st: margins with nocons option in Discrete choice model

[Robert Duval <rduval@gmail.com>]

Dear Friends,

I estimate a discrete choice model with heteroskedasticity (similar to
a bivariate probit but with some variations).

This means that in my likelihoold i have terms that look like

P(y=1)=F(xb1/exp(zb2))

where xb1 is the main index of interest and exp(zb2) is a term
included to model potential heteroskedasticity (see Wooldridge's
EACSPD (2010) p.601-602).

The index zb2 must exclude a constant. However, when I run margins
after the model has estimated I get the message that margins are
non-estimable (I have dummies entering zb2).

I checked the H matrix and indeed there are entries different than
(-1,0,1) associated to the zb2 equation. Also, I read in this list
that for some reason margins doesn't always produces results when the
'nocons' option is specified (as in my case for the equation zb2).

My question is whether it would be OK to proceed and estimate the
margins with the noestimcheck option for this type of (heteroskedastic
DV) models.

Many thanks in advance,
robert
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/