Skip to contents

Bundles the priors for the intercept, the regression coefficients and the auxiliary parameter. Any component left as NULL uses rstanarm's default, weakly informative, autoscaled prior (see vignette("priors", package = "rstanarm")). To request an improper flat prior explicitly, use prior_spec("flat").

Usage

prior_config(intercept = NULL, slope = NULL, aux = NULL)

Arguments

intercept, slope, aux

Priors created with prior_spec(). Native rstanarm prior objects (e.g. rstanarm::normal(0, 1)) and the list format used by bmbeR 1.x (e.g. list(type = "normal", mu = 0, sigma = 1)) are also accepted.

Value

An object of class bmb_prior_config.

Details

The intercept prior applies to the intercept after centring the predictors (i.e. the expected outcome, on the link scale, at the predictor means). This is how rstanarm parameterises the model; you do not need to centre the predictors yourself.

Examples

prior_config(
  intercept = prior_spec("normal", location = 80, scale = 20),
  slope     = prior_spec("normal", location = 0, scale = c(1, 10))
)
#> <bmb_prior_config>
#>   intercept : normal(location = 80, scale = 20)
#>   slope     : normal(location = 0, scale = [1, 10])
#>   aux       : rstanarm default