Dear all,
My apologies if this basic question was previously asked, I could not find it in the forum.
We are performing a Poisson regression in which the dataset contains a separate line for each person each week. In each such line any number of variables could have different values (including the exposure, the covariates and of course the outcome). So a person could appear in the dataset multiple times, though could only experience the outcome once.
Does this model need a random effect per person for the variance estimates to be valid?
On the one hand, these are repeated measurements, so it seems like it would.
On the other hand, it is widely known that survival models (e.g. Cox) do not need random effects when using time varying covariates, so maybe it’s similar here?
Would be thankful for any assistance,
Noam Barda