Bootstrap CI for Eval Accuracy

Problem

Seed random with seed, draw n_boot bootstrap resamples of the 0/1 outcomes with replacement, and return the percentile interval [low, high] at alpha/2 and 1 - alpha/2 using linear interpolation between closest ranks. Return [0.0, 0.0] when outcomes is empty, n_boot <= 0, or alpha lies outside (0, 1).

Hint

Resample indices with int(random.random() * n) using the seeded global RNG.

Test cases

case 1

in: [[1,1,0,1,0,1,1,0],2000,42,0.05]

expected: [0.25,1]

case 2

in: [[1,0],500,7,0.1]

expected: [0,1]

case 3

in: [[],100,1,0.05]

expected: [0,0]

case 4

in: [[1,1,1],100,3,0.2]

expected: [1,1]

Starter codePython

import random
def bootstrap_accuracy_ci(outcomes, n_boot, seed, alpha):
    # Your code here
    pass

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