Hi everyone!
I am currently working on my Masters' thesis, I'm working with Poisson Fixed Effects but some things remain unclear to me.
My independent variable is job satisfaction (scale 0-10), my dependent variables are quantity smoked, alcohol consumption, days of exercise and days of eating particular healthy products; all non-negative count data with many zeros.
The control variables are lagged dependent variable, age and i.weekly_date.
I am using an unbalanced panel dataset (T=10 n=5,000).
I am running xtpoisson regressions with robust standard errors, since this is the most robust model (I think).
To validate my choice for this model I performed Hausman tests for my main regressions.
Five of my regressions show the expected p-value of 0.0000 but two them show chi2<0, one shows chi2=0 and one regression does not result in an answer, the re regression is still running after 3 hours.
What does this say about my model/regressions?
Are there Goodness of fit tests which can be used for a model like this?
Any help is greatly appreciated.
Please let me know if any complementary information is necessary.
Kind regards,
Marijke
0 Response to xtpoisson fe, Hausman and Goodness of Fit
Post a Comment