using glimmix for multi-level modeling

#1
I have a question about how to conduct multi-level analysis in glimmix.

I have administered a peer nomination survey to a group of elementary students. The students were asked to indicate if their fellow classmates matched a series of descriptions (let's just label them, X, Y, and Z). Therefore, the outcomes are binary or dichotomous (yes the classmate matched the description or no the classmate did not match). There is not a way to group them into continuous variables--otherwise I would do a social relations model.

The children could provide unlimited nominations. There are 8 classes of children.

I would like to do an analysis to determine the odds of being nominated given the gender of the nominating student and the gender of the student being nominated. I would like to do this in glimmix or another procedure in SAS.

I understand the analysis conceptually but am having difficulty translating this into SAS code. The children are nested within classrooms. The data are arranged in a many-to-many way: every student can nominate every other student and every student can be nominated by all their classmates.

Does anyone have any suggestions about how to set up this analysis in GLIMMIX?

Thank you,

Charles
 

hlsmith

Omega Contributor
#2
Yeah, this feels a little awkward. You may have another layer if students can rate more than one other student. So students nested in classes and nominations clustered in students. Still seems a little off though.
 
#3
I think you misunderstood. The children were able to nominate as many students from their class as they wanted. Or they could choose not to nominate anyone at all.