Bootstrap采样方法的python实现

本文只考虑相对简单的一维的情况：

import numpy as np

def bootstrap(data, num_samples, statistics, alpha):
    n = len(data)
    idx = np.random.randint(0, n, size=(num_samples, n))
    samples = data[idx]
    stat = np.sort(statistics(samples, 1))
    return (stat[int((alpha/2)*num_samples)], stat[int((1-alpha/2)*num_samples)])