Question

鉴于下面的示例数据，我试图计算条件概率。列表示在序列（S1，S2 ......）中发生的事件A-E。

前两个示例按预期工作并计算P（S2 | S1）和P（S2，S3 | S1）。当条件包括多个列时，该方法会中断，如示例三所示，我预期会计算P（S3 | S1，S2）。

我很欣赏有关为什么这不起作用以及哪种替代方法可以获得P（S3 | S1，S2）的期望结果的见解。例如，我希望输出包含行A,D,B,0.25和A,D,C,0.75。

谢谢！

MWE代码：

import pandas as pd

data = { 'S1' : ['A','A','A','B','B','A','A'],
         'S2' : ['B','D','D','A','D','D','D'],
         'S3' : ['C','C','C','D','C','B','C'],
         'S4' : ['D','B','E','C','A','C','E'] }

df = pd.DataFrame(data)

print (df)
print ((df.groupby(['S1','S2']).agg({'S4':'count'}) /
        df.groupby('S1').agg({'S4':'count'})).rename(columns={'S4':'Freq'}))
print ((df.groupby(['S1','S2','S3']).agg({'S4':'count'}) / 
        df.groupby('S1').agg({'S4':'count'})).rename(columns={'S4':'Freq'}))
print ((df.groupby(['S1','S2','S3']).agg({'S4':'count'}) / 
       df.groupby(['S1','S2']).agg({'S4':'count'})).rename(columns={'S4':'Freq'}))

输出：

  S1 S2 S3 S4
0  A  B  C  D
1  A  D  C  B
2  A  D  C  E
3  B  A  D  C
4  B  D  C  A
5  A  D  B  C
6  A  D  C  E
       Freq
S1 S2      
A  B    0.2
   D    0.8
B  A    0.5
   D    0.5
          Freq
S1 S2 S3      
A  B  C    0.2
   D  B    0.2
      C    0.6
B  A  D    0.5
   D  C    0.5
Traceback (most recent call last):
  File "test.py", line 13, in <module>
    print ((df.groupby(['S1','S2','S3']).agg({'S4':'count'}) / df.groupby(['S1','S2']).agg({'S4':'count'})).rename(columns={'S4':'Freq'}))
NotImplementedError: merging with more than one level overlap on a multi-index is not implemented

Pandas groupby分裂错误

0 个答案: