I am working with a dataset with survey design effects set (svyset) and using multiple imputation (mi). I am estimating probit models using endogenous treatment effects, so using eprobit with the entreat() option. These models estimate fine using
However, postestimation is causing big headaches.
doesn't work even outside the mi environment, because the treatment variable and its predictors within the entreat() clause usually return:
nconsistent estimation sample levels 0 and 1 of factor TREATMENTVAR
works outside the mi environment and without the svy: commands, but using estat teffects following svy: eprobit (regardless of mi status) returns:
invalid subcommand teffects
and running the same model under mi estimate: (even without svy
returns
estat teffects not valid
In short: I would like a marginal effects measure (via predict or via teffects) for arguably the most important question: the effect of the treatment on the dependent variable. Any advice will be much appreciated.
Code:
mi estimate, cmdok: svy: eprobit...
Code:
margins, dydx(*) predict(pr)
nconsistent estimation sample levels 0 and 1 of factor TREATMENTVAR
Code:
estat teffects
invalid subcommand teffects
and running the same model under mi estimate: (even without svy
returnsestat teffects not valid
In short: I would like a marginal effects measure (via predict or via teffects) for arguably the most important question: the effect of the treatment on the dependent variable. Any advice will be much appreciated.
