Yates Sample Ratio Mismatch

Problem

Compute the Yates-corrected sample ratio mismatch chi-square. With expected_i = total * ratio_i, add max(0, abs(observed_i - expected_i) - 0.5)^2 / expected_i for positive expected counts. Return 0.0 when the observed total is 0.

Hint

A large mismatch between observed and planned traffic share suggests the randomization is broken.

Test cases

case 1

in: [[45,55],[0.5,0.5]]

expected: 0.81

case 2

in: [[50,50],[0.5,0.5]]

expected: 0

case 3

in: [[30,70],[0.5,0.5]]

expected: 15.21

case 4

in: [[],[]]

expected: 0

case 5

in: [[10],[1]]

expected: 0

Starter codePython

def yates_sample_ratio_mismatch(observed, expected_ratios):
    # Your code here
    pass

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