GroupBy with transform for normalization in Pandas - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of group normalization using transform

What is the output of this code that normalizes values within each group by subtracting the group mean?

Pandas

import pandas as pd

df = pd.DataFrame({
    'group': ['A', 'A', 'B', 'B', 'B'],
    'value': [10, 20, 30, 40, 50]
})

df['norm'] = df.groupby('group')['value'].transform(lambda x: x - x.mean())
print(df)

A

  group  value  norm
0     A     10  -5.0
1     A     20   5.0
2     B     30 -10.0
3     B     40   0.0
4     B     50  10.0

B

  group  value  norm
0     A     10  10.0
1     A     20  20.0
2     B     30  30.0
3     B     40  40.0
4     B     50  50.0

C

  group  value  norm
0     A     10  -10.0
1     A     20  -20.0
2     B     30  -30.0
3     B     40  -40.0
4     B     50  -50.0

D

  group  value  norm
0     A     10   5.0
1     A     20  -5.0
2     B     30  10.0
3     B     40   0.0
4     B     50 -10.0

💡 Hint

Think about how subtracting the mean affects each value within its group.

The transform subtracts the mean of each group from each value, so values in group A (10,20) become (10-15, 20-15) = (-5,5), and in group B (30,40,50) become (30-40, 40-40, 50-40) = (-10,0,10).

❓ data_output

intermediate

Number of normalized values above zero per group

After normalizing values within groups by subtracting the group mean, how many values in group 'B' are greater than zero?

Pandas

import pandas as pd

df = pd.DataFrame({
    'group': ['A', 'A', 'B', 'B', 'B'],
    'value': [5, 15, 10, 20, 30]
})

df['norm'] = df.groupby('group')['value'].transform(lambda x: x - x.mean())
count = df[(df['group'] == 'B') & (df['norm'] > 0)].shape[0]
print(count)

A 0

B 1

C 2

D 3

💡 Hint

Calculate the mean of group B and count values above it.

Group B values are 10, 20, 30 with mean 20. Normalized values are -10, 0, 10. Values greater than zero are only 10, so count is 1. But 0 is not greater than zero. So only one value > 0. Wait, check carefully: 10,20,30 mean 20, norm = value - 20: 10-20=-10, 20-20=0, 30-20=10. Only one value > 0 (30). So correct answer is 1.

🔧 Debug

advanced

Identify the error in group normalization code

What error does this code raise when trying to normalize values by group?

Pandas

import pandas as pd

df = pd.DataFrame({
    'group': ['X', 'X', 'Y'],
    'value': [1, 2, 3]
})

df['norm'] = df.groupby('group')['value'].transform(lambda x: x / x.mean())
print(df)

A TypeError

B KeyError

C ZeroDivisionError

D No error, outputs normalized values

💡 Hint

Check if dividing by mean is valid for these values.

Dividing each value by the mean of its group is valid and does not raise an error. The code outputs normalized values.

🚀 Application

advanced

Apply group normalization with standard deviation scaling

Which code snippet correctly normalizes 'score' within each 'team' by subtracting the mean and dividing by the standard deviation?

A df['norm'] = df.groupby('team')['score'].apply(lambda x: (x - x.mean()) / x.std())

B df['norm'] = (df['score'] - df.groupby('team')['score'].mean()) / df.groupby('team')['score'].std()

C df['norm'] = df.groupby('team')['score'].transform(lambda x: (x - x.mean()) / x.std())

D df['norm'] = df['score'] / df.groupby('team')['score'].transform('std')

💡 Hint

Remember transform keeps the original index and length.

Option C uses transform with a lambda that subtracts mean and divides by std per group, preserving the original DataFrame shape. Option C returns a Series with group indices only, causing misalignment. Option C tries to subtract and divide Series of different lengths without alignment. Option C divides by std but does not subtract mean, so not normalized.

🧠 Conceptual

expert

Why use transform instead of apply for group normalization?

What is the main reason to use groupby with transform instead of apply when normalizing data within groups?

A Transform returns a Series aligned with the original DataFrame, preserving index and length, enabling direct assignment.

B Apply is faster and more memory efficient than transform for group operations.

C Transform can only be used with numeric data, while apply works with any data type.

D Apply automatically sorts groups, while transform does not.

💡 Hint

Think about the shape and alignment of the output from each method.

Transform returns a Series with the same index as the original DataFrame, so it can be assigned directly as a new column. Apply returns a reduced or differently indexed result, which may not align for direct assignment.