topic Re: Proc genmod with normal distribution, log link, unstructured variance-covariance and strata in Statistical Procedures

Proc genmod with normal distribution, log link, unstructured variance-covariance and strata

Liamb — Tue, 02 Jul 2024 16:37:10 GMT

Hello,

I'm looking for SAS code to calculate the difference between two groups using the Generalized Estimating Equation (GEE) with normal distribution assumptions, log link functions, unstructured variance-covariance structure and stratification factor (strata).

I found the following code in the SAS help but it doesn't specify the variance-covariance structure and strata options

proc genmod data=nor;
   model y = x / dist = normal
                 link = log;
   output out       = Residuals
          pred      = Pred
          resraw    = Resraw
          reschi    = Reschi
          resdev    = Resdev
          stdreschi = Stdreschi
          stdresdev = Stdresdev
          reslik    = Reslik;
run;

Thanks

Re: Proc genmod with normal distribution, log link, unstructured variance-covariance and strata

StatDave — Tue, 02 Jul 2024 17:50:22 GMT

Assuming that you want to use GEE because you have multiple observations or subjects within the strata and you want to allow for those responses to be correlated, then the following does the analysis you are suggesting. PROC GEE is the recommended procedure for fitting GEE models, though you could use the same syntax with GENMOD. But note that if the number of observations within a stratum is large, the model with the unstructured correlation matrix could be hard to estimate. Since GEE protects for incorrect structure, many just use simple structures like TYPE=IND or EXCH.

proc gee;
class strata group;
model y=group / link=log;
repeated subject=strata / type=un;
lsmeans group / diff;
run;

Re: Proc genmod with normal distribution, log link, unstructured variance-covariance and strata

Liamb — Tue, 02 Jul 2024 18:32:55 GMT

Thanks a lot