04-02-2012 10:28 AM
I wonder if survey procedures only apply to survey data with a sampling design frame? I read a paper which analyzed non-sampled data (patient claims) and the author used survey procedures (surveyreg and surveylogistic) to do the analysis with "cluster" effects (provider). I think the regular hierachical model (LME, GLE or GEE) should be applied to this situation, but I am not quite sure about this. Can anyone give me some suggestions?