python pandas 按不同轴上的总和和平均值分组

Question

I need to group my data and calculate mean on one axis and sum on another. I've been looking for similar questions but I can't find a proper solution.

I have a similar df:

df = pd.DataFrame ({'A': ['XX','XX','XX','XX','XX','XX','XX','XX','XX',
                          'YY','YY','YY','YY','YY','YY','YY','YY','YY',
                          'ZZ','ZZ','ZZ','ZZ','ZZ','ZZ','ZZ','ZZ','ZZ'],
                    
                    'B': ['ind1','ind2','ind3','ind1','ind2','ind3','ind1','ind2','ind3',
                          'ind1','ind2','ind3','ind1','ind2','ind3','ind1','ind2','ind3',
                          'ind1','ind2','ind3','ind1','ind2','ind3','ind1','ind2','ind3'],   
                                        
                    'C': ['2017','2017','2017','2018','2018','2018','2019','2019','2019',
                          '2017','2017','2017','2018','2018','2018','2019','2019','2019',
                          '2017','2017','2017','2018','2018','2018','2019','2019','2019'],
                    
                    'D': np.random.randint(0,100,size=27)})

I need the following df:

A   ind1    ind2    ind3    TOTAL
XX  52.33   73.00   37.00   162.33
YY  40.67   51.33   54.33   146.33
ZZ  84.00   28.67   62.00   174.67

Where columns ind1, ind2, ind3 are means by axis =0, while TOTAL is the sum of ind1,ind2,ind3 by axis = 1

I tried the following but is not working:

print(df.groupby('A')['D'].agg(['sum','mean']))

Any help would be fantastic.

Answer 1

I believe you need pivoting by crosstab or DataFrame.pivot_table and then add new column with sums by DataFrame.assign :

np.random.seed(20)
    
df = pd.DataFrame ({'A': ['XX','XX','XX','XX','XX','XX','XX','XX','XX',
                          'YY','YY','YY','YY','YY','YY','YY','YY','YY',
                          'ZZ','ZZ','ZZ','ZZ','ZZ','ZZ','ZZ','ZZ','ZZ'],
                    
                    'B': ['ind1','ind2','ind3','ind1','ind2','ind3','ind1','ind2','ind3',
                          'ind1','ind2','ind3','ind1','ind2','ind3','ind1','ind2','ind3',
                          'ind1','ind2','ind3','ind1','ind2','ind3','ind1','ind2','ind3'],   
                                        
                    'C': ['2017','2017','2017','2018','2018','2018','2019','2019','2019',
                          '2017','2017','2017','2018','2018','2018','2019','2019','2019',
                          '2017','2017','2017','2018','2018','2018','2019','2019','2019'],
                    
                    'D': np.random.randint(0,100,size=27)})

df = (pd.crosstab(df['A'], df['B'], df['D'], aggfunc='mean')
        .assign(Total = lambda x: x.sum(axis=1)))

print (df)
B        ind1       ind2       ind3       Total
A                                              
XX  67.666667  46.000000  60.000000  173.666667
YY  69.333333  45.666667  67.333333  182.333333
ZZ  16.333333  57.666667  32.333333  106.333333

Or:

df = (df.pivot_table(index='A',columns='B',values='D')
        .assign(Total = lambda x: x.sum(axis=1)))

Answer 2

This is another method if you are not familiar with cross_tab or pivot table.

df_n = df.groupby(['A','B'])['D'].mean().unstack()
df_n['Total'] = df_n.sum(axis=1)

Output would be:

B        ind1       ind2       ind3       Total
A                                              
XX  67.666667  46.000000  60.000000  173.666667
YY  69.333333  45.666667  67.333333  182.333333
ZZ  16.333333  57.666667  32.333333  106.333333