Unbalanced nested random effects in PROC GLIMMIX

[MeasuringHealth]

Hi everyone,

Is it a problem if the subgroups specified by nesting in my RANDOM statement are unbalanced? For example, below you can see id(Diagnosis) where there are 5 diagnoses but the number of ids (patients) with each diagnosis range from 8 to 16 (e.g. 8 patients have diabetes, 16 patients have cancer, 10 patients have epilepsy). Do I need to adjust for this in my syntax? If so, what syntax do I need to add? 

PROC GLIMMIX DATA=CONC_long; 
         CLASS Time (ref= '1') Diagnosis; 
         MODEL LoHiSev = Composite / dist=binomial link=logit solution; 
         RANDOM intercept / subject=id(Diagnosis) type=CS;
         OUTPUT out=preAUCnested1 pred=xbeta pred(ilink)=predprob pred(ilink noblup)=fix_predprob;
RUN;

Thank you, SAS Community!

Best,

MeasuringHealth

[Haris]

Generically speaking, SAS takes care of the imbalance for you. But, if you give us the research question you are trying to test and more details on your design you may get useful input. For example, do you know that Diagnosis is not used in the model structure of this GLIMMIX model outside of distinguishing among IDs across different diagnoses? Your model has a g-matrix with N*N dimensions. Diagnosis is not part of your model and, therefore, your balancing question is not relevant.