Skip to content

Instantly share code, notes, and snippets.

@sampsyo sampsyo/

Last active Oct 31, 2018
What would you like to do?
the Clopper-Pearson method for getting a confidence interval for an estimated a Bernoulli parameter
import scipy.stats
import math
import random
def clopper_pearson(x, n, alpha=0.05):
"""Estimate the confidence interval for a sampled Bernoulli random
`x` is the number of successes and `n` is the number trials (x <=
n). `alpha` is the confidence level (i.e., the true probability is
inside the confidence interval with probability 1-alpha). The
function returns a `(low, high)` pair of numbers indicating the
interval on the probability.
b = scipy.stats.beta.ppf
lo = b(alpha / 2, x, n - x + 1)
hi = b(1 - alpha / 2, x + 1, n - x)
return 0.0 if math.isnan(lo) else lo, 1.0 if math.isnan(hi) else hi
# As a test, estimate the probability of a fair coin (p=0.5) using 100 flips.
if __name__ == '__main__':
total = 100
successes = sum(random.randint(0, 1) for i in range(total))
lo, hi = clopper_pearson(successes, total)
print('95% confidence interval: {:.2f}-{:.2f}'.format(lo, hi))
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
You can’t perform that action at this time.