Dear all,

I would like to implement a version of -heckman- with too many fixed effects than my computer can compute if I include them explicity. As I've seen, Stata 16 thankfully has a dedicated command -xtheckman- for this, but unfortunately I'm scheduled to get my Upgrade to Stata 16 only in 2-3 months, while I would like to obtain some estimates now.

In principle I suppose I should be able to either estimate first the 1st stage with xtprobit and then the second stage with xtreg manually, but I am not sure how exactly to code this while ensuring that the standard errors of my second stage will be correct.

Would anyone be able to translate
xtset crossecvar timevar
xtheckman y x1 x2 x3
into a correct "manual" implementation?

Thank you so much!
PM