Summarize predictions — summarize_predictions • powerup

Collect predictions across POWERUP models. Models without prediction outputs, including models skipped during training, are omitted.

Usage

summarize_predictions(
  models,
  format = c("wide", "long"),
  response_cutoff = NULL
)

Arguments

models: A list of model objects generated by fit_powerup_models and with predictions added using add_powerup_predictions.
format: Output format. "wide" returns the sample-by-perturbation prediction matrix. "long" returns one row per sample and perturbation with prediction uncertainty and target-event metadata for posterior updating.
response_cutoff: Optional finite response cutoff used to calculate target-event probabilities when format = "long". If NULL, each model's stored cutoff is used.

Value

If format = "wide", a data frame with samples in rows and perturbations in columns. If format = "long", a tibble with one row per sample-perturbation pair containing prediction distributions, uncertainty intervals, response cutoffs, Gaussian target-event probabilities, and transformed target-event probabilities.

Examples

if (FALSE) { # \dontrun{
predictions <- summarize_predictions(my_models)
prediction_distributions <- summarize_predictions(my_models, format = "long")
} # }