Roberto,
I'm no expert on ordered probit with a selection equation,
but if the 'normal' heckman twostage model needs a
correction than the 'ordered probit' heckman twostage model
will probably also need a correction. However, I don't know
the correction, nor know of any published work on it (but
since I am no expert, that does not mean much).
Alternatively, you could implement the model using Stata's
-ml- command. The maximum likelihood estimates don't need a
correction.
Since this isn't much of an answer, I have forwarded this
message to the statalist in the hope somebody else can give
you more advise. (Even if I knew the answer it would be good
to let such a discussion go over the statalist, because:
a) I might make a mistake and somebody else might catch it,
and b) other people may be following this thread and want to
know the answer.
HTH,
Maarten
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting adress:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
--- Roberto Fontana wrote:
> I agree with you that proit - oprobit is the best option.
> To my knowledge, simply using the IMR does not correct the standard errors
> in the second stage (or does it?). Are you aware of a procedure in Stata
> that allows to correct them? More in general do you know a published paper
> that have employed the two-stage procedure with probit-oprobit and Mills ratio?
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
