I'm trying to fit various models (I had been doing negative binomial, but then had to switch to Poisson in some cases because of the model being saturated) and am unsure whether I should be doing type 1 or type 3 analysis.

I have 3 study areas and hospital admissions over time (yearly), so I want to compare admissions between the 3 study areas AND over time (so mainly interested in the area*period term rather than area or period on its own). From my understanding, type 1 analysis is sequential and the variables are added one class variable at a time whereas type 3 compares the full model to the model without the class variable in question. What I read said to use type 3 by default, but someone who had helped me initially told me to use type 1. So I'm not sure which I should use for what I need to be doing - is it necessary for the variables to be added sequentially to the model or is type 3 what I should be using instead? Thanks!