Population Stability Index

Problem

Compute the population stability index between expected and actual bin counts: normalize each list, clip each bin share to at least 1e-4, and sum (actual - expected) * ln(actual / expected).

Return 0.0 if either total is 0.

Hint

PSI is symmetric-ish divergence that grows when the actual distribution drifts from expected.

Test cases

case 1

in: [[20,30,50],[10,30,60]]

expected: 0.08754687373538998

case 2

in: [[50,50],[0,100]]

expected: 4.60431846666895

case 3

in: [[10,10],[10,10]]

expected: 0

Starter codePython

import math
def population_stability_index(expected_counts, actual_counts):
    # Your code here
    pass

3 test cases · run your code against them in the editor