Random Effects model vs Mixed Effects model for an Ordered Logit Panel. Is there a test to decide which model is the most appropriete?

[Osmar Molina]

Hello fellow statalisters.

This is my first post ever in this forum. Also, sorry if I commit any grammar mistake, english is not my native language.

I’m thinking of doing a ordered logistic model for a panel data in Stata 15.1. Until now, I’ve only worked with linear regression panel data, always making the Breusch-pagan test or the hausman test or others (depending of the panel) to decide if I use a RE model or a FE model. However, since I’m doing an ordered logit panel, I don’t know exactly what are the tests for this regression and I see Stata 15 doesn’t have the option to make a fixed effects model for a oredered logistic regression, so it's only random effects or mixed effects.

Any ideas of what test should work for an ordered logistic panel regression?

[Carlo Lazzaro]

Osmar:
welcome to this forum.
When Stata does not include a given option, is for good theoretical reasons (e.g., the statistic is not sufficient; the estimator is biased).
Hence, -fe- is out of debate with -xtologit-, and you can only go -re-.
As a side note, many of us (like me) are non-English mother tongue.
However, after many years of continuous learning from Statalist, I cannot witness any occasion in which native English speakers considered my non-native English as a justification to discard my queries.

[Osmar Molina]

Thank you so much for your response Carlo!

I'm doing an RE model then.

Yes, I think I was just insecure about my english, but as you say and although I've not been so much here, the people I've seen is very kind and supportive in this forum. I'm going to be more confident around here from now on.