[R] NLME / LMER: Multiple Membership Models

[Rense Nieuwenhuis]

Dear all,

I'm investigating on how to estimate a specific kind of cross- 
classified multilevel model. I think it is often referred to as a  
multiple-membership model.

To problem is this: I want to study changing attitudes of people, for  
which I use (amongst other things) contextual data of the  
neighborhood they live in. I have two surveys (panel-design) and I  
know the neighborhood people lived in during both surveys.

Until this point, it can be estimated using simple multilevel  
modeling. The problem is that many people moved to other  
neighborhoods (I still have all their information). `Basic' cross- 
classified models (with contexts of both neighborhoods) do not  
suffice, for two reasons:
	- not all respondents moved, and I want to model them all  
simultaneously
	- not all respondents moved at the same time, so weight-factors need  
to be applied to contexts

Does any of you have ideas on how to estimate such models? Preferably  
using the NLME / LMER packages. In the literature these kinds of  
models are referred to as multiple membership models.

Thanks in advance,

Rense Nieuwenhuis