Covariance Parameter Estimates GENMOD?

JonKetchup — Wed, 11 May 2022 15:28:44 GMT

Is there a way to output the covariance parameter estimates in proc genmod in a similar way to the covtest statment in glimmix? I know I can output the working correlation structure. I prefer the method of moments method over likelihood based techniques for GEEs, hence why I am trying to use genmod.

PROC GLIMMIX data = DATA 
class CLUSTER OUTCOME VAR1 VAR2;
model OUTCOME (event='1')= VAR1 VAR2
			/ dist=bin link=logit ;
RANDOM _residual_ / SUBJECT=CLUSTER TYPE=cs 
covtest '1'/ wald;
RUN;

PROC GENMOD data = DATA 
class CLUSTER OUTCOME VAR1 VAR2;
model OUTCOME (event='1')= VAR1 VAR2
			/ dist=bin link=logit;
REPEATED SUBJECT=CLUSTER /  TYPE=cs corrw ;
RUN;

Re: Covariance Parameter Estimates GENMOD?

StatDave — Wed, 11 May 2022 16:43:24 GMT

No. GENMOD does not fit a random effects model so there are no random effects to estimate covariances for. As you note, GENMOD fits a Generalized Estimating Equations model and instead estimates a correlation matrix among the repeated measures. GEE is not a likelihood-based estimation method. The estimation method is outlined in the Details: Generalized Estimating Equations section of the GENMOD documentation.

topic Covariance Parameter Estimates GENMOD? in Statistical Procedures

Covariance Parameter Estimates GENMOD?

Re: Covariance Parameter Estimates GENMOD?