holisticai.bias.metrics.zscore_diff

holisticai.bias.metrics.zscore_diff(group_a, group_b, y_pred, q=0)

ZScore Difference

This function computes the spread in Zscores between group_a and group_b. The Zscore is a normalised version of Disparate Impact.

If q is a vector, this function returns a vector with the respective result for each given quantile in q.

A value of 0 is desired. The Zscore will approximate the number of standard deviations away from the mean. In particular values that exceed 2 are statistically significant with 95% probability.

group_aarray-like

Group membership vector (binary)

group_barray-like

Group membership vector (binary)

y_predarray-like

Predictions vector (regression)

qfloat, array-like, optional

quantile of predictions considered, default=0.

float

ZScore Difference

:math:`

rac{ exttt{AVgroup_a} - exttt{AVgroup_b}}{ exttt{STD_pool}}`

>>> import numpy as np
>>> from holisticai.bias.metrics import zscore_diff
>>> group_a = np.array([1, 1, 1, 1, 0, 0, 0, 0, 0, 0])
>>> group_b = np.array([0, 0, 0, 0, 1, 1, 1, 1, 1, 1])
>>> y_pred = np.array([0.8, 0.9, 0.2, 0.1, 0.2, 0.9, 0.3, 0.6, 0.3, 0.5])
>>> zscore_diff(group_a, group_b, y_pred)
0.1166919931983158