df = DataFrame(data = {'col1':[1,2,3,4],'col2':[1,2,3,4]}, index =MultiIndex.from_tuples([('a','b'),('a','c'),('a','b'),('a','d')]))
df.sort_index(inplace=True)
df
        col1   col2
a   b    1      1
    b    3      3
    c    2      2
    d    4      4

df.ix['a','b']

        col1  col2
a   b    1     1
    b    3     3

df.reset_index(inplace=True)
df.set_index(['level_0','level_1','col1'])

col2
level_0 level_1 col1
  a        b     1      1
                 3      3
           c     2      2
           d     4      4

df.ix['a','b',1,'col2'] = 3
IndexingError: only tuples of length <= 2 supported

In [12]: df.ix[('a','b'),'col2'] = 3

In [13]: df
Out[13]:
     col1  col2
a b     1     3
  b     3     3
  c     2     2
  d     4     4

[4 rows x 2 columns]

df.ix['a','b']

In [19]: df.reset_index()
Out[19]:
  level_0 level_1  col1  col2
0       a       b     1     1
1       a       c     2     2
2       a       b     3     3
3       a       d     4     4

[4 rows x 4 columns]

In [20]: df.reset_index().set_index(['level_0','level_1','col1'])
Out[20]:
                      col2
level_0 level_1 col1
a       b       1        1
        c       2        2
        b       3        3
        d       4        4

[4 rows x 1 columns]

In [21]: df2 = df.reset_index().set_index(['level_0','level_1','col1'])

In [22]: df2.ix[('a','b',1)] = 10

In [23]: df2
Out[23]:
                      col2
level_0 level_1 col1
a       b       1       10
        c       2        2
        b       3        3
        d       4        4

[4 rows x 1 columns]