Lprior?

  • Operating System: macOS
  • brms Version: 2.17.0

Since brms 2.17.0, several of the post-processing functions return draws from an lprior parameter, which seems new. Here’s a quick example:

library(brms)


fit <- brm(
  data = epilepsy,
  family = poisson,
  count ~ 1
)

posterior_summary(fit)
as_draws_df(fit)
                Estimate   Est.Error         Q2.5        Q97.5
b_Intercept     2.110184 0.022768989     2.065641     2.153492
lprior         -1.970318 0.003358909    -1.976841    -1.963891
lp__        -1644.344975 0.741538245 -1646.383320 -1643.840654
# A draws_df: 1000 iterations, 4 chains, and 3 variables
   b_Intercept lprior  lp__
1          2.1     -2 -1644
2          2.1     -2 -1644
3          2.1     -2 -1644
4          2.1     -2 -1644
5          2.1     -2 -1644
6          2.1     -2 -1644
7          2.1     -2 -1644
8          2.1     -2 -1644
9          2.1     -2 -1644
10         2.1     -2 -1644
# ... with 3990 more draws
# ... hidden reserved variables {'.chain', '.iteration', '.draw'}

I didn’t see any documentation on lprior in the updated brms reference manual and it wasn’t mentioned in the NEWS file. So my questions are:

  • What is the new lprior parameter?
  • What should applied researchers do with lprior?

Looks like support for automatic prior/likelihood sensitivity analysis using priorsense package GitHub - n-kall/priorsense: priorsense: an R package for prior diagnostics and sensitivity

1 Like

Thank you for the insights, @avehtari. It looks like I have another package to familiarize myself with.

1 Like