About "fillna()"...kind like a bug?

[j-takurou]

Actually I wanna replace NaN with "mode" value of the variable.
But the following code doesn't work.

# Your code here
a = df.Embarked.mode()
df.Embarked[df.Embarked.isna()].replace(a.values[0], inplace=True)
df.Embarked[df.Embarked.isna()]
>>>61     NaN
829    NaN
Name: Embarked, dtype: object

[j-takurou]

I don't know why fillna() doesn't work....
Let me know if this is a bug or my error :(

[MarcoGorelli]

Hi Jumpo-523 :)

First, what's df.Embarked? You're typically expected to write examples that others can reproduce. For example:

import pandas as pd 
import numpy as np 

df = pd.DataFrame({'Embarked': [1, 2, np.nan]}) 
a = df.Embarked.mode() 
df.Embarked[df.Embarked.isna()].replace(a.values[0], inplace=True) 
df.Embarked[df.Embarked.isna()]                                         

Actual output:

2   NaN
Name: Embarked, dtype: float64

Expected output: empty Series.

Second, it's often useful to know what versions you're using, so if you could fill out this part of the issue template, I think people here will appreciate it:

#### Output of ``pd.show_versions()``

<details>

[paste the output of ``pd.show_versions()`` here below this line]

</details>

[j-takurou]

I'm so sorry!
The part of data is like the following. Maybe you could replicate it.

[1]data = pd.DataFrame({"PassengerId":{"50":51,"51":52,"52":53,"53":54,"54":55,"55":56,"56":57,"57":58,"58":59,"59":60,"60":61,"61":62,"62":63,"63":64,"64":65,"65":66,"66":67,"67":68,"68":69},
    "Embarked":{"50":"S","51":"S","52":"C","53":"S","54":"C","55":"S","56":"S","57":"C","58":"S","59":"S","60":"C","61":np.nan,"62":"S","63":"S","64":"C","65":"C","66":"S","67":"S","68":"S"}})
[2]data.Embarked.fillna(data.Embarked.mode(), inplace=True)
[3]data
>>>   PassengerId Embarked
50           51        S
51           52        S
52           53        C
53           54        S
54           55        C
55           56        S
56           57        S
57           58        C
58           59        S
59           60        S
60           61        C
61           62      NaN
62           63        S
63           64        S
64           65        C
65           66        C
66           67        S
67           68        S
68           69        S
[4]pd.__version__
>>>'0.25.1'

If you don't mind it, please try it.

SIncerely

[MarcoGorelli]

Sure. I'm not a maintainer, but I don't think this is a bug, as .replace is meant to be used by first specifying the values to replace, and then the values to replace them with. You have specified only the replacement value.

I believe you should use fillna to replace null values - in your example:

df.Embarked.fillna(a.values[0], inplace=True)

EDIT

sorry, just got confused between your original post (when you used replace) and your new one where you used fillna, will update now

UPDATE

data.Embarked.fillna(data.Embarked.mode().values[0])

will work for your example.

If you pass a Series to fillna, then pandas uses its index to decide which values to replace, see the docs

Values not in the dict/Series/DataFrame will not be filled.

[j-takurou]

First of all, I tried to use "fillna" but doesn't work then I tried "replace" and either does't work...
With your comment, I understood the solution but could not understand the mechanism.
I will check documents you recommended!
Thank you so much! :) !!!!

[TomAugspurger]

Thanks @MarcoGorelli.