Multiple membership or cross classified model?

[FloHarrison]

Hi,

I'm constructing a model looking at children's physical activity. Children were measured on three occasions, in Years 5, 6 and 9. So for the first two measurement periods they were at primary schools, but had moved to secondary schools for measurement 3. I am modelling my outcome on a day-by-day basis, so have days nested within children nested within schools. On each day a child is only ever at one school.

Should I use a multiple membership model, which would allow for some weighting to be placed on primary school for measurements taken at the secondary school, or should I try a cross classified model, which I think would treat primary schools and secondary schools as separate units.

Thanks,

Flo Harrison

[joneskel]

Interesting question! There is some debate about the choice of models in the literature; I would recommend looking at

Alastair H Leyland and Øyvind Næss (2009) The effect of area of residence over the life course on subsequent mortality, J R Stat Soc Ser A Stat Soc. 2009 June; 172(3): 555–578.
doi: 10.1111/j.1467-985X.2008.00581.x

Leyland AH, Næss Ø. (200&) Multilevel modelling of the longitudinal influence of neighbourhoods on health [e-letter]. Journal of Epidemiology & Community Health Online 2007:18 January open access