Programming the statistical procedures from SAS

Mixed models for aggregate data analysis

Posts: 58

Mixed models for aggregate data analysis

I am running a straightforward Poisson regression model using aggregate data where the unit of analysis is Zip_code. The basic Poisson regression model looks like this. Since it is likely cases within Zip_Code are highly correlated, I was wondering if the modelling approach should be a random intercept model. Also, is it meaningful to use a mixed effects or random intercept model for aggregate data? The dataset was originally at individual level and analyzed within a mixed model framework to account for clustering within zip code which led to convergence issues.

Zip_Code Year Number_of_Cases Total_Population

123  2011  12  1200

123  2012  15  1250

123  2013  24  1200

999  2011  07  1000

999  2012  08  1100

999  2013  15  1000

proc GENMOD data=have;

     class Year;

     model Number_of_Cases=Year / link=log dist=Poisson offset=log(Total_Population);

     lsmeans Year / ilink;


Ask a Question
Discussion stats
  • 0 replies
  • 1 in conversation