I have a study where clustering occurs in one condition but not the other. In the treatment group I have repeated measurements on individuals who are nested within families. And in the control condition I only have repeated measurements on individuals, but no clustering due to family.
Some dummy data would look like this
time id family group y
1 0 1 1 treatment 0.58407458
2 1 1 1 treatment 0.57629394
3 2 1 1 treatment 1.16558208
4 0 2 1 treatment -1.03117769
5 1 2 1 treatment 0.87066744
6 2 2 1 treatment 0.42714038
7 0 3 1 treatment 0.62503878
8 1 3 1 treatment 0.11275242
9 2 3 1 treatment 0.66396118
10 0 4 2 treatment 0.07094150
11 1 4 2 treatment -0.44600018
.. .... . ....... .......
28 0 10 NA control -1.66283938
29 1 10 NA control 1.17574655
30 2 10 NA control -0.59692375
31 0 11 NA control -1.94929165
32 1 11 NA control 0.88162730
33 2 11 NA control -2.38991654
.. .... . ....... .......
If both conditions were nested within families I would run
lmer(y ~ time*group + (1 | family/id), data=study_data)
But I don't think this fits the right model in this scenario? What would be the correct way to fit a partially clustered design?
