I want to test the association between childhood behaviors at age six and trajectories of welfare use in early adulthood (3 levels). I’m using a multinomial logistic regression model and need to control for the clustering at age 6 at the school and classroom levels, all within a multiple imputations framework (i.e. MICE).

From what I’ve read, Stata doesn’t have a dedicated multilevel command for multinomial logistic regression – it can only be done using the gsem command. I’ve also read that the gsem command does not support multiple imputations, although it can be forced using the cmdok command.

The model I want to run is this:
mi estimate: ?command 3_level_outcome pred1 pred2 etc || school: || class:

Are there alternatives to using gsem for this kind of problem? If not, what would the gsem model look like? Note that I don’t plan to interpret effects at the school and classroom level, I just want to control for them.

Thank you.