Setting values on pandas DataFrame with multiindex

Question

The following is a minimal example of what I am trying to do. I have a pandas DataFrame with multiindex as follows

import pandas as pd
import numpy as np

arrays = [['bar', 'bar', 'baz', 'baz', 'foo', 'foo', 'qux', 'qux'],
          ['one', 'two', 'one', 'two', 'one', 'two', 'one', 'two']]
tuples = list(zip(*arrays))
index = pd.MultiIndex.from_tuples(tuples, names=['first', 'second'])
s = pd.DataFrame(np.random.randn(8,2), index=index)

So the DataFrame I have is

                     0         1
first second                    
bar   one    -3.174428 -0.314160
      two     0.968316  0.278967
baz   one     0.171292 -0.789257
      two     1.420621  0.100964
foo   one    -1.001074 -0.517729
      two    -0.211823  0.951422
qux   one     1.173289  0.313692
      two    -0.159855  0.149710

What I want is to set all the observations with the index "second" equal to two as -1. What I have in mind is using .loc, something as follows:

s.loc[(:,'two')]

but .loc would not accept the ":" operator.

Could someone help here?

MaxU - stand with Ukraine · Accepted Answer · 2018-02-22T21:14:38.020

2

Option 1:

In [127]: s.loc[pd.IndexSlice[:, 'two'], :] = -1

In [128]: s
Out[128]:
                     0         1
first second
bar   one    -0.581647  0.225254
      two    -1.000000 -1.000000
baz   one     0.705050 -1.414695
      two    -1.000000 -1.000000
foo   one     0.359795  1.468521
      two    -1.000000 -1.000000
qux   one    -0.481149 -0.241922
      two    -1.000000 -1.000000

Option 2:

In [137]: s.loc[(slice(None),'two'), :] = -11

In [138]: s
Out[138]:
                      0          1
first second
bar   one      2.144487   0.024400
      two    -11.000000 -11.000000
baz   one     -0.177128  -1.088566
      two    -11.000000 -11.000000
foo   one     -0.780979   2.701814
      two    -11.000000 -11.000000
qux   one     -0.981635  -0.202875
      two    -11.000000 -11.000000

edited Feb 22 '18 at 21:14

answered Feb 22 '18 at 10:51

MaxU - stand with Ukraine

205,989
36
386
419

Thanks for the reply, but the first method does not work. I don't work which version you are using, but the latest version returns KeyError message. The second method works, but the syntax seems complicated. I guess if there are no better options, I will use the second method. – user3821012 Feb 22 '18 at 20:48
@user3821012, i'm using `pandas: 0.22.0` and I do prefer a second option as it much clearer (for me)... – MaxU - stand with Ukraine Feb 22 '18 at 20:50
As the results of the first method shows, it does not return the desired results--the values for "two" do not change. – user3821012 Feb 22 '18 at 21:00
and If you use "s.loc[(None, 'two'), :]", it returns error. – user3821012 Feb 22 '18 at 21:01
@user3821012, i don't get any error message... Please post your desired data set – MaxU - stand with Ukraine Feb 22 '18 at 21:03
Your second method gives the desired results, except for the last line with "NaN" – user3821012 Feb 22 '18 at 21:06

Setting values on pandas DataFrame with multiindex

1 Answers1