ROCAUC Difference

Overview

Measures the difference in ROC AUC between privileged and unprivileged groups. It quantifies the absolute disparity in discrimination performance across demographic categories.

Usage

Manually

Python

# Calculate difference in rocauc score

rocauc_diff = roc_auc_difference(df, protected_attribute, privileged_group, labels, positive_label, y_true)
print(rocauc_diff)

Using Fairness Object

result = fo.compute(roc_auc_difference)

Results

0.35416666666666663

These results are obtained by using the input data given in the Create Example Data page under Getting Started

Interpretation

A result > 0 indicates that the privileged group has a better ROC AUC score than the unprivileged group while a result < 0 indicates that the unprivileged group has a better ROC AUC score. A score around 0 is desirable