带有条件的百分比列

Question

I have this df:

data = {'A':[102, 102, 102, 102, 312, 312, 312], 
        'B':[1001,1001,1001,1001,1001,1001,1001],
        'C':[3005,3005,3005,3005,3005,3005,3005],
        'D':[2004,2004,2004,2004,2002,2002,2002],
        'E':[1,3,5,999,1,5,999],
        'F':[300,1,192,837,19,1,1037]} 

df = pd.DataFrame (data, columns = ['A','B','C','D','E','F'])

df.head(7)

一行代码计算了一个百分比，该百分比不同于我希望它排除E列中的行值为（999）的计数值：

df['Percentage'] = 100 * df['F'] / df.groupby('A')['F'].transform('sum')

百分比应显示：

Percentage
60.85193
0.20284
38.94523
(Blank)
95
5
(Blank)

任何帮助将不胜感激

Answer 1

应用蒙版并将选定的行设置为无。按照您提供的示例执行所有操作，然后添加以下行：

mask = data['E'] == '999'
df['Percentage'][mask] = np.nan

Answer 2

您可以细分框架，然后将transform细分为该特定部分，然后将结果重新分配回去：

>>> grp = df[df['E'].ne(999)]
>>> grp['F'].mul(100).div(grp.groupby('A')['F'].transform('sum'))
0    60.851927
1     0.202840
2    38.945233
4    95.000000
5     5.000000
Name: F, dtype: float64
>>> df['Percentage'] = grp['F'].mul(100).div(grp.groupby('A')['F'].transform('sum'))
>>> df
     A     B     C     D    E     F  Percentage
0  102  1001  3005  2004    1   300   60.851927
1  102  1001  3005  2004    3     1    0.202840
2  102  1001  3005  2004    5   192   38.945233
3  102  1001  3005  2004  999   837         NaN
4  312  1001  3005  2002    1    19   95.000000
5  312  1001  3005  2002    5     1    5.000000
6  312  1001  3005  2002  999  1037         NaN