Bias-Corrected Cramer's V

Problem

Return the bias-corrected Cramer's V: sqrt(max(0, chi2/n - (r-1)(c-1)/(n-1)) / min(r-1, c-1)) for a 2D table, where chi2 uses expected counts row_total*col_total/n. Return 0.0 for tables with fewer than two rows or columns, a total of at most one, or any zero expected count.

Hint

The correction subtracts the association expected by chance.

Test cases

case 1

in: [[[10,20],[30,40]]]

expected: 0

case 2

in: [[[10,0],[0,10]]]

expected: 0.9733285267845753

case 3

in: [[[50,10],[10,50]]]

expected: 0.6603340693163625

case 4

in: [[[5,5],[5,5]]]

expected: 0

case 5

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

expected: 0.3871174808473405

Starter codePython

def bias_corrected_cramers_v(observed):
    # Your code here
    pass

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