't need to use -xtmelogit-, you can use -xtlogit- which is a lot faster (and therefore makes it easier to get to the bottom of your problem):
mi xtset school
mi estimate: xtlogit y i.x1 x2 x3 x4 x5