Bootstrapped two-sample t-test in Python

bootstrap.py

from __future__ import division
import numpy as np
import pandas as pd
import random

def sample(data):
    sample = [random.choice(data) for _ in xrange(len(data))]
    return sample

def bootstrap_t_test(treatment, control, nboot = 1000, direction = "less"):
    ones = np.vstack((np.ones(len(treatment)),treatment))
    treatment = ones.conj().transpose()
    zeros = np.vstack((np.zeros(len(control)), control))
    control = zeros.conj().transpose()
    Z = np.vstack((treatment, control))
    tstat = np.mean(treatment[:,1])-np.mean(control[:,1])
    tboot = np.zeros(nboot)
    for i in xrange(nboot):
        sboot = sample(Z)
        sboot = pd.DataFrame(np.array(sboot), columns=['treat', 'vals'])
        tboot[i] = np.mean(sboot['vals'][sboot['treat'] == 1]) - np.mean(sboot['vals'][sboot['treat'] == 0]) - tstat
    if direction == "greater":
        pvalue = np.sum(tboot>=tstat-0)/nboot
    elif direction == "less":
        pvalue = np.sum(tboot<=tstat-0)/nboot
    else:
        print 'Enter a valid arg for direction'

    print 'The p-value is %f' % (pvalue)