Modeling paired data with GLMM?

[ksb39]

I am trying to test for differences in the ages at when two different behaviors emerge in nestlings birds. Age distributions of the two behaviors are reasonably normally distributed.  We have the ages of both behaviors recorded for each of 53 individuals, so I could use a paired t-test, but need to control for non-independence of sibling groups. The 53 nestlings come from 21 different nests, thus some nestlings were not independent (siblings had the same parents, habitat, etc.,).  Here is my code:

Proc GLIMMIX data=...;

CLASS nest ID behavior;

MODEL age = behavior / dist=poisson link=log solution;

RANDOM  nest;

RANDOM residual / type=vc subject=ID;

run;quit;

Does this seem appropriate?  Thanks!

[SteveDenham · Accepted solution]

This should model what you are looking for:

Proc GLIMMIX data=...;

CLASS nest ID behavior;

MODEL age = behavior / dist=poisson link=log solution;

RANDOM  nest;

RANDOM behavior / residual type=vc subject=ID;

run;quit;

I hope this helps.

Steve Denham