部分 label 索引上的 Pandas MultiIndexing KeyError

Question

I'm trying to reproduce the "Basic indexing on axis with MultiIndex" portion of the pandas user guide .

arrays = [['bar', 'bar', 'baz', 'baz', 'foo', 'foo', 'qux', 'qux'],
          ['one', 'two', 'one', 'two', 'one', 'two', 'one', 'two']]
df = pd.DataFrame(np.random.randn(8, 4), index=arrays)
print(df)
print(df['bar'])

I can't see why my code is any different from what is shown in the user guide. I have a KeyError: 'bar' in the last line.

Answer 1

If you want to access bar you need to use loc , or transpose it, because bar is not in columns:

>>> df.loc['bar']
            0         1         2         3
one  0.888182  0.066730  1.397408 -0.550522
two -0.258916 -1.859689 -0.294348 -0.646791

transpose:

>>> df.T['bar']

        one       two
0  0.888182 -0.258916
1  0.066730 -1.859689
2  1.397408 -0.294348
3 -0.550522 -0.646791

EDIT:

As pointed out, very helpfully, in the comments, in case you want to keep the multiindex:

>>> df.xs('bar',drop_level=False)
                0         1         2         3
bar one -0.857271  1.271094  0.565691 -0.523375
    two  0.826911  0.244787  0.991158 -0.484815

Even though it will depend on your use-case, seems like you actually need:

>>> arrays = [['bar', 'bar', 'baz', 'baz', 'foo', 'foo', 'qux', 'qux'],
          ['one', 'two', 'one', 'two', 'one', 'two', 'one', 'two']]
>>> df = pd.DataFrame(np.random.randn(4, 8), columns=arrays)

>>> df

        bar                 baz                 foo                 qux          
        one       two       one       two       one       two       one       two
0  1.600817 -1.420187 -0.798078  1.632550 -0.737740 -1.036077 -1.034157  1.576907
1  0.111148 -1.830283  0.507195 -0.042425  0.260859 -1.600065 -0.449921  0.657582
2 -1.054305 -0.885309  0.325678 -0.253772 -0.444176  0.331933  0.332281  0.127738
3  1.071590  0.947280 -0.973616  0.677141  0.133742  1.352731 -0.210731  2.079073

# Then you can do either df['bar'] directly

References:

1. df.xs

2. df.T

3. df.loc

4. The link in your question. I would suggest you to go through it thoroughly, it is very well explained.

Answer 2

The part of the user guide you are referring to ( this ) is for when the dataframe's columns are multiindex. You should look at this part for when the index is multiindex.

To access bar , you should do:

print(df.loc["bar"])