pandas: Selecting index and then columns on multiindex slice

Question

I have a pandas dataframe with a multiindex of 2 levels, and four data columns, and I want to select, first some indices of both levels, and then certain columns. However, I do not find a way to do it. Let me illustrate with an example:

>>> import numpy as np
>>> import pandas as pd
>>> a = pd.MultiIndex.from_arrays(np.array([np.repeat([1, 2], 6), np.tile(np.repeat(["a", "b", "c"], 2), 2)] ) )
>>> df = pd.DataFrame(np.random.randn(12,4), index=a, columns=list('ABCD'))
>>> df
            A         B         C         D
1 a  0.688481 -2.380630  1.586150  0.008736
  a  0.481300  1.276254 -0.159889  0.568995
  b -0.250297 -0.161532 -1.099416  0.362548
  b -0.588712  0.403857 -0.573104  1.420252
  c -1.026832 -1.297500 -0.084957  0.072306
  c -0.262771  0.865048 -0.278121 -1.406952
2 a -0.873142 -0.871211 -1.790412  1.073330
  a -2.252715  1.772763  1.784698  1.716706
  b  1.108562 -0.622361  0.036408  1.170411
  b -0.157282 -0.660653 -1.230480 -0.295144
  c -0.241862 -0.471965  0.363945  0.125543
  c -0.396109  0.245829  0.544339 -0.957016

So I am able to do this:

>>> df.loc["1","a"]  
            A         B         C         D
1 a  0.688481 -2.380630  1.586150  0.008736
  a  0.481300  1.276254 -0.159889  0.568995

And this:

>>> df.loc[:,["A","B"]]
            A         B
1 a  0.688481 -2.380630
  a  0.481300  1.276254
  b -0.250297 -0.161532
  b -0.588712  0.403857
  c -1.026832 -1.297500
  c -0.262771  0.865048
2 a -0.873142 -0.871211
  a -2.252715  1.772763
  b  1.108562 -0.622361
  b -0.157282 -0.660653
  c -0.241862 -0.471965
  c -0.396109  0.245829

And even this:

>>> df.loc["1","a"]["A"]
1  a    0.688481
   a    0.481300
Name: A, dtype: float64

But when I do either of these:

>>> df.loc["1","a",["A","B"]]
>>> df.loc["1","a"]["A","B"]

I get an error. What I want to obtain is this:

            A         B
1 a  0.688481 -2.380630
  a  0.481300  1.276254

I am new to pandas, and I am sure this is quite a basic question, but I cannot find an answer.

Thanks!

Answer 1

Use slicers for complicated selecting:

idx = pd.IndexSlice
a = df.loc[idx['1','a'],idx['A','B']]
print (a)
            A         B
1 a  0.290874  0.279663
  a  0.435876 -0.318957

---

a = df.loc[idx['1',['a', 'b']],idx['A','B']]
print (a)
            A         B
1 a -1.329079  0.278034
  a  0.955832 -0.877614
  b  1.875801  0.415695
  b  0.765055 -0.828989

---

a = df.loc[idx['1',['a', 'c']],:]
print (a)
            A         B         C         D
1 a  0.142426  0.541231  1.340099 -1.569256
  a -0.510343 -0.447771  0.937850 -0.356663
  c -0.900009 -0.931002 -1.222737 -0.393311
  c -0.957582  2.056467 -1.888492 -1.128331

Answer 2

Use iloc . iloc gets rows (or columns) at particular positions in the index:

df.iloc[:2, :2]