I am working on data for 10 years that has counts of hospitalization and population size by countries. I also have other covariates that I need to control for.
Using GEE Poisson regression, and autoregressive correlation structure, the p-value for z-statistics for parameter estimate and Type3(Score test) results don't match for some continuous variables.
(IN SAS I have used PROC GENMOD; REPEATED statement with link=Poisson, Type=AR and offset=popsize)
What can be going on?
I am not able to share the data or results due to data confidentiality.
