大熊猫：多列求和

Question

考虑这个数据帧：

STUDENT T_1 T_2 T_3 T_4 
0 A   PASS FAIL PASS FAIL 
1 B   PASS FAIL FAIL FAIL 
2 C   FAIL FAIL PASS PASS 
3 D   PASS FAIL PASS PASS 

列T_1 - > T_4代表测试。在这种情况下，T_1和T_3是'X'类型的测试，而T_2和T_4是'Y'类型的测试。这些列是分类值。我想得到每个测试类型的％分布（即X/Y）。所以，我想这一点：

STATUS X    Y 
0 PASS 0.75 (6/8) 0.25 (2/8) 
1 FAIL 0.25 (2/8) 0.75 (6/8) 

我知道我可以在一系列使用s.value_counts（）/ s.count（）来获取每列％状态分布，但如何聚集在多个列（即，结合T_1/T_3，T_2/T_4，因为我知道它们属于特定的测试类型）

Answer 1

这里有一种方法可以做到这一点。

import pandas as pd 
import numpy as np 

# just try to simulate your data 
student_id = np.array('A B C D E F G H I G'.split()).reshape(10, 1) 
test_results = np.random.choice(['PASS', 'FAIL'], size=(10, 4), p=[0.7, 0.3]) 
data = np.concatenate([student_id, test_results], axis=1) 
df = pd.DataFrame(data, columns=['STUDENT', 'T_1', 'T_2', 'T_3', 'T_4']) 

# set index as student names 
df.set_index('STUDENT', inplace=True) 
# add multi-level index to columns 
df.columns = pd.MultiIndex.from_tuples([('T_1', 'X'), ('T_2', 'Y'), ('T_3', 'X'), ('T_4', 'Y')]) 
# transpose the df, groupby X,Y 
by = df.T.groupby(level=1) 


def count_func(group): 
    num_pass = (group.values == 'PASS').sum() 
    num_fail = (group.values == 'FAIL').sum() 
    pass_rate = '{:>3.2f}% ({}/{})'.format(num_pass/(num_pass + num_fail), num_pass, num_pass + num_fail) 
    fail_rate = '{:>3.2f}% ({}/{})'.format(num_fail/(num_pass + num_fail), num_fail, num_pass + num_fail) 

    return pd.Series({'PASS RATE': pass_rate, 'FAIL_RATE': fail_rate}) 


result = by.apply(count_func) 

Out[5]: 
     FAIL_RATE  PASS RATE 
X 0.25% (5/20) 0.75% (15/20) 
Y 0.25% (5/20) 0.75% (15/20)