Dear Brian, dear all,
Thanks for your quick response!
Not knowing what exactly Bart was looking for, I envisioned - as you
correctly assumed - a model in which the endogenous regressor is an ordered
categorical variable and the dependent variable is continuous (i.e. the
first stage is fitted via -oprobit- and the second stage via -regress-).
Now, my follow-up question referred to the corrected standard errors from
the second stage regression. By elaborating on the code provided (see
below), how could the corrected standard errors be "made" robust (as Stata
would do if you added the -, robust- or -, cluster(var)- options to an
existing/implemented estimation command) in a simple way.
Your guess that the treatment effects literature may have something to say
about that is most probably correct. Being a novice to this literature, I'm
wondering whether you could provide me with a couple of more precise
references? Thanks!
Best regards,
Tobias
Regarding robust covariance matrices for two-step estimators, there was an
article by James Hardin in the Stata Journal (2002, v. 2, #3). The tricky
part is in getting all of the required derivatives.
