Poisson Regression Residuals

[StataPerson]

I'm modeling the number of doctor visits (a count variable) on factors such as income, chronic condition, insurance, etc. I use the canned Stata command poisson y x1 x2, etc. but I am encountering the following issue when trying to find the sample average of the residuals:

Stata does not allow me to execute predict resids, residuals after I've ran the poisson regression above.. I then tried to generate the residuals using the formula "y-yhat", yet I get a sample average of residuals with a negative average mean (-6.96e-09...) - is this 2nd approach even correct ? Also, I tried fitting the model with glm, f(poisson)and then running

predict resids, residuals ,but Stata only allowed me to run it as "predict resids" and automatically added " (option mu assumed; predicted mean docvis)" and as a result the sample average of residuals with this approach is a constant value of 3.81 throughout all observations!

Any thoughts here?

[Maarten Buis]

That question was asked and answered here:
http://stats.stackexchange.com/quest.../108694#108694

Note that the Statalist FAQ asks you to mention that you cross-posted your question elsewhere.

Also note that the Statalist FAQ asks you to post under your real name. This has a long tradition on this list, and we believe that it helped maintain a friendly and professional tone on this list. To change your name you can use the "contact us" button on the bottom right of the page.

[Nick Cox]

Although Stata Person (please change to your real name, as Maarten underlines) should have mentioned the post on Cross Validated, the extra problem with predict after glm is explained by Stata itself: you just have to read the help carefully. The syntax for predict includes options

 
 mu              g_inverse(xb); the default
 anscombe        Anscombe residuals
 deviance        deviance residuals
 likelihood      a weighted average of standardized deviance and standardized Pearson residuals
 pearson         Pearson residuals
 response        differences between the observed and fitted outcomes
 working         working residuals

so to get a flavour of residuals, just use one of the appropriate options. As Stata is telling you, predict defaults to showing the mean.