Hello, I am using PROC GLIMMIX for the first time and do not have any expertise in spatial analyses. I have Zip Code level data with positive spatial autocorrelation. I am interested in running a regression model to examine the association of some area level predictors with the number of people with the illness of interest in an area. I am using PROC GLIMMIX to account for clustering at Zip Code level and for spatial autocorrelation. When I use the following statements as Step 1 (to simply account for clustering at Zip Code level), PROC GLIMMIX model converges and I get an output that makes practical sense however with overdispersion (the model does not converge when I use ‘random _residual_;’ statement for overdispersion, or any other DDFM method). Count is the number of people with the illness. All predictors are continuous variables. White, black, adults, older_adults are proportions. Offset is the log of population. There are about 12,000 observations at ZC level in the ZC1 dataset. Step 1: Proc Glimmix data = library.ZC1 NOCLPRINT ; class ZC; model Count = white black adults older_adults averageHHsize/ dist = poisson offset = lnPop solution DDFM=BW; random intercept / Subject = ZC ; run; In Step 2, I run the same statements but this time adding the ZC centroid information (X and Y represent the latitude and longitude of the ZC). Model converges once again but I get an output exactly the same as in Step 1 and still have overdispersion. Here are the SAS statements from Step 2. Can someone please tell me what am I doing wrong here? Thanks a lot. Step 2: Proc Glimmix data = library.ZC1 NOCLPRINT ; class ZC; model Count = white black adults older_adults averageHHsize/ dist = poisson offset = lnPop solution DDFM=BW; random intercept / Subject = ZC type=sp(exp)(X Y); run;
... View more