Feature request: zero and one inflated proportional models

[ericward-noaa]

Thanks for all the work on brms Paul -- it's an outstanding package.

I know brms can handle zero-inflated proportions. Some collaborators and I recently developed a model for responses that includes proportional data, as well as 0s and 1s. The package is built using Stan: zoid and source

At a recent conference, a few colleagues wondered about porting it over to brms to take advantage of smooths, random effects, etc. If there's enough interest, happy to brainstorm on this more, or work on a PR etc. Thanks!

[paul-buerkner]

Have you checked out the zero_one_inflated_beta distribution of brms?

[ericward-noaa]

Apologies, I wasn't clear -- our model is a zero and one inflated Dirichlet. I had looked at the zero_one_inflated_beta in brms, and it looks great for univariate cases.

[paul-buerkner]

I see. Yes, adding such a family could be interesting indeed. How shall we proceed with the discussion?

[paul-buerkner]

I will close this issue for now to clean up the issue tracker a bit.