-nlogit- has no cluster option, i.e. does not allow observations to be
correlated, put differently: assumes i.i.d. observations. However,
sometimes, this is too strong an assumption.
In the restaurant example in the Stata manual, restaurant choices of
families in the same neighborhood may be correlated because families in
the same neighborhood have similar (unobserved) tastes.
Is there some way to allow for non i.i.d. observations when estimating
nested logits? Is it possible to do this in -gllamm- ?
