I suspect that it is not possible to condition out the fixed effects
in the zero inflated negbin model (even though it possible to do so in
the fixed effect logit model and the negbin model). If I am wrong, I
am sure someone else will correct me.
BUT:
why do you need this? even if you could run such a model, the
estimates would only be valid under a set of rather restrictive (and
difficult to test) distributional assumptions. I would argue that
xtpoisson with QML standard errors (and not the ML standard errors
produced by xtpoisson) is actually what you need. The only thing it
assumes is that you can write E[y|X]=exp(X'beta), which will be true
even if there is a big mass point at 0 for the dependent variable.
Nothing about the variance and higher order moments. Tim Simcoe at
Toronto has developed a variation on xtpoisson that will do what you
need.
For more on the QML vs ML and count models, refer to chapter 19 in
Wooldridge's econometric textbook.
Pierre
---------------------------
Hello,
I would like to know if anyone has written code to implement the fixed
effects ZINB? I would really appreciate your assistance. The standard
set up in Stata only implements fixed effects models for the standard
Poisson and negative binomial models.
Thanks.
Sharon James-Wade
Assistant Professor
Management & Human Resources
744 Fisher Hall
The Ohio State University
2100 Neil Avenue
Columbus, OH 43210
O: (614) 292-0363
F: (614) 292-7062
sjwade@cob.osu.edu
--------------------------------------------------
Pierre Azoulay
Assistant Professor of Strategy
Massachusetts Institute of Technology
Sloan School of Management
50 Memorial Drive — E52-555
Cambridge, MA 02142-1947
Tel [Sloan]: (617) 258-9766
Tel [NBER]: (617) 868-1464
Fax: (617) 253-2660
*
* 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/