Trouble with Margins and Marginplots in Imputed GSEM Models

[Reza Tayari Ashtiani]

Hello,

I am trying to get margins and marginplots for the interaction in my model but I get the "last estimate not found" error:

this is the code I am using: mi estimate, cmdok: margins ethgrp, predict(equation(between_collective_efficacy)) at(between_deprivation2=(-2(1)2))

and this is my model:

mi estimate, cmdok: gsem (between_deprivation2 -> connectedness, ) (c.between_deprivation2##i.ethgrp -> between_collective_efficacy, ) (between_deprivation2 -> log_loneliness, ) .................. covstruct(_l
> exogenous, diagonal) latent(M1 ) cov( e.connectedness*e.log_loneliness) nocapslatent

Am I using the correct command? I would really appreciate it if you could help me with this.

[Clyde Schechter]

No, you can't use -margins- on the results of multiple imputation estimates. Trying to force it with -mi estimate, cmdok- still fails. The problem is that -mi estimate: estimation command- does not leave the multiply imputed estimation results posted in e(b) and e(V) the way an ordinary estimation command does. There is a separate command -mimrgns-, by Dan Klein, available from SSC that you can use instead.

Do read the help file carefully, and take heed of the various warnings given therein.