Durbin-Watson

Problem

Compute the Durbin-Watson statistic: sum((e_t - e_{t-1})^2) / sum(e_t^2).

Return 0.0 with fewer than two residuals or a zero denominator.

Hint

Values near 2 indicate no autocorrelation; near 0 suggests positive autocorrelation.

case 1

in: [[1,-1,1,-1]]

expected: 3

case 2

in: [[1,2,3]]

expected: 0.14285714285714285

case 3

in: [[0,0,0]]

expected: 0

case 4

in: [[2,1,2,1]]

expected: 0.3

Starter codePython

def durbin_watson(residuals):
    # Your code here
    pass

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