Counterfactual Prediction Residual

Problem

Compute the residual between the observed outcome and the counterfactual prediction under the observed treatment: y_obs - (y1_pred if treatment == 1 else y0_pred).

Hint

Only the potential outcome matching the treatment actually received is observed.

Test cases

case 1

in: [5,3,6,1]

expected: -1

case 2

in: [2,3,6,0]

expected: -1

case 3

in: [4,4,7,1]

expected: -3

case 4

in: [0,0,0,0]

expected: 0

Starter codePython

def counterfactual_residual(y_obs, y0_pred, y1_pred, treatment):
    # Your code here
    pass

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