Cluster SE and Fixed Effects using mlogit

[juliana pinto]

Dear Statalists,

I do not have stata17 so I cannot use xtmlogit. The femlogit does not allow cluster SE - cluster(id) and does not allow to use i.variables in the regression.
Can I use mlogit and use i.id i.time to represent the fixed effects? Does this give me the same results as xtmlogit?

Many thanks!

[Carlo Lazzaro]

Juliana:
the following thread can be helpful: https://www.statalist.org/forums/for...or-panel-model

[juliana pinto]

Carlo Lazzaro Many thanks for answering.

So, does this mean that if I don't have stata17 and xtmlogit, I cannot run a multinomial logit model for panel data with standard errors clustered by id?

The alternative would be femlogit, but its help document says that " robust uses the robust or sandwich estimator of variance. This is valid
only for quasi-maximum-likelihood interpretation (Wooldridge 2010,
502ff.). It can be interpreted only as heteroskedasticity
robustness, not as panel robustness."

In the link you sent, there is a suggestion of transforming the dep. var. categories into just 2, but this is not good for my data. My dep. var. are : dropout, pass, fail, missing. Is there another stata 16 alternative to run the model with cluster(id)?

Many thanks!!