RE: st: Prediction after xtpoisson, fe

[Maarten buis <maartenbuis@yahoo.co.uk>]

--- david reinstein wrote
> I am regressing a non-negative non-count variable (charitable giving)
> on a host of other variables ("price", income etc) using poisson
> regression with a fixed effect <snip> I then try to generate
> predictions (and residuals) for charitable giving for each
> (household/year) observation, using (as suggested in stata help):
> -predict psnfexb_tot, xb-
>
> But this does not seem to generate reasonable predictions .
> Summarizing the data, I get a mean prediction of 1.87, versus a mean
> dependent variable of  2092.  My guess is it is leaving out the fixed
> effect in the prediction.

The main reason for this difference is the fact that you did not 
predict the predicted donation but the linear predictor (xb). Try
-predict yhat, nu0- instead. These would still be off a bit, for the 
reason that you hypothesized, but they will be much more reasonable.

Hope this helps,
Maarten

-----------------------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://home.fsw.vu.nl/m.buis/
-----------------------------------------


      

*
*   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/