Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Fixed Effects Poisson Model with Unstructured Correlation Matrix


From   nicola.baldini2@unibo.it
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Fixed Effects Poisson Model with Unstructured Correlation Matrix
Date   Sun, 27 Apr 2008 18:53:46 +0200

-xtgee- does the so-called population-averaged estimators. You can't have FE from it, and I'm not knowledged enough to suggest you to insert dummies by yourself while using -xtgee- (to my unexperienced and non-statistical brain it sounds not a good idea). 
As a second-best, -xtpqml- (which you can download from SSC) fits conditional fixed-effects Poisson (quasi-maximum likelihood) models with clustered robust standard errors (no ways to specify the structure).
About the conditioning issue, I copy and paste from Allison and Waterman, FIXED-EFFECTS NEGATIVE BINOMIAL REGRESSION MODELS, Sociological Methodology, 32, page 249:
"For Poisson regression, on the other hand, these two estimation methods —unconditional maximization of the likelihood and conditional likelihood— always yield identical estimates for b and the associated covariance matrix (Cameron and Trivedi, 1998)."
The reference in there is Cameron, A. Colin, and Pravin K. Trivedi. 1998. Regression Analysis of Count Data. Cambridge, England: Cambridge University Press.

Warning: I didn't set Statalist-disturbing automatic replies. Thus please note in advance that I will be offline for the next 8 days at least, and thus unable to continue this discussion for a while.
Nicola

At 02.33 26/04/2008 -0400, "ippab ippab" wrote:
>Hi Statalisters,
>
>I have an unbalanced panel data set with 1150 panels (groups) and the
>dependent variable is a count.  I am estimating a fixed effect model
>with "xtpoisson, fe" command.  I am wondering if there is any way to
>specify the structure of the correlation matrix.  In particular, I am
>interested in estimating the model with "unstructured" correlation
>matrix.
>
>Relatedly, I would really appreciate if you can tell me how to do
>fixed effect models with xtgee.  Is there a way to create the fixed
>effect dummies in xtgee command?  I suppose I can specify a structure
>in xtgee.  Then, I am worried whether or not the unconditional fixed
>effects model would provide me drastically different estimates than
>the conditional fixed effect model (xtpoisson, fe).  I would
>appreciate your insights on this issue.
>
>Thanks a lot!
>
>Best regards,
>
>ippab 


*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index