Confounding Bias Magnitude

Problem

Return the bias from an omitted confounder: slope * covariate_gap, where slope = correlation * sd_outcome / sd_covariate is the outcome-on-confounder regression coefficient and covariate_gap is the treated-minus-control mean of the confounder. Return 0.0 when sd_covariate is zero.

Hint

Convert the correlation to a slope first.

Test cases

case 1

in: [0.5,2,4,1.5]

expected: 1.5

case 2

in: [0,2,4,3]

expected: 0

case 3

in: [-0.4,5,2,2.5]

expected: -0.4

case 4

in: [0.3,0,2,1]

expected: 0

Starter codePython

def confounding_bias(correlation, sd_covariate, sd_outcome, covariate_gap):
    # Your code here
    pass

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