difference between logit and xtlogit

[Marzieh Goodarzi]

hi

What is the difference between logit and xtlogit codes in Stata? In terms of assumptions applies to these two models in stata.
For example, do different assumptions apply to error term in logit and xtlogit?

Any help on the topic is appreciated.

[Joro Kolev]

You should read a textbooks such as Wooldridge, J.M., 2010. Econometric analysis of cross section and panel data. MIT press; or Maddala, G.S., 1986. Limited-dependent and qualitative variables in econometrics (No. 3). Cambridge university press.

A forum cannot replace systematic study of statistics/econometrics.

Having said this, and as far as I remember from the times when I actually read the above textbooks' chapters on the topic:

1. -xtlogit- controls for entity level fixed effect, while -logit- does not. E.g., if you -xtset person id time- , then -xtlogit- will control for person specific fixed effects.

2. The possibility of controlling for fixed effects in logit is a perculiaritty coming from the logit error assumption. For example there does not exists such a thing as -xtprobit- because the joint normality does not allow for elimination of the fixed effect, unlike the joint logit distribution.

3 The joint logit distribution comes at a cost. I cant recall exactly what was the cost, but a manifestation of this cost is that the multinomial logit model is a lot less general than the multinomial probit, that is, the former works only if a particular assumption holds. (I think the assumption was called "independence of the irrelevant option", or something similar.)