Run the grant-recipient match — pf_match

Splits the prepared grants into in-scope and out-of-sample (see pf_oos_details()), runs the cascade on the in-scope side only, and reports the elapsed time of the match itself.

Usage

pf_match_grants(
  grants,
  ref,
  config = npmatch::np_config(),
  method = "hier",
  passes = pf_default_passes(),
  recover = FALSE,
  verbose = TRUE,
  check_reference = c("auto", "stop", "warn", "ignore")
)

Arguments

grants: A prepared frame from pf_prepare_grants().
ref: A reference bundle from pf_build_reference(), or a normalized np_reference.
config: An npmatch::np_config().
method: Scoring method passed to npmatch::np_score().
passes: Blocking passes; defaults to pf_default_passes().
recover: Add the stage-4 acronym and prefix blocking keys to both sides before matching. Required by pf_recovery_passes(); harmless otherwise, but it costs a pass over both frames, so it is FALSE by default.
verbose: Print the per-stage cascade report.

Value

A pf_match object: the np_tiered result plus the in-scope / out-of-scope split, the per-stage table, and timings.

Details

The timing covers the cascade only — blocking, comparison, scoring, veto, selection, tiering. Reading and normalizing the BMF, and preparing the grant extract, are excluded (do them once via pf_build_reference() / pf_prepare_grants()).