Descriptive score-by-group disparity

Usage

morie_fairness_predpol_score_disparity(score, group, reference = NULL)

Arguments

score: Continuous risk score per individual.
group: Protected attribute per individual.
reference: Optional reference group label (default: lowest-mean).

Value

morie_fairness_result; $value is the spread (max - min) of per-group mean scores.

Examples

set.seed(7); n <- 80L
group <- rep(c("X", "Y"), each = n / 2L)
score <- ifelse(group == "X", rnorm(n, 0, 1), rnorm(n, 2, 1))
morie_fairness_predpol_score_disparity(score, group)
#> Predictive-Policing Score Disparity (descriptive)
#> =================================================
#>   Group-mean spread  1.8528
#>   ANOVA F            72.42
#>   ANOVA p-value      9.6468e-13
#>   Reference group    X
#> 
#> Group mean risk scores span 1.85 points (reference 'X'). A one-way ANOVA finds the between-group difference statistically significant (F = 72.42, p = 0.0000). Note: a score gap is not itself evidence of bias — it can reflect genuine base-rate differences. Pair this with morie_fairness_predpol_calibration_audit, which compares the score against realised outcomes.