RMSE-LOO Metric — RmseLooMetric • bayesim

RMSE estimated via PSIS-LOO: the LOO-weighted posterior-mean prediction is computed with loo::E_loo() and compared to the observed response (sqrt(mean((y - yloo)^2))). This matches the construction underlying brms' loo_predict(type = "mean"). The max Pareto-k is returned as a diagnostic. Falls back to NA when the PSIS object or observed response is unavailable.

Constructor for RmseLooMetric.

Usage

RmseLooMetric(
  name = "rmse_loo",
  needs = c("loo", "epred"),
  required = FALSE,
  summary_type = "mean",
  schema = list(value = list(role = "estimate", aggregation = "mean", mcse = "sd"), elpd
    = list(role = "estimate", aggregation = "mean", mcse = "sd"), pareto_k_max =
    list(role = "diagnostic", aggregation = "mean", mcse = "sd"))
)

rmse_loo_metric(name = "rmse_loo")

Arguments

name: Character string naming the metric. Defaults to "rmse_loo".

Value

A RmseLooMetric object.

Examples

rmse_loo_metric()
#> <bayesim::RmseLooMetric>
#>  @ name        : chr "rmse_loo"
#>  @ needs       : chr [1:2] "loo" "epred"
#>  @ required    : logi FALSE
#>  @ summary_type: chr "mean"
#>  @ schema      :List of 3
#>  .. $ value       :List of 3
#>  ..  ..$ role       : chr "estimate"
#>  ..  ..$ aggregation: chr "mean"
#>  ..  ..$ mcse       : chr "sd"
#>  .. $ elpd        :List of 3
#>  ..  ..$ role       : chr "estimate"
#>  ..  ..$ aggregation: chr "mean"
#>  ..  ..$ mcse       : chr "sd"
#>  .. $ pareto_k_max:List of 3
#>  ..  ..$ role       : chr "diagnostic"
#>  ..  ..$ aggregation: chr "mean"
#>  ..  ..$ mcse       : chr "sd"