Pandas: for all set of duplicate entries in a part

问题:

I have a large Dataframe that looks similar to this:

     ID_Code    Status1    Status2
0      A          Done       Not
1      A          Done       Done
2      B          Not        Not
3      B          Not        Done
4      C          Not        Not
5      C          Not        Not
6      C          Done       Done

What I want to do is calculate is for each of the set of duplicate ID codes, find out the percentage of Not-Not entries are present. (i.e. [# of Not-Not/# of total entries] * 100)

I'm struggling to do so using groupby and can't seem to get the right syntax to perform this.

回答1:

I may have misunderstood the question, but you appear to be referring to when values of Status1 and Status2 are both Not, correct? If that's the case, you can do something like:

df.groupby('ID_Code').apply(lambda x: (x[['Status1','Status2']] == 'Not').all(1).sum()/len(x)*100)

ID_Code
A     0.000000
B    50.000000
C    66.666667
dtype: float64

回答2:

IIUC using crosstab

pd.crosstab(df['ID_Code'],(df['Status1'].eq('Not'))&(df['Status2'].eq('Not')),normalize ='index')
Out[713]: 
col_0       False     True 
ID_Code                    
A        1.000000  0.000000
B        0.500000  0.500000
C        0.333333  0.666667



#pd.crosstab(df['ID_Code'],(df['Status1'].eq('Not'))&(df['Status2'].eq('Not')),normalize ='index')[True]

回答3:

Using sum and a boolean mask:

df.filter(like='Status').eq('Not').all(1).groupby(df.ID_Code).mean().mul(100)

ID_Code
A     0.000000
B    50.000000
C    66.666667
Name: flag, dtype: float64