In order to test for alpha = 0 you have test [lnalpha]_con = - infinity; where
lnalpha is the equation name.
Something like - test [lnalpha]_con = -500 - will produce output (on the
assumption that -500 is pretty close to negative infinity). However, given the
comments in the first technical note, page 6 [R]N-R, about the test occuring on
the boundary makes me doubt that -test- will give you the correct results in
this case.
Hope this helps,
Scott
----- Original Message -----
From: "Nevbahar Ertas" <[email protected]>
To: <[email protected]>; <[email protected]>
Sent: Wednesday, January 28, 2004 4:08 PM
Subject: Re: st: LR test result for negative binomial with clusturing
> Thanks. I checked that stata FAQ, but when I run for a wald test (by the
command .test) instead of LR, it tests for all the independent variables in the
model. It does not give me the test for alpha=0 in negative binomial. Any ideas?
>
> >>> [email protected] 01/28/04 16:03 PM >>>
> At 03:42 PM 1/27/2004 -0500, Nevbahar Ertas wrote:
> >I am trying to estimate a negative binomial regression model, and I am
> >also correcting for clusturing. I noticed that the output doesn't include
> >the LR test for alpha at the end of the table. Any ideas on how to get the
> >LR test result? Thanks.
>
> There is a Stata FAQ entitled "Why should I not do a likelihood-ratio test
> after an ML estimation (e.g., logit, probit) with clustering or pweights?"
> that might address your question. See
>
> http://www.stata.com/support/faqs/stat/lrtest.html
>
>
*
* 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/
