PROC GLIMMIX estimate of the random effect

Silvana — Fri, 16 Mar 2018 12:13:31 GMT

I'm running a 3- level multilevel regression model using proc glimmix and one of the classes is HPSU so when i run the following code the estimate for HPSU random effect is 9.13E-11 and the standard error is zero.

Is there something i'm missing?

Here is the code I'm using:

PROC GLIMMIX DATA=DIRT METHOD=LAPLACE ;
CLASS HPSU HNUMBER;
MODEL HEPBRESF (EVENT=LAST) = / CL DIST=BINARY LINK=LOGIT ALPHA=0.1 SOLUTION ODDSRATIO (DIFF=FIRST
LABEL) DDFM=BW ;
WEIGHT HPWGT ;
RANDOM INTERCEPT / SUBJECT=HPSU TYPE=VC SOLUTION CL ;
RANDOM INTERCEPT / SUBJECT=HNUMBER (HPSU) TYPE=VC SOLUTION CL ;
COVTEST / WALD;
RUN;

Re: PROC GLIMMIX estimate of the random effect

Consuelo — Wed, 21 Mar 2018 20:10:04 GMT

Run the model with HPSU as fixed-effect and check the F-value for HPSU,

Some good advice found in

Tips and Strategies for Mixed Modeling with SAS/STAT® Procedures

http://support.sas.com/resources/papers/proceedings12/332-2012.pdf

and

Usage Note 40724: Comparing covariance structures, testing covariance parameters using the COVTEST statement in PROC GLIMMIX

http://support.sas.com/kb/40/724.html

topic PROC GLIMMIX estimate of the random effect in Statistical Procedures

PROC GLIMMIX estimate of the random effect

Re: PROC GLIMMIX estimate of the random effect