www.cmm.bristol.ac.uk/forum

Hello,

I'm wondering if there is a benchmark figure regarding the appropriateness of running a Multiple Membership model.

I'm analysing a dataset of patients nested within hospitals and around 2% of patients are "members" of multiple hospitals - does this figure of 2% justify running a multiple membership model?

Thanks
Valerie

Hi Valerie,

Good question, I am not aware of any studies relating the percentage of multiple membership to parameter biases.
Clearly, the more mobility, the less appropriate a naive hierarchical model and the more appropriate a multiple membership model.
However, with only 2% of patients belonging to 2 hospitals I think that the naive hierarchical model will be an adequate approximation in your study.
Put another way, the naive 2-level model is appropriate for 98% of your sample, which sounds fine!

Best wishes

George

Hi George,

Many thanks for your response - this is very useful!

Best regards,
Valerie