Re: st: xtnbreg discontinous region encountered after rescaling

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- On Fri, 10/12/10, hl wrote:
> I am running a series of fixed
> effects negative binomial panel regressions (xtnbreg, fe) on
> different dependent variables. I have encountered some
> specifications that do not converge or cannot be computed
> due to discontinous regions before. However, one specific
> regression that worked fine suddenly returns
> 
> discontinuous region encountered
> cannot compute an improvement
> r(430);
> 
> after I rescaled one independent variable. I didnt change
> anything substantial, just divided by factor 100 - how is
> that possible?

A negative binomial model needs a lot of information to be
estimated, and a fixed effects models throws a lot of 
information away (all comparisons between individuals), so
I am not surprised that you have trouble estimating your 
model. In these cases numerical issues can easily matter.
It is ofcourse hard for us to determine what happend given 
the information you provided, but I guess that that is what 
you ran into.

Hope this helps,
Maarten

Ps. Note the remarks on anonymous question in the Statalist 
FAQ section 2.1 sub 3:
<http://www.stata.com/support/faqs/res/statalist.html#tojoin>

--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://www.maartenbuis.nl
--------------------------


      

*
*   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/