Multilevel Logistic Regression in GLIMMIX - Subject effect and number of levels

[Amanda92]

Hi,

I'm using SAS On Demand for Academics and new to multilevel modeling as well. I have a dataset with 8300 children who are nested in Mothers nested in household nested in clusters. But I have predictor variables only for child level, mother level and household level. I used PROC GLIMMIX to fit the empty model of the multilevel logistic regression model first with the following code.

PROC GLIMMIX data=FINAL.DATABASE NOCLPRINT;
class
Cluster_number Household_number Mother_Number;
model Living_Status=/dist=binomial link=logit;
random int/subject=Mother_Number*Household_number*Cluster_number;
COVTEST/ WALD;
RUN;

So is it ok to use Cluster_number in the Subject=Mother_Number*Household_number*Cluster_number even though there are no predictors to be added in the full model for cluster level?

Also below code does not give the covariance parameter values of Cluster_number,Household_number*Cluster_number,Mother_Number*Household_number*Cluster_number and residual covariance parameter together. Is there any modification needs to be done to the code to get those values? I need to calculate the ICC for each level.

PROC GLIMMIX data=FINAL.DATABASE NOCLPRINT;
class
Cluster_number Household_number Mother_Number;
model Living_Status=/dist=binomial link=logit;
random int/subject=Cluster_number;

random int/subject=Household_number*Cluster_number;

random int/subject=Mother_Number*Household_number*Cluster_number;
COVTEST/ WALD;
RUN;

[SteveDenham]

What does your output look like?  Are the estimates in the Covariance Parameters table all missing?  Are there warnings in the log, or included as part of the output?  More info will definitely help address your issue.

SteveDenham