search

paul-buerkner / brms

brms R package for Bayesian generalized multivariate non-linear multilevel models using Stan
https://paul-buerkner.github.io/brms/
GNU General Public License v2.0
1.28k stars 184 forks source link

Adding ordered beta regression to brms #1191

Open saudiwin opened 3 years ago

saudiwin commented 3 years ago

Hi -

I've been working on a new model for [0,1] responses. It's been posted to the Stan discourse and has gotten some usage in the wider community. In comparison to existing methods like zero-one inflation for Beta regressions, it implements a much simpler parameterization that is essentially a Beta regression plus two ordered cutpoints. This allows for a single linear model to be used instead of multiple models 0, 1 and (0,1). A working paper describing the approach is available here:

https://osf.io/preprints/socarxiv/2sx6y/

I already also implemented it as a custom family. I works fairly well, although the user has to add 0 + Intercept to the formula to get the cutpoints to work right. More info is available in this vignette:

https://htmlpreview.github.io/?https://github.com/saudiwin/ordbetareg/blob/master/estimate_with_brms.html

Anyhow, would love to see this get in to brms proper.

Thanks much,

Bob

paul-buerkner commented 3 years ago

Thanks! I will take a closer look at the provided material but from a quick look this sounds like a super nice addition to brms.

saudiwin commented 3 years ago

Thanks!

sjwild commented 1 year ago

I want to follow up on this. I think I have an idea for how this could be implemented as a supported family in brms. I'm happy to start the work behind it if there is still interest given there is also the {ordbetareg} package.

paul-buerkner commented 1 year ago

I think this would be a nice addition to brms and appreciate people working on this feature.