nested logit (nlogit) and nest-specific covariates?

[Irene Prix]

Hello forum,

This may be a bit of a stretch, but I was wondering if it is at all possible to fit a nested logit model (nlogit) in Stata with nest-specific variables, i.e., variables which apply only to individuals in particular nests and are thus not available for all bottom-level alteratives/individuals?

Any hints are greatly appreciated.

[Charlie Joyez]

Welcome to the list Irene,

I'm afraid that you cannot run a such nested logit, either with nlogit or nlogitrum.
I don't know your topic and your variables but would that make sense to set this variable equal to zero for the set of alternatives in the nests where this variable isn't specified?

This is unlikely, but could happen. If it doesn't make sense, do not set these values to zero.

Anyway, a nested logit is made to compare the relative influence of the same determinants over several case of alternatives, and you can't compare on different basis (not the same explanatory variables.) So I doubt the nested logit model is adapted to what you want to do.

Best,
Charlie

[Irene Prix]

Many thanks for your input, Charlie!

Perhaps it works the way you suggested, setting the variable to zero for the nonapplicable alternative -- I've just noticed that this appears to be the way Yves Croissant deals with the matter in an example for nested logit in the vignette of his mlogit-package in R. I'll have to think about it and see whether it makes sense for what I'm trying to do.

In any case, thank you!