http://endlesspint.com/2016-12-23-clean-food-please/

0_stat_signif_shuffle

sample Python code of running a shuffling and sampling of two populations

1_eg_shuff_samp.py

import numpy as np
import pandas as pd

import matplotlib.pyplot as plt
import matplotlib
matplotlib.style.use('ggplot')
%matplotlib inline


### scores

craft_mean = np.mean(frsq_311.groupby(['dba', 'street', 'inspection_date'])['score'].mean())
print craft_mean
# Out:
# 12.890173410404625

other_mean = np.mean(manHealth[~manHealth.camis.isin(frsq_311.camis)].groupby(['dba', 'street', 'inspection_date'])['score'].mean())
print other_mean
# Out:
# 14.67555658952399

score_diff = other_mean - craft_mean
score_diff
# Out:
# 1.785383179119366

print len(scores_craft)
scores_all = np.append(scores_craft, scores_other)
print scores_all
# Out:
# 177


### shuffle, sample, diff

mean_diff_list = []

for i in range(10000):

    np.random.shuffle(scores_all)
    x = scores_all[177:]
    y = scores_all[:177]
    
    mean_diff_list.append(x.mean() - y.mean())
    
print np.sum(np.array(mean_diff_list) > score_diff)/float(len(mean_diff_list))
# Out:
# 0.0082


### plot

plt.figure(figsize=(10,5))
plt.hist(mean_diff_list, bins=50)
plt.plot([score_diff, score_diff], [0, 700])

plt.title("Ten Thousand Shuffle & Compare Simulations")
plt.xlabel("Difference in Score Means")

plt.text(-3.5, 10.0, 'endlesspint.com',
               fontsize=12, color='gray',
               ha='left', va='bottom', alpha=0.3)

### save file localy w high(er) resolution
# plt.savefig('img/diff_hist.png', dpi=300)