civic_calibrate: Assess probability calibration (binary classification)

In civic.icarm: Interpretable Civic-Accountable and Responsible Machine Learning

Assess probability calibration (binary classification)

Description

Evaluates whether predicted probabilities are well-calibrated: a model predicting 0.7 should be correct ~70 Returns Brier score and Expected Calibration Error (ECE).

Usage

civic_calibrate(object, data, outcome, positive = NULL, n_bins = 10L)

Arguments

object	A 'civic_model' (binary classification only).
data	A data frame for evaluation.
outcome	Character. Outcome column name.
positive	Character. Positive class level.
n_bins	Integer. Number of probability bins (default '10').

Value

An object of class 'civic_calibration' (a list) with: 'bins' (tibble), 'brier_score', 'ece', 'positive', 'outcome', 'model'.

Examples

m   <- civic_fit(voted ~ age + education, civic_voting)
cal <- civic_calibrate(m, civic_voting, "voted", "yes")
print(cal)
civic_plot_calibration(cal)

civic.icarm documentation built on June 18, 2026, 1:06 a.m.