Question

I would like to modify DataFrame entries for some rows at a specific level of a hierarchical index. Here is a canonical example:

>>> index = pd.MultiIndex.from_arrays([['a','a', 'b', 'b'], [1,2,1,2]], 
...                                   names=['first', 'second'])
>>> data = pd.DataFrame(np.random.rand(len(index)), index=index, columns=['A'])
>>> print data
                     A
first second          
a     1       0.587781
      2       0.560407
b     1       0.492996
      2       0.267799

I would like to set rows for which second==2 to 0 (for example). I tried using DataFrame.xs method but it returns a copy and not a view:

>>> selected = data.xs(2, level='second')
>>> print selected
              A
first          
a      0.560407
b      0.267799

>>> selected['A']=0
>>> print data

                    A
first second          
a     1       0.587781
      2       0.560407
b     1       0.492996
      2       0.267799

The last assignment did not affect data (it changed values in selected of course).

Answer 1

In recent version of pandas help(data.xs) shows how to get a view on the data.

It can be done by using xs arg copy=False.