New group level in RStanarm output

Interfaces rstanarm
1

I was looking at the example_model in rstanarm and a bit unclear about some of the output.

library(rstanarm)
example(example_model)

tail(example_model$stan_summary[, 1:2])

which gives the following output

#>                                              mean     se_mean
#> b[(Intercept) herd:14]                 1.02671072 0.016200257
#> b[(Intercept) herd:15]                -0.62123581 0.019984998
#> b[(Intercept) herd:_NEW_herd]          0.00726711 0.020258006
#> Sigma[herd:(Intercept),(Intercept)]    0.65040705 0.021683219
#> mean_PPD                               1.77508929 0.006152789
#> log-posterior                       -114.76141580 0.268875928

Is b[(Intercept) herd:_NEW_herd] here the mean that each group level intercept is drawn from? Or is it a posterior predictive distribution for a new hypothetical group?

I couldn’t find this documented, but would appreciate a reference!

2

It’s a parameter for a hypothetical new group that is used under the hood when posterior_predict is called asking for a new group. I didn’t actually realize that it’s included it in the output. I think originally we were intending to keep it hidden and just use it internally as needed.

3

Got it, thanks!