Value counts for all combinations of groups

Question

I have a dataframe with multiple group columns and a value column.

   a  b  val
0  A  C  1
1  A  D  1
2  A  D  1
3  A  D  2
4  B  E  0

For any one group, for eg a==A , b==C I can do value_counts on the series slice. How can I get the value counts of all possible combinations of the group columns in a dataframe format similar to:

  a  b val counts
0 A  C 1   1
1 A  D 1   2
2 A  D 2   1
2 B  E 0   1

Answer 1

is that what you want?

In [47]: df.groupby(['a','b','val']).size().reset_index()
Out[47]:
   a  b  val  0
0  A  C    1  1
1  A  D    1  2
2  A  D    2  1
3  B  E    0  1

or this?

In [43]: df['counts'] = df.groupby(['a','b'])['val'].transform('size')

In [44]: df
Out[44]:
   a  b  val  counts
0  A  C    1       1
1  A  D    1       3
2  A  D    1       3
3  A  D    2       3
4  B  E    0       1