I have a panel dataset (covering 3 periods), in which we've stacked the data so that each individual respondent has 4-8 observations. The independent and dependent variables are both on a Likert scale. My general idea is: dependent = indep1 + gend + indep1*gend, with fixed effects.

I found this paper that seems to do a similar thing (ordinal logistic regression with fixed effects, feologit in Stata). I'm basically just wondering whether this still works for stacked data?

My dataset looks like this:

Resp  depen Indep1  gend
 1       1      1   f
 1       2      2   m
 1       4      3   f
 2       2      4   f
 2       4      1   m
 3       3      2   f

A paper did a similar thing to what I intend to do and used alternative specific multinomial probit - but I have the Likert scale outcome, more time periods and I'd rather not estimate separate models for the gender (gend).


1 Answer 1


This is probably a place for a proportional odds ordinal logistic model with random effects (random intercepts) for subjects. See for example the R mixor package or the R rmsb package.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.