Stabilized Weights

Problem

Compute stabilized inverse propensity weights. Let p1 be the treated fraction; treated units get p1/e and controls get (1-p1)/(1-e), with e clipped to [1e-6, 1 - 1e-6].

Return the list of weights; empty input returns [].

Hint

Multiplying by the marginal treatment probability keeps weights near 1.

Test cases

case 1

in: [[1,0],[0.8,0.4]]

expected: [0.625,0.8333333333333334]

case 2

in: [[1,1,0,0],[0.5,0.5,0.5,0.5]]

expected: [1,1,1,1]

case 3

in: [[],[]]

expected: []

Starter codePython

def stabilized_weights(treatment, propensity):
    # Your code here
    pass

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