Pandas value_counts 组总和取决于 groupby.agg 中的另一列 function

Question

I am currently aggregating a set of variables grouped and aggregated by variables var1 and var2 . Let's assume I do have continuous variables var3, var4, var5 where I calculate the mean, std, min, max and median values within aggregated groups easily. However, I also do have two more variables var6 and var7 where var6 is a categorical variable and continuous var7 variable that shows the size of var6 . My dataframe looks as follows:

var1 var2 var3 var4 var5 var6 var7
1    a    3    8     9   0    125
1    a    4    0     12  0    12 
1    a    12   4     12  2    3
1    b    24   5     1   1    45
2    a    1    19    4   1    76
2    a    2    37    12  1    12
2    c    3    93    156 1    341
2    c    57   1     87  2    73
2    c    42   4     95  2    95
3    b    12   11    0   0    11
3    b    119  0     901 0    5

The first part, where I calculate grouped aggregation is easy as follows (as an example):

desired_df=my_df.groupby(['var1', 'var2']).agg(
    max_var3=('var3', 'max')
    mean_var4=(var4, 'min'))

What I desire to do on top of this aggregation is to group var6 and var7 and put them as new columns next to the aggregation. Below is what I like to do:

var1 var2 var6_group0_sum var6_group1_sum var6_group2_sum
1    a        137         0               3
1    b        0           0               45
2    a        0           88              0
2    c        0           341             168
3    b        16           0               0

How can I achieve this within grouped aggregation? Any help appreciated.

Answer 1

Get the aggregation without the categorical columns (copied and pasted your code):

left = (df.groupby(['var1', 'var2'])
         .agg(max_var3=('var3', 'max'),
              mean_var4=('var4', 'min'))
        )

Get the aggregation of just the categorical column:

right = (df.groupby(['var1', 'var2', 'var6'])
           .var7
           .sum()
           .unstack(-1, fill_value = 0)
           .rename(columns = lambda col: f'var6_group{col}_sum')
           .rename_axis(columns = None)
          )

Combine the two dataframes, note that they share the same index:

pd.concat([left, right], axis = 1)

           max_var3  mean_var4  var6_group0_sum  var6_group1_sum  var6_group2_sum
var1 var2                                                                        
1    a           12          0              137                0                3
     b           24          5                0               45                0
2    a            2         19                0               88                0
     c           57          1                0              341              168
3    b          119          0               16                0                0