set_index with multindex on columns destroys column-multiindex

[adamklein]

Per Chris Billington on pystatsmodels

In [8]: df
Out[8]: 
          a         b         c  routine1            routine2       top          
                                  result1   result2   result1        OD          
                                                                     wx        wy
0 -0.914239 -0.230830 -0.421134  0.131751 -0.902347  0.382893  0.334615  0.909842
1 -0.533285 -0.878316 -3.046224 -0.760047  0.299502 -0.268321 -0.709366 -0.878868
2  0.804353 -1.203268 -1.221442 -1.948503  0.156240 -0.439375 -0.921254  1.119102

In [9]: df2 = df.set_index([('a','','')])

In [10]: df2
Out[10]: 
<class 'pandas.core.frame.DataFrame'>
Index: 3 entries, -0.914238702285 to 0.804353019197
Data columns:
('b', '', '')                  3  non-null values
('c', '', '')                  3  non-null values
('routine1', 'result1', '')    3  non-null values
('routine1', 'result2', '')    3  non-null values
('routine2', 'result1', '')    3  non-null values
('top', 'OD', 'wx')            3  non-null values
('top', 'OD', 'wy')            3  non-null values
dtypes: float64(7)

In [11]: df2.columns
Out[11]: 
Index([('b', '', ''), ('c', '', ''), ('routine1', 'result1', ''),
       ('routine1', 'result2', ''), ('routine2', 'result1', ''),
       ('top', 'OD', 'wx'), ('top', 'OD', 'wy')], dtype=object)

[adamklein]

Code to construct df:

arrays = [[  'a',   'b',   'c',    'top',    'top', 'routine1', 'routine1', 'routine2'],
        [   '',    '',    '',     'OD',     'OD', 'result1',   'result2', 'result1'],
        [   '',    '',    '',     'wx',     'wy',        '',          '',        '']]

tuples = zip(*arrays)
tuples.sort()
index = MultiIndex.from_tuples(tuples)

df = DataFrame(randn(3,8), columns=index)