Why does the Zelig package simulate different categories depending on order of model specification?

Question

I have register data of compulsory hospitalisations. The risk of compulsory hospitalisation varies with age and gender. I want to describe the geographic variation after con