Set_prior to Dirichlet distribution in multilevel bernoulli model

[divas...@gmail.com]

Dear Paul,

Hello. Now I built up a two-level binary model family=bernoulli(link = "logit"). I used RMLE first to estimate all coefficients and found a set of predictors are significant (p<0.05). However, when I used brms, some predictors became insignificant. I just wander that maybe it is due to the multicollinearity among the predictors and a lot of them are categorical variables. Therefore, I want to set priors of these variables using Dirichlet distribution. However, everytime I tried to define it in brms, it gave an error. And I can not find any document to tell how to define Dirichlet prior in brms. Could you please give me an idea? Many thanks.

Diva

[Paul Buerkner]

It is not possible in brms to set a dirichlet prior on the regression coefficients. How would such a model look like in your opinion?

--
You received this message because you are subscribed to the Google Groups "brms-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to brms-users+unsubscribe@googlegroups.com.
To post to this group, send email to brms-...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/brms-users/1f89afbc-a1a2-4a27-9815-0b41f7fdf0dd%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.