group_matrices: Group confusion matrices

In ModelOriented/FairModels: Flexible Tool for Bias Detection, Visualization, and Mitigation

Group confusion matrices

Description

Calculates confusion matrices for each subgroup

Usage

group_matrices(protected, probs, preds, cutoff)

Arguments

protected	vector containing protected variable
probs	character name of column with probabilities
preds	numeric, vector with predictions
cutoff	numeric cutoff for probabilities, default = 0.5

Value

group_matrices object It is a list with values:

For each subgroup:

• subgroup

  • tp - number of true positives

  • fp - number of false positives

  • tn - number of true negatives

  • fn - number of false negatives

Examples

data("compas")

glm_compas <- glm(Two_yr_Recidivism ~ ., data = compas, family = binomial(link = "logit"))
y_prob <- glm_compas$fitted.values

y_numeric <- as.numeric(compas$Two_yr_Recidivism) - 1

gm <- group_matrices(compas$Ethnicity,
  y_prob,
  y_numeric,
  cutoff = list(
    Asian = 0.45,
    African_American = 0.5,
    Other = 0.5,
    Hispanic = 0.5,
    Caucasian = 0.4,
    Native_American = 0.5
  )
)

gm

ModelOriented/FairModels documentation built on Aug. 30, 2022, 5:48 p.m.