Using Generalized Estimated Equations (GEE) to analyze correlations?

[rbrui]

I have repeated measures data (meaning 3 measures per type of animal for one treatment only; there are 2 types in total) that are unbalanced. There are 2 treatments. Additionally, these animal 'types' are nested within group, which are nested within a cage.

Now, because of the difficulty of this data I thought to use GEE (which can process the unbalanced data and cope with the nested design) to analyse for potential correlation of certain behaviours in my data per treatment. I reckon that this is possible to do in GEE, however, I am not sure whether this is 'allowed'/ a valid way?

Many thanks for any comment(s)!

 

[hlsmith]

Probably feasible, but I am not familiar with it. I know another approach is application of multilevel modeling for longitudinal data. This is a pretty standard approach for dealing with within and between cluster variance/covariance. Though, I don't get much longitudinal data, so I am not much help beyond the recommendation.