Hi Folks:
I'm analyzing spatially correlated data and using proc countreg. My health outcome is the number infectious disease counts at each city. The covariates rate_c1, rate_c2 and rate_c3 are at city level (250 cities). And, two other covariates are at province level (17 provinces): rate_p1 and rate_p2. I thought having covariates at different municipal levels cause random error. Therefore, I defined groupid (PID which identifies each province) and errorcomp in the model as shown below.
However, I get an error:
ERROR: Panel data models with spatial effects are not supported.
Am I using the correct SAS procedure when covariates are different levels and data has spatial auto-correlation?
Any pointers regarding the correct procedure to analyze a data with spatial effects with multi-level variables is greatly appreciated.
proc countreg data=combined Wmat=Wmatrix groupid=PID;
model n_outcome = rate_c1 rate_c2 rate_c3 rate_p1 rate_p2
/ errorcomp=random dist=negbin offset=ln;
spatialeffects rate1 rate3 rate4 rate_p1 rate_p2;
run;