I am using cmp to estimate a fractional probit and a probit at the same time.
There are about 5,000 variables, most of which are binary fixed effects.
The number of observations varies between 52,000 and 1,100,000 depending on the model.
cmp runs fine, it takes a little longer than I would like, but it's not crazy.
Using mfx afterward takes far too long to be practical.
mfx compute, varlist(Indepdent variables) predict(pr equation(outcome1))
I've also tried adding the "force" option to that mfx command which did not make a noticable difference.
It has run for over 50 hours before I just decided to break it and ask here for help.
I feel like stata is iterating far too many times or calculating effects for each observation in an inefficient manner, is there a better command for estimating marginal effects or a way to make mfx run more efficiently? I'm currently trying it with the margins command, but I doubt it's any different.
Thanks in advance!
There are about 5,000 variables, most of which are binary fixed effects.
The number of observations varies between 52,000 and 1,100,000 depending on the model.
cmp runs fine, it takes a little longer than I would like, but it's not crazy.
Using mfx afterward takes far too long to be practical.
mfx compute, varlist(Indepdent variables) predict(pr equation(outcome1))
I've also tried adding the "force" option to that mfx command which did not make a noticable difference.
It has run for over 50 hours before I just decided to break it and ask here for help.
I feel like stata is iterating far too many times or calculating effects for each observation in an inefficient manner, is there a better command for estimating marginal effects or a way to make mfx run more efficiently? I'm currently trying it with the margins command, but I doubt it's any different.
Thanks in advance!
