Conformal Interval

Problem

Build a conformal prediction interval from calibration absolute residuals: take the k-th smallest residual with k = ceil((n + 1) * (1 - alpha)) clamped to [1, n], then return [prediction - q, prediction + q].

Empty residuals return [0.0, 0.0].

Hint

Conformal intervals carry a finite-sample coverage guarantee under exchangeability.

Test cases

case 1

in: [[1,2,3,4],0.25,10]

expected: [6,14]

case 2

in: [[1,2,3],0.5,0]

expected: [-2,2]

case 3

in: [[],0.1,5]

expected: [0,0]

Starter codePython

import math
def conformal_interval(residuals, alpha, prediction):
    # Returns [lower, upper]
    # Your code here
    pass

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