I'm trying to replace values in multiple columns of a dataframe with numpy.where in Python by doing the following:
df['X, Y, Z'] = np.where(df['X, Y, Z'] < 1, 0, df['X, Y, Z'])
However, it gives me the following error: KeyError: 'X, Y, Z'
I have already tried doing the strings separately, like 'X', 'Y', 'Z', but it doesn't work either.
How do I resolve this?
What about passing a proper list of keys ['X', 'Y', 'Z'] to your dataframe instead of a long string 'X, Y, Z':
import numpy as np
import pandas as pd
data = {'X': np.linspace(0,2,8), 'Y': np.linspace(0,2,8)*2, 'Z': np.linspace(0,2,8)*4}
df = pd.DataFrame.from_dict(data)
which gives:
>>> df
>>> 0 0.000000 0.000000 0.000000
>>> 1 0.285714 0.571429 1.142857
>>> 2 0.571429 1.142857 2.285714
>>> 3 0.857143 1.714286 3.428571
>>> 4 1.142857 2.285714 4.571429
>>> 5 1.428571 2.857143 5.714286
>>> 6 1.714286 3.428571 6.857143
>>> 7 2.000000 4.000000 8.0000000
df[['X', 'Y', 'Z']] = np.where(df[['X', 'Y', 'Z']] < 1, 0, df[['X', 'Y', 'Z']])
and now with no longer KeyError:
>>> df
>>> 0 0.000000 0.000000 0.000000
>>> 1 0.000000 0.000000 1.142857
>>> 2 0.000000 1.142857 2.285714
>>> 3 0.000000 1.714286 3.428571
>>> 4 1.142857 2.285714 4.571429
>>> 5 1.428571 2.857143 5.714286
>>> 6 1.714286 3.428571 6.857143
>>> 7 2.000000 4.000000 8.000000
df['X', Y', 'Z'] is not the same as df[['X', 'Y', 'Z']] that is why. In one dimensional case df['X'] is indeed the same as df[['X']]. Python doesn't split 'X, Y, Z' and interpret it as 3 distincts keys nor it does understand 3 distincts keys, but it needs to iterate through list of keys ['X', 'Y', 'Z']one at a time, hope it's clear now.8 samples that span a range between 0 and 2 where your test condition is meaningful (since you failed to provide a Minimal Reproducible Example). I multiplied by 2 and 4 to get different columns where the condition applied differently. This is arbitrary but shows you that it works for various data :)