Skip to contents

Recomputes the audit categorization across a grid of confidence thresholds so the researcher can see how many cases drop into "low_conf_agreement" or "uncertainty" at each tau.

Usage

tau_sens(
  annotations,
  thresholds = c(0.5, 0.55, 0.6, 0.7),
  low_threshold = 0.31
)

Arguments

annotations

As in audit.

thresholds

Numeric vector of confidence thresholds to scan (e.g. c(0.5, 0.55, 0.6, 0.7)).

low_threshold

Passed to audit.

Value

A long data frame: one row per (threshold, audit_status) with column n.