| ![](/includes/contimages/spacer.gif) |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: RE: Random Effects vs GEE NBREG Models
I'm trying to estimate a panel model with an overdispersed count response. The response variable was assessed for an average of 157 days following baseline. All of the predictors are time invariant subject characteristics assessed at baseline. The assumption that the dispersion parameter varies randomly by subject seems plausible. But my understanding is that the model dispersion parameter is constant for all elements in the same group (here, days for each individual). Would this assumption be violated if we expected the count response to vary considerably from day to day for each individual? A comparison of the random effects model gives rather different conclusions than a population averaged GEE model using an exchangeable correlation structure. How does one go about choosing the appropriate model specification? If someone could point me toward appropriate references I'd also be most appreciative. Thanks much in advance!
Brad
*
* 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/