Group-R2 prior for high dimensional regression

[jear2412]

Hi Paul,

Would it be possible to add the Group-R2 prior to the family of shrinkage priors currently supported in brms?

This prior extends the existing R2-based framework by incorporating known group structures among covariates. The total R2 is decomposed in two stages:

1. Across groups: The user specifies which covariates belong to each group, and the group-level variance decomposition follows a Dirichlet distribution.
2. Within groups: Each group has its own Dirichlet prior to partition the group-specific explained variance among its covariates.

The main reference is here: https://arxiv.org/abs/2507.11833

Let me know if you need anything from my side. Thanks!

Best,
Javier

[paul-buerkner]

That would indeed be great! I have to think about how to extend the interface for the R2D2 prior for this, but I think at least the interface part should be straightword with just a single argument addition.