MultiIndex-based indexing in pandas [duplicate]

Answer 1

xs may be what you want. Here are a few examples:

In [63]: df.xs(('B',), level='Alpha')
Out[63]:
                  I        II       III        IV         V        VI       VII
Int Bool                                                                       
0   True  -0.430563  0.139969 -0.356883 -0.574463 -0.107693 -1.030063  0.271250
    False  0.334960 -0.640764 -0.515756 -0.327806 -0.006574  0.183520  1.397951
1   True  -0.450375  1.237018  0.398290  0.246182 -0.237919  1.372239 -0.805403
    False -0.064493  0.967132 -0.674451  0.666691 -0.350378  1.721682 -0.791897
2   True   0.143154 -0.061543 -1.157361  0.864847 -0.379616 -0.762626  0.645582
    False -3.253589  0.729562 -0.839622 -1.088309  0.039522  0.980831 -0.113494

In [64]: df.xs(('B', False), level=('Alpha', 'Bool'))
Out[64]:
            I        II       III        IV         V        VI       VII
Int                                                                      
0    0.334960 -0.640764 -0.515756 -0.327806 -0.006574  0.183520  1.397951
1   -0.064493  0.967132 -0.674451  0.666691 -0.350378  1.721682 -0.791897
2   -3.253589  0.729562 -0.839622 -1.088309  0.039522  0.980831 -0.113494 

Edit:

For the last requirement you can chain get_level_values and isin:

Get the even values in the index (other ways to do this too)

In [87]: ix_vals = set(i for _, i, _ in df.index if i % 2 == 0)
         ix_vals

Out[87]: set([0L, 2L])

Use these with isin

In [89]: ix = df.index.get_level_values('Int').isin(ix_vals)
In [90]: df[ix]
Out[90]:                I        II       III        IV         V        VI       VII
Alpha Int Bool                                                                       
A     0   True  -1.315409  1.203800  0.330372 -0.295718 -0.679039  1.402114  0.778572
          False  0.008189 -0.104372  0.419110  0.302978 -0.880262 -1.037645 -0.264265
      2   True  -2.414290  0.896990  0.986167 -0.527074  0.550753 -0.302920  0.228165
          False  1.275831  0.448089 -0.635874 -0.733855 -0.747774 -1.108976  0.151474
B     0   True  -0.430563  0.139969 -0.356883 -0.574463 -0.107693 -1.030063  0.271250
          False  0.334960 -0.640764 -0.515756 -0.327806 -0.006574  0.183520  1.397951
      2   True   0.143154 -0.061543 -1.157361  0.864847 -0.379616 -0.762626  0.645582
          False -3.253589  0.729562 -0.839622 -1.088309  0.039522  0.980831 -0.113494 

Answer 2

You can use pd.IndexSlice for an intuitive way (Inspired from this answer). Some examples (using pandas 0.18.0):

df.sort_index(inplace=True)
idx = pd.IndexSlice
evens = np.arange(2,max(df.index.levels[1])+1,2)

df.loc[idx[['A','B'],evens,True],['III','V']]
Out[]: 
                     III         V
Alpha Int Bool                    
A     2   True -1.041243 -0.561155
B     2   True  0.381918 -0.148990

df.loc[idx[:,evens,:],:]
Out[]: 
                        I        II       III        IV         V        VI  \
Alpha Int Bool                                                                
A     2   False  0.791142  0.333383  0.089767 -0.584465  0.295676 -1.323792   
          True  -1.023160 -0.442004 -1.041243  1.613184 -0.561155  0.397923   
B     2   False  0.383229 -0.052715 -0.214347 -2.041429 -1.101059 -0.374035   
          True  -0.183386 -0.855367  0.381918 -0.261106 -0.148990  0.621537   

                      VII  
Alpha Int Bool             
A     2   False  0.717301  
          True  -0.133701  
B     2   False  0.166314  
          True   0.517513