Abstract
The paper deals with the small sample performance of estimators and tests in the overdispersed Poisson regression model. Maximum likelihood and semiparametric estimators and corresponding t-test statistics are compared in a Monte Carlo experiment. Three tests of overdispersion are evaluated. The results suggest that differences between estimators are small. The semiparametric estimator and the corresponding t-test perform well. Two of the studied overdispersion tests are found to have reasonable size properties. Extensions to limited dependent Poisson regression models are discussed.

This publication has 11 references indexed in Scilit: