McNemar test麥克尼馬爾檢驗

 python金融風控評分卡模型和數據分析微專業課（博主親自錄制視頻）：http://dwz.date/b9vv

模型項目統計聯系QQ:231469242

McNemar’s Test用於配對的2*2表格，例如如果你要比較兩個醫生對同一個病人治療效果。一個病人自控，葯物治療前和治療后結果比較。

McNemar’s Test用於檢驗(a+b)/N 和（a+c/N）是否顯著？

McNemar’s Test的兩個分類變量不是獨立的，而是相關的，因為(a+b)/N 和（a+c/N）都包含a

McNemar’s Test This is a matched pair test for 2  *   2 tables. For example,
if you want to see if two doctors obtain comparable results
when checking (the same) patients, you would use this test.

 

 

 

 

 

http://www.doc88.com/p-7337013497692.html

 

 

 

 

 

 

 

 

 python代碼測試一致

 

# -*- coding: utf-8 -*-
'''
QQ:231469242,by Toby
'''
# Import standard packages
import numpy as np
import scipy.stats as stats
import pandas as pd

# additional packages
from statsmodels.sandbox.stats.runs import cochrans_q, mcnemar
#葯物測試，對疾病是否治愈，要求相同對象群體
#obs = np.array([[101, 121],[59, 33]])
obs = np.array([[2, 4],[0, 4]])
def Mcnemar(obs):
    '''McNemars Test should be run in the "exact" version, even though approximate formulas are
    typically given in the lecture scripts. Just ignore the statistic that is returned, because
    it is different for the two options.
    
    In the following example, a researcher attempts to determine if a drug has an effect on a
    particular disease. Counts of individuals are given in the table, with the diagnosis
    (disease: present or absent) before treatment given in the rows, and the diagnosis
    after treatment in the columns. The test requires the same subjects to be included in
    the before-and-after measurements (matched pairs).
    '''
    (statistic, pVal) = mcnemar(obs)
    print('\nMCNEMAR\'S TEST -----------------------------------------------------')
    print('p = {0:5.3f}'.format(pVal))
    if pVal < 0.05:
        print("There was a significant change")  
    else:
       print("There was no significant change")


Mcnemar(obs)      
   
'''
p = 0.125
There was no significant change
'''   

 

 

 

 

例題2

 

python結果無顯著差異

 

 

例題3

 

 

 

 

 

練習

 

 

 python機器學習生物信息學系列課（博主錄制）：http://dwz.date/b9vw