Build context for metrics computation — build_metric

Precomputes predictions, log_lik, loo based on metric needs. Only computes what is needed by the metrics, and only if the fitter supports it.

Usage

build_metric_context(fit_result, fitter, data_bundle, metrics, seed = NULL)

Arguments

fit_result: A bayesim_fit_result object from a successful fit
fitter: S7 Fitter object
data_bundle: A validated data bundle list
metrics: List of S7 Metric objects

Value

A named list containing any of:

predictions: Prediction results from the fitter
log_lik: Pointwise log-likelihood matrix
loo: LOO-CV results

Details

The function inspects the needs property of each metric to determine what context elements are required. It then checks if the fitter supports each capability and computes them. Any errors during computation result in NULL values for that context element.