Re: st: problem with truncreg

[Rich Steinberg <rsteinbe@iupui.edu>]

Thanks to you and to Jay Verkuilen.  We have done a fair amount of 
transforming in tobit, and the form we choose worked best there so we 
were hoping to keep it the same here.  Our ultimate goal was to manually 
do something like a double hurdle (truncreg and probit), but we finally 
found a way to get our real double hurdle model to converge.

Austin Nichols wrote:
> Rich Steinberg <rsteinbe@iupui.edu> :
> Perhaps the model (truncated normal error) is simply not very
> appropriate for your data--you might try transforming your variables
> if that makes sense in your setting.
> webuse laborsub, clear
> truncreg whrs kl6 k618 wa we, ll(0)
> g s=(whrs)^2
> truncreg s kl6 k618 wa we, ll(0)
>
> On Tue, Sep 22, 2009 at 4:44 PM, Rich Steinberg <rsteinbe@iupui.edu> wrote:
>   
> > I've been trying truncreg and no matter what maximization technique I
> > try, I get a few "backups" and then the iteration terminates with the
> > following message:
> >
> > numerical derivatives are approximate
> > nearby values are missing
> > numerical derivatives are approximate
> > nearby values are missing
> > cannot compute an improvement -- discontinuous region encountered
> > r(430);
> >
> > Shortening my list of control variables did not eliminate the problem.
> >
> > I understand problems with concavity.  Although they are present, they
> > do not seem to be the source of this error.
> > Perhaps I am being dense, but what could cause the likelihood to become
> > discontinuous? Is there a fix up or is this problem inherent.  (Note --
> > this is unconstrained).
> >
> >     
>
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/statalist/faq
> *   http://www.ats.ucla.edu/stat/stata/
>   
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/