Crosstab on multiple columns

user_3346095 0
Viewed 12664

I have a dataframe with a name, day, and location. For each name-day-location triple, I want to know what proportion of the rows with that name-day have that location.

In code, I am starting with df and looking for expected.

import pandas as pd

df = pd.DataFrame(
    [
        {"name": "Alice", "day": "friday", "location": "left"},
        {"name": "Alice", "day": "friday", "location": "right"},
        {"name": "Bob", "day": "monday", "location": "left"},
    ]
)

print(df)



expected = pd.DataFrame(
    [
        {"name": "Alice", "day": "friday", "location": "left", "row_percent": 50.0},
        {"name": "Alice", "day": "friday", "location": "right", "row_percent": 50.0},
        {"name": "Bob", "day": "monday", "location": "left", "row_percent": 100.0},
    ]
).set_index(['name', 'day', ])
print(expected)

Printed:

In [13]: df                                                                                                                                                                                  
Out[13]: 
      day location   name
0  friday     left  Alice
1  friday    right  Alice
2  monday     left    Bob




In [12]: expected                                                                                                                                                                            
Out[12]: 
             location  row_percent
name  day                         
Alice friday     left         50.0
      friday    right         50.0
Bob   monday     left        100.0
pandas python
1 Answers

Using groupby and value_counts:

df.groupby(['name', 'day']).location.value_counts(normalize=True).mul(100)


name   day     location
Alice  friday  left         50.0
               right        50.0
Bob    monday  left        100.0
Name: location, dtype: float64

With a bit more cleaning for your desired output:

out = (df.groupby(['name', 'day']).location.value_counts(normalize=True).mul(100)
          .rename('row_percent').reset_index(2))


             location  row_percent
name  day
Alice friday     left         50.0
      friday    right         50.0
Bob   monday     left        100.0

out == expected


              location  row_percent
name  day
Alice friday      True         True
      friday      True         True
Bob   monday      True         True
user_34832030
Related