Missing Data

Remarks#

Should we include the non-documented ffill and bfill?

Filling missing values

In [11]: df = pd.DataFrame([[1, 2, None, 3], [4, None, 5, 6], 
                            [7, 8, 9, 10], [None, None, None, None]])

Out[11]: 
     0    1    2     3
0  1.0  2.0  NaN   3.0
1  4.0  NaN  5.0   6.0
2  7.0  8.0  9.0  10.0
3  NaN  NaN  NaN   NaN

Fill missing values with a single value:

In [12]: df.fillna(0)
Out[12]: 
     0    1    2     3
0  1.0  2.0  0.0   3.0
1  4.0  0.0  5.0   6.0
2  7.0  8.0  9.0  10.0
3  0.0  0.0  0.0   0.0

This returns a new DataFrame. If you want to change the original DataFrame, either use the inplace parameter (df.fillna(0, inplace=True)) or assign it back to original DataFrame (df = df.fillna(0)).

Fill missing values with the previous ones:

In [13]: df.fillna(method='pad')  # this is equivalent to both method='ffill' and .ffill()
Out[13]: 
     0    1    2     3
0  1.0  2.0  NaN   3.0
1  4.0  2.0  5.0   6.0
2  7.0  8.0  9.0  10.0
3  7.0  8.0  9.0  10.0

Fill with the next ones:

In [14]: df.fillna(method='bfill')  # this is equivalent to .bfill()
Out[14]: 
     0    1    2     3
0  1.0  2.0  5.0   3.0
1  4.0  8.0  5.0   6.0
2  7.0  8.0  9.0  10.0
3  NaN  NaN  NaN   NaN

Fill using another DataFrame:

In [15]: df2 = pd.DataFrame(np.arange(100, 116).reshape(4, 4))
         df2
Out[15]: 
     0    1    2    3
0  100  101  102  103
1  104  105  106  107
2  108  109  110  111
3  112  113  114  115

In [16]: df.fillna(df2) #  takes the corresponding cells in df2 to fill df
Out[16]: 
       0      1      2      3
0    1.0    2.0  102.0    3.0
1    4.0  105.0    5.0    6.0
2    7.0    8.0    9.0   10.0
3  112.0  113.0  114.0  115.0

Dropping missing values

When creating a DataFrame None (python’s missing value) is converted to NaN (pandas’ missing value):

In [11]: df = pd.DataFrame([[1, 2, None, 3], [4, None, 5, 6], 
                            [7, 8, 9, 10], [None, None, None, None]])

Out[11]: 
     0    1    2     3
0  1.0  2.0  NaN   3.0
1  4.0  NaN  5.0   6.0
2  7.0  8.0  9.0  10.0
3  NaN  NaN  NaN   NaN

Drop rows if at least one column has a missing value

In [12]: df.dropna()
Out[12]:
     0    1    2     3
2  7.0  8.0  9.0  10.0

This returns a new DataFrame. If you want to change the original DataFrame, either use the inplace parameter (df.dropna(inplace=True)) or assign it back to original DataFrame (df = df.dropna()).

Drop rows if all values in that row are missing

In [13]: df.dropna(how='all')
Out[13]: 
     0    1    2     3
0  1.0  2.0  NaN   3.0
1  4.0  NaN  5.0   6.0
2  7.0  8.0  9.0  10.0

Drop columns that don’t have at least 3 non-missing values

In [14]: df.dropna(axis=1, thresh=3)
Out[14]: 
     0     3
0  1.0   3.0
1  4.0   6.0
2  7.0  10.0
3  NaN   NaN

Interpolation

import pandas as pd
import numpy as np

df = pd.DataFrame({'A':[1,2,np.nan,3,np.nan],
                   'B':[1.2,7,3,0,8]})

df['C'] = df.A.interpolate()
df['D'] = df.A.interpolate(method='spline', order=1)

print (df)
     A    B    C         D
0  1.0  1.2  1.0  1.000000
1  2.0  7.0  2.0  2.000000
2  NaN  3.0  2.5  2.428571
3  3.0  0.0  3.0  3.000000
4  NaN  8.0  3.0  3.714286

Checking for missing values

In order to check whether a value is NaN, isnull() or notnull() functions can be used.

In [1]: import numpy as np
In [2]: import pandas as pd
In [3]: ser = pd.Series([1, 2, np.nan, 4])
In [4]: pd.isnull(ser)
Out[4]: 
0    False
1    False
2     True
3    False
dtype: bool

Note that np.nan == np.nan returns False so you should avoid comparison against np.nan:

In [5]: ser == np.nan
Out[5]: 
0    False
1    False
2    False
3    False
dtype: bool

Both functions are also defined as methods on Series and DataFrames.

In [6]: ser.isnull()
Out[6]: 
0    False
1    False
2     True
3    False
dtype: bool

Testing on DataFrames:

In [7]: df = pd.DataFrame({'A': [1, np.nan, 3], 'B': [np.nan, 5, 6]})
In [8]: print(df)
Out[8]: 
     A    B
0  1.0  NaN
1  NaN  5.0
2  3.0  6.0    

In [9]: df.isnull()  # If the value is NaN, returns True.
Out[9]: 
       A      B
0  False   True
1   True  False
2  False  False

In [10]: df.notnull()  # Opposite of .isnull(). If the value is not NaN, returns True.
Out[10]: 
       A      B
0   True  False
1  False   True
2   True   True