Calibration Sample Count

Problem

Estimate how many calibration samples are needed to observe a rare case.

Return the smallest n with 1 - (1 - frequency)^n >= confidence, computed as ceil(log(1 - confidence) / log(1 - frequency)). Return 0 when frequency or confidence is not positive, and 1 when frequency >= 1.

Test cases

case 1

in: [0.95,0.01]

expected: 299

case 2

in: [0.99,0.001]

expected: 4603

case 3

in: [0.5,0.1]

expected: 7

case 4

in: [0.95,0]

expected: 0

Starter codePython

def calibration_samples(confidence, frequency):
    # Your code here
    pass

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