Effectsize-centered prior in brms

Hey everyone,

I have a question regarding the specification of my prior for a mixed effects model in brms. I would like to use an effectsize-centered prior, which is the combination of two t-distributions with a non-zero non-centrality parameter corresponding to the expected effect size. The first distribution is centered around the positive value for the expected effect size, the second is centered around the negative value. This effect-size prior is explored in detail by Klauer et al. (2024), which created their own package emBayesFactor (which I am using for every test so far), for anyone interested in diving deeper. I’m attaching a visualization of the prior below. My question now is if I can specify this prior in brms.

Please also provide the following information in addition to your question:

  • Operating System: MAC OS 14.1 Sonoma
  • brms Version: 2.22.0

Thanks,

Max

I’m not aware of a simple way to do this with brms. Perhaps you could extend some of the custom-likelihood capabilities for your case (see this vignette).

1 Like

It looks like this is your first time posting. Welcome to the Stan forums, @MEck!

Hi Solomon,

thanks for the advice and the warm welcome. I’m trying to create a custom family following the vignette and will post an update for everyone interested in the subject once I succeed (or ultimately fail).

1 Like