Multiple membership multilevel modelling

[Jean-Sebastien Bournival]

Hi,
I'm looking to assess the influence of kin on living socioeconomic context of Ego. I use dyadic data (separated pairwise datasets like parent-child, siblings, etc.) and follow the dyads over 7 consecutive censuses. My outcome variable is dichotomous (whether or not two individuals live in the same socioeconomic context for a specific year).

I'm trying to refine this simple multilevel model :

melogit y x1 x2 x3 || dyad:

I'm not sure this is correct since individuals can be part of multiple dyads (e.g. cousins, siblings, uncles, etc.) and for what I understand the model would treat each dyad as being independant as if they were formed by different people (like in couple analyses). I was wondering how to add that specification to tell the model that individuals may be part of multiple dyads. The next question would probably be related to weighing options. It is not rare to see people with more than a hundred cousins, intuitively I feel the need to weigh this kind of overrepresentation.

Thanks for your advices.
JS

[Carlo Lazzaro]

Jean-Sebastian:
is there a way to impose a clustered standard errors on multiple dyads?

[Jean-Sebastien Bournival]

Mr Lazzaro,
I'll look into that. It seems possible for linear regression though. I'm not as familiar as I would like to be with statistics, but from what I just read online you may be right. However, could a dyad be part of more than one cluster?
Thanks

[Carlo Lazzaro]

Jean-Sebastian:
this is the methodologcal issue I ignore, because I',m not familiar with the literature of your research field.
I figured something like the -group- function available from -egen- to create higher-level groups from dyads on which to cluster the standard errors.
As an aside, please call me Carlo, just like all on (and many more off) this list do. Thanks.