My dependent variable is a positive integer that I am modeling with Poisson regression. I have two instrumental variables that I need to incorporate into my model as well. I am not sure how to proceed with this. I know that proc qlim allows for IV analysis but I have only seen examples using logistic regression. This procedure doesn't work for Poisson, correct? Do I need to use non-linear two stage least squares regression to incorporate these instruments?
Not a solution (but it could be with some work), but it appears that PROC MODEL will handle a Poisson distributed dependent variable, and implements a non-linear two stage LS regression to do so. QLIM looks to be restricted to qualitative responses of various sorts.