9.9 Using default rstanarm priors

Authors recommend using the default prior from rstanarm:

bike_model_default <- rstanarm::stan_glm(
  rides ~ temp_feel, data = bikes, 
  family = gaussian,
  # here very specific prior on sd
  prior_intercept = normal(5000, 2.5, autoscale = TRUE), # <- see autoscale arg.
  prior = normal(0, 2.5, autoscale = TRUE), 
  prior_aux = exponential(1, autoscale = TRUE),
  chains = 4, iter = 5000*2, seed = 84735)

prior_summary(bike_model_default)
Priors for model 'bike_model_default' 
------
Intercept (after predictors centered)
  Specified prior:
    ~ normal(location = 5000, scale = 2.5)
  Adjusted prior:
    ~ normal(location = 5000, scale = 3937)

Coefficients
  Specified prior:
    ~ normal(location = 0, scale = 2.5)
  Adjusted prior:
    ~ normal(location = 0, scale = 351)

Auxiliary (sigma)
  Specified prior:
    ~ exponential(rate = 1)
  Adjusted prior:
    ~ exponential(rate = 0.00064)
------
See help('prior_summary.stanreg') for more details

It uses weakly informative priors using the scale of the data.

Figure 9.1: Fig 9.5 and 9.14 from Bayes Rules!