Hello,
I am trying to estimate the exchangeable working correlation for classroom and school clustering in a 3-level model.
I can estimate the ICC using GLIMMIX, however, I am wondering if it is possible to estimate a 3-level model using GENMOD as well?
Currently, my two level model is producing an exchnageable working correlation with the folloiwng code:
proc genmod daya=sel;
class schoolid sel;
model sel (ref=first)=/D=B link=logit;
repeated subject schoolid/type=exch covb;
run;
When I try to add in a third level, the following code works but only provides one exchangeable working correlation. Is there a way to separate the class from the school exchangeable working correlation in GENMOD? If so, how do I fix the following code?
proc genmod daya=sel;
class schoolid classid sel;
model sel (ref=first)=/D=B link=logit;
repeated subject schoolid/type=exch covb;
repeated subject classid(schoolids)/type=exch covb;
run;
Thanks in advance!