Conditional probability after eprobit

[Nikolaos Kanellopoulos]

Dear all,

I am using eprobit in Stata 15 to estimate a probit regression with endogenous sample selection.
After the estimations I want to calculate the probability of success conditional on selection and the selection probability.
I know I can use heckprobit to that (see following example), but I wonder if it is possible to do it with eprobit.

Regards,
Nikos

clear *
webuse heartsm
eprobit attack age bmi i.exercise, select(full = age bmi i.checkup)

// This is equivalent to the above
heckprobit attack age bmi i.exercise, select(full = age bmi i.checkup)

// This is want I want to estimate with eprobit
margins , predict(psel)
margins , predict(pcond)

[Richard Williams]

If there is I don’t see it. This would seem like a good thing to add. But why would you want to if heckprobit does it?

[Nikolaos Kanellopoulos]

Dear Richard thank you for the clarification.
I prefer eprobit to heckprobit as in fact the model I calculate includes an endogenous variable. I just preferred to keep my question as simple as possible.

One solution I thought of is to use cmp as in the example below.

clear
webuse heartsm
eprobit attack age i.exercise, select(full = age bmi i.checkup) endogenous(bmi = age exercise exintensity )
cmp setup
cmp (attack = age i.exercise bmi) (full = age bmi i.checkup) (bmi = age exercise exintensity ) , nolr ind(full*$cmp_probit $cmp_probit $cmp_cont)
margins,         predict(pr eq(full))
margins,         predict(pr eq(attack) condition(0 ., eq(full)))