Let's consider following data frame:
I want to change string-type elements of this DataFrame into NaN. Example of an solution would be:
frame.replace("k", np.NaN)
frame.replace("s", np.NaN)
However it would be very problematic in bigger data sets to go through each element, checking if this element is string and changing it at the end. Is there an easier solution?
Desired table:
Use pd.to_numeric to transform all non numeric values to nan:
frame = frame.apply(pd.to_numeric, errors='coerce')
Use df.replace regex
import numpy as np
df.replace(regex='[A-Za-z]', value=np.nan)
You can use astype(str) and .str.digit for each column to get a mask for values that are numbers, and then just index the dataframe with that mask to make NaN the values that aren't masked:
df = df[df.astype(str).apply(lambda col: col.str.isdigit())]
Output:
>>> df
0 1 2
0 1 2 NaN
1 2 NaN 4
2 5 NaN 1