Alteratives to -mi estimate-

[paulvonhippel]

The Stata prefix mi estimate: helps users estimate a variety of models with multiply imputed data, but it's not compatible with every Stata command. For example, you can't run the user-written csdid command after mi estimate.

Are there alternatives to mi estimate, perhaps user-written, that might be compatible with the csdid command?

[daniel klein]

I have never used csdid (from SSC, I suppose).

As long as the command stores coefficients in e(b) and covariances in e(V) and if you are confident that mi estimate is theoretically justified, try the option cmdok, as in

mi estimate , cmdok : csdid ...

Other than that, various versions of the quite ancient mim package are floating around on SSC, SJ, and the author's websites. I believe it offers a bit more flexibility regarding where results are stored. But then, it should not be hard to write a wrapper around csdid that puts stuff where mi estimate expects it.