Ordinal regression models with random effects

[Michael Ratajczak]

Hi,

Is there any way to include random effects in ordinal regression models fitted using rstanarm? stan_polr doesn't seem to have this capability -- is this something that may be added in the future?

Thanks,

Michael

[Bob Carpenter]

I'm just calling out RStanArm in the title. I don't know
all the models it can fit, but Ben and Jonah will.

- Bob

> --
> You received this message because you are subscribed to the Google Groups "Stan users mailing list" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to stan-users+...@googlegroups.com.
> To post to this group, send email to stan-...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.

[Paul Buerkner]

As far as I know, rstanarm does currently not support random effects in ordinal models. but the brms package, which is also based on Stan, offers this feature.

[Andrew Gelman]

I thought rstanarm did allow ordered logit with varying intercepts and slopes using lmer4-style notation.  But maybe Ben and Jonah haven't put this in yet, I'm not sure.

A

[Michael Ratajczak]

Hi Paul and Andrew,

Thank you for your replies.

Paul: I am aware of the brms package, but I prefer rstanarm because it is faster. Also, I have many predictors and I would like to fit the model using the R-squared prior.

Andrew: I think stan_polr is the only modelling function in rstanarm that models an ordinal response; when I try to vary the intercepts for each participant using lme4-style notation, the following error message appears:

Error: Constant variable(s) found: 1 | Subject

Michael

[Michael Ratajczak]

I am wondering, are there plans for incorporating random effects in the stan_polr function in the future? Or for creating an additional function allowing for random effects in ordinal regression models?

[Andrew Gelman]

There is definitely a plan to do this.  It might be there already, actually.

A