panel data ordered logit model and fixed effects

[Lisa Space]

Dear community,

I have a small follow up question to a recent problem of mine where I cannot find an answer to.
The main problem is that I have panel data (I used xtset), an ordered categorical variable (which is why I want to use the ordered logit model),I did perform the hausmann test (which suggested fixed effects), plus I have heteroscedasticity (which is why I wanted to use the robust command).

However, I cannot perform the xtologit command with the fe command.

stata says:
option fe not allowed

Any suggestions how to perform the ordered logit regression for panel data with using fixed effects?

Best

Lisa

[Phil Bromiley]

As you can see, the documentation for xtologit says it does random effects ordered logit. Generally, if Stata doesn't provide an option you should suspect that option doesn't make sense. You need to read up on panel ordinal logits. I suspect you'll find that the fixed effects model is not identified. This is the case for xttobit. Sometimes folks use dummy variables in place of fixed effects, but these often do not result in consistent estimates.

[Pat Carpet]

Lisa Space I have a similar problem now. Did you figure out a way to do fixed effects on an ordered logit model in Stata?

Edit: Apparently, the user-written ado 'feologit' might be a solution.