topic Re: Error code in SAS when choosing unstructured working correlation within Proc Genmod in Statistical Procedures

Error code in SAS when choosing unstructured working correlation within Proc Genmod

Mi1122 — Tue, 25 Oct 2022 15:41:41 GMT

Hi everyone, I'm getting an error code (no specific one, just following note: Stopped by error!) when trying to run following Proc Genmod procedure with an unstructured working correlation:

PROC Genmod DATA= table descending;
Class ID Year (ref='1');
Model TreatmentType=Year/ Dist=Binomial Link=Logit;
Repeated Subject=VID/ Type=UN corrw;

The working correlation matrix in the output looks like this (looks like it stopped running):

Col1 Col2 Col3 Col4 Col5 Col6
Row1 2.9295 -1.0000 0.0000 0.0000 0.0000 0.0000
Row2 -1.0000 0.3414 0.0000 0.0000 0.0000 0.0000
Row3 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Row4 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Row5 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000
Row6 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000

However, when I change the working correlation to exchangeable the procedure will run.

How does this come?

Thanks in advance!

Re: Error code in SAS when choosing unstructured working correlation within Proc Genmod

StatDave — Sat, 29 Oct 2022 14:19:07 GMT

The unstructured matrix requires the estimation of most correlation parameters - 15 in your case - as opposed to much simpler structures, like exchangeable, which only requires one to be estimated. As with any iterative model fitting algorithm, increasing the number of parameters to estimate, either model parameters or correlation parameters, increases the likelihood of fitting problems. This is particularly true for binary response models where data sparseness can make these problems more likely. That said, the more modern PROC GEE is the recommended procedure to fit GEE models rather than the older PROC GENMOD. It is possible, though by no means certain, that you'll have better luck there.

Re: Error code in SAS when choosing unstructured working correlation within Proc Genmod

lvm — Wed, 02 Nov 2022 21:08:41 GMT

I use MIXED and GLIMMIX much more than I do GENMOD. But.... I think you should make VID a class variable (add it to the CLASS statement). Also, if you have more than one observation for each combination of year and VID, you will have a problem.