ABSTRACT

Poisson regression model with an offset is explained. Short comparison with linear regression. The constant term (intercept) actually estimates a rate ratio, not a rate. How standard errors are estimated. The iteration log and header statistics are discussed. Log likelihood, deviance, and the likelihood ratio statistic are presented. Generalized linear model command shown for the same model. An example shows how to estimating changes in rates over time. A regression trick is shown; alternative parameterization of the Poisson model using importance weights.