Dear all
I have a similar problem to the previous query "Model fit index after xt command".
I am using Poisson regression on district death counts with my exposure as the expected number of deaths (as part of an ecological study examining risk factors for COPD mortality in UK districts). My data are overdispersed. The rpoisson command, Poisson regression with frailty, gives me the lowest deviance by a long way - close to the number of degrees of freedom - the next best command nbreg, negative binomial regression, gives me a deviance an order of magnitude higher than the degrees of freedom.
However, model comparison looks to be difficult. Fitstat doesn't work after rpoisson. I don't think I'm able to calculate the likelihood ratio or AIC manually as I'm not sure how many extra parameters the model uses to estimate the overdispersion. Can anyone help with this?
Thanks
Anna Hansell
Wellcome Research Fellow in Clinical Epidemiology
Dept Epidemiology & Public Health
Imperial College London
Norfolk Place
LONDON W2 1PG
UK
-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of Shige Song
Sent: 02 April 2004 06:55
To: [email protected]
Subject: st: Model fit index after xt command
Hi,
Fitstat has been an invaluable tool for model comparison, but it does not
seem to work with xt command series. Does anyone know whether there are
other options out there to compute AIC and BIC automatically after xt
commands? It is not difficult to manually calculate them, but it'll be
nice to automate the process. Thanks!
Shige
*
* 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/
*
* 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/
