The asymptotic distribution of the differences in log
pseudo-likelihoods has a complicated distribution which is the
difference of two sums like
sum_j a_j v_j
where a_j are eigenvalues of [vce(opg) times inverse(vce(oim))], and
v_j are independent chi^2_1 variables. Of course the resulting
distribution will be non-standard. (People familiar with structural
equation modeling would easily recognize Satorra-Bentler scaled
difference test in this :)). Are you really sure you want to deal with
that crap? Wald test with your robust standard errors will do just
fine to compare two nested models.
On Mon, Dec 14, 2009 at 6:24 AM, Thomas Klausch
<[email protected]> wrote:
> Dear Stata listeners,
>
> I want to compare nested negative binomial regression models with
> robust standard errors (using -nbreg, robust-). Is it permissable to
> compare the pseudo-likelihoods of the nested models with a likelihood
> ratio test? Is this test statistic still chi-squared distributed? I
> think I have come about something like a pseudo-likelihood ratio test,
> but do not find any details in my literature on this question. Thank
> you in advance.
>
> Thomas
> *
> * 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/
>
--
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/
