How to select the elements of a Pandas DataFrame given a Boolean mask?-CodePudding

I was wondering wether, given a boolean mask, there is a way to retreive all the elements of a DataFrame positioned in correspondance of the True values in the mask.

In my case I have a DataFrame containing the values of a certain dataset, for example let's take the following :

l = [[5, 3, 1], 
     [0, 3, 1], 
     [7, 3, 0], 
     [8, 5, 23], 
     [40, 4, 30],
     [2, 6, 13]]
df_true = pd.DataFrame(l, columns=['1', '2', '3'])
df_true

Then I randomly replace some of the values with 'np.nan' as follows:

l2 = [[5, 3, np.nan], 
     [np.nan, 3, 1], 
     [7, np.nan, 0], 
     [np.nan, 5, 23], 
     [40, 4, np.nan],
     [2, np.nan, 13]]
df_nan= pd.DataFrame(l2, columns=['1', '2', '3'])
df_nan

Let's say that after applying some imputation algorithm I obtained as a result:

l3 = [[5, 3, 1], 
     [2, 3, 1], 
     [7, 8, 0], 
     [8, 5, 23], 
     [40, 4, 25],
     [2, 6, 13]]
df_imp= pd.DataFrame(l3, columns=['1', '2', '3'])
df_imp

Now I would like to create two lists (or arrays), one containing the imputed values and the other one the true values in order to compare them. To do so I first created a mask m = df_nan.isnull() which has value True in correspondance of the cells containing the imputed values. By applying the mask as df_imp[m] I obtain:

     1       2       3
0   NaN     NaN     1.0
1   2.0     NaN     NaN
2   NaN     8.0     NaN
3   8.0     NaN     NaN
4   NaN     NaN     25.0
5   NaN     6.0     NaN

Is there a way to get instead only the values without also the Nan, and put them into a list?

CodePudding user response：

You can use df.values to return a numpy representation of the DataFrame then use numpy.isnan and keep other values.

import numpy as np
arr = df.values
res = arr[~np.isnan(arr)]
print(res)
# [1. 2. 8. 8. 25. 6.]