Simple cross-tabulation in pandas

Question

I stumbled across pandas and it looks ideal for simple calculations that I\'d like to do. I have a SAS background and was thinking it\'d replace proc freq -- it looks like it\'l

Answer 1

v0.21 answer

Use pivot_table with the index parameter:

df.pivot_table(index='category', aggfunc=[len, sum])

           len   sum
         value value
category            
AB           2   300
AC           1   150
AD           1   500

---

<= v0.12

It is possible to do this using pivot_table for those interested:

In [8]: df
Out[8]: 
  category  value
0       AB    100
1       AB    200
2       AC    150
3       AD    500

In [9]: df.pivot_table(rows='category', aggfunc=[len, np.sum])
Out[9]: 
            len    sum
          value  value
category              
AB            2    300
AC            1    150
AD            1    500

Note that the result's columns are hierarchically indexed. If you had multiple data columns, you would get a result like this:

In [12]: df
Out[12]: 
  category  value  value2
0       AB    100       5
1       AB    200       5
2       AC    150       5
3       AD    500       5

In [13]: df.pivot_table(rows='category', aggfunc=[len, np.sum])
Out[13]: 
            len            sum        
          value  value2  value  value2
category                              
AB            2       2    300      10
AC            1       1    150       5
AD            1       1    500       5

The main reason to use __builtin__.sum vs. np.sum is that you get NA-handling from the latter. Probably could intercept the Python built-in, will make a note about that now.