Dataframe access second level of MultiIndex

Question

I have been trying to figure out the best way to deal with a multiIndex. Especially when I want to access some values on the second level of the multiIndex. For example:

df = pd.DataFrame([np.random.randint(3,size=20),np.random.rand(20)]).T
df.columns = ['A','B']
g = df.groupby('A').describe()

Let's say I'm trying to look at the mean values of this output. I can do something like this:

idx = [True if a[1]=='mean' else False for a in g.index.tolist()]
p.loc[idx,:]

It works but there must be a better way to do this. Is there a better way to access the second level multiIndex?

Answer 1

I found a couple of easy answers:

g.xs('mean', level=1)

Another one:

idx = pd.IndexSlice
g.loc[idx[:,'mean'],:]

Answer 2

You should read over the documentation on multi-index data frames. IndexSlice is the way to handle this. Something like this should work.

import pandas as pd
idx = pd.IndexSlice
g.loc[idx[:,"mean"],:]

Answer 3

You can swap the order of indexes in the multiindex:

g.reorder_levels([1,0]).loc['mean']
#            B
#A            
#0.0  0.515745
#1.0  0.451534
#2.0  0.483014

Answer 4

You can also do this

g.loc[:, ("B", "mean")]

  A
0.0    0.381882
1.0    0.450356
2.0    0.497692
Name: (B, mean), dtype: float64

Check out advanced indexing with hierarchical index