 Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# Re: st: Poisson regression - Test of goodness of fit

 From Steven Samuels To statalist@hsphsun2.harvard.edu Subject Re: st: Poisson regression - Test of goodness of fit Date Mon, 28 Feb 2011 12:36:09 -0500

```Alexandra-

1. If you are interested only in effects of predictors on the mean count, then failure of the Poisson assumption might not matter much.
2. If you wish to predict the individual probabilities as well as the mean count, then other survey commands offer more flexible models, e.g. -svy: nbreg-  -svy: zip-.  See the -help- for "survey estimation".
3. You _must_ use the survey version of one of these commands.  The output you show is for the non-survey commands. There will be no likelihood for the survey version.
4. For testing the fit of the model for the mean, you can use -linktest- (with some continuous predictors), add polynomial terms (or fractional polynomials -fpoly-), interactions and -test- augmented against reduced models.

Steve

Steven J. Samuels
Consulting Statistician
18 Cantine's Island
Saugerties, NY 12477 USA
Voice: 845-246-0774
Fax:   206-202-4783
sjsamuels@gmail.com

On Feb 28, 2011, at 6:10 AM, Alexandra Boing wrote:

I have doubts about the use of poisgof - Test of goodness of fit (command Stata 8= poisgof Stata 9= estat gof) in a Poisson regression in which I have to run with svy. The poisgof and estat gof not accepted. Is acceptable only when run a Poisson regression without the svy.
How should I proceed as it is necessary to use in my case the svy? What is problem run Test of goodness of fit (poisgof OR estat gof) with Poisson regression without svy. This is a problem?
Another question I have a model output and doubt in the interpretation of the output: pseudo R2=0.0562, LRchi2(6)=71,46, goodnessfit=789.0331, chi2(1665)=1.0000
This is good? Is appropriate? What is best option of analyse?

Poisson regression                           Number of obs   =       1672
LR chi2(6)      =      71.46
Prob > chi2     =     0.0000
Log likelihood = -599.51655                  Pseudo R2       =     0.0562

Goodness-of-fit chi2  =  789.0331
Prob > chi2(1665)     =    1.0000

Att. Alexandra

*
*   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/

*
*   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/
```