Pandas: Group by a column that meets a condition

Question

I have a data set with three colums: rating , breed, and dog.

import pandas as pd
dogs = {'breed': ['Chihuahua', 'Chihuahua', 'Dalmatian', 'Sphynx'],
        'dog': [True, True, True, False],
        'rating': [8.0, 9.0, 10.0, 7.0]}

df = pd.DataFrame(data=dogs)

I would like to calculate the mean rating per breed where dog is True. This would be the expected:

  breed     rating
0 Chihuahua 8.5   
1 Dalmatian 10.0  

This has been my attempt:

df.groupby('breed')['rating'].mean().where(dog == True)

And this is the error that I get:

NameError: name 'dog' is not defined

But when I try add the where condition I only get errors. Can anyone advise a solution? TIA

Answer 1

Once you groupby and select a column, your dog column doesn't exist anymore in the context you have selected (and even if it did you are not accessing it correctly).

Filter your dataframe first, then use groupby with mean

df[df.dog].groupby('breed')['rating'].mean().reset_index()

       breed  rating
0  Chihuahua     8.5
1  Dalmatian    10.0

Answer 2

An alternative solution is to make dog one of your grouper keys. Then filter by dog in a separate step. This is more efficient if you do not want to lose aggregated data for non-dogs.

res = df.groupby(['dog', 'breed'])['rating'].mean().reset_index()

print(res)

     dog      breed  rating
0  False     Sphynx     7.0
1   True  Chihuahua     8.5
2   True  Dalmatian    10.0

print(res[res['dog']])

    dog      breed  rating
1  True  Chihuahua     8.5
2  True  Dalmatian    10.0