Machine-readable diagnostics for a calibration
Source:R/calibrate.R
diagnostics.agent_calibration.RdOne-row diagnostic summary of an agent_calibrate() result: the method and
estimand, the gold and total sample sizes, the naive bias (the plug-in
estimate minus the corrected estimate, for the first/headline term), the
corrected interval width, and the labeled-set accuracy and Krippendorff alpha.
A naive_bias far from zero is the signal that plugging predicted labels in
directly would have been misleading.
Usage
# S3 method for class 'agent_calibration'
diagnostics(x, ...)Value
A one-row tibble with method, estimand, n_labeled, n_total,
naive_bias, ci_width, accuracy, alpha.