Re: st: Mixed negative binomial models?

[Stas Kolenikov <skolenik@gmail.com>]

As far as I understand the mathematics of this, a negative binomial is
a mixed Poisson with multiplicative gamma random effects. So if you
continue mixing it with independent gammas, you will continue having
the negative binomial -- in other words, there is no mixed negative
binomial per se. You can mix it with something else (e.g. normal, as
-gllamm- would do with Poisson), but the results will probably differ
by very little. I may be missing something, so if there are people on
the list with more solid knowledge of GLMs and their extensions (Joe
Hilbe?), you might want to wait and hear their input.

On Mon, Aug 30, 2010 at 7:24 PM, Troy Payne <paynetc@gmail.com> wrote:
> I've been through Rabe-Hesketh and Skrondal's (2008) book on
> multilevel modeling a few times and haven't seen a way to estimate
> mixed negative binomial models.  Can Stata run mixed negative binomial
> models?  I'm looking for something like xtmixed but for overdispersed
> count (Poisson) distributions.  Did I miss a way to do this in the
> myriad of options for gllamm?
>
> If there's not a method to estimate mixed negative binomial models in
> Stata, does anyone know offhand if other packages (e.g., HLM) can
> estimate these models?



-- 
Stas Kolenikov, also found at http://stas.kolenikov.name
Small print: I use this email account for mailing lists only.

*
*   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/