Introduction
Before running an A/B test, a cross-validation fold, or a pilot study, you need groups
that are genuinely random and roughly balanced in size — not an artifact of how the
data happened to be sorted. CreateRandomSampleGroups answers "how do I split this
population into n unbiased, reproducible groups?" by shuffling rows and assigning
them evenly, with a random seed for reproducibility. Reach for it whenever the
validity of a downstream comparison depends on the groups being genuinely random, not
just conveniently split.
Teaching Note
The function is particularly useful for:
- A/B testing and randomized controlled trials (assigning users to treatment/control)
- K-fold cross-validation (manually partitioning data into folds)
- Pilot studies where a representative sample is needed from a larger population
- Stress testing systems by splitting load into different processing buckets
- Marketing campaigns where different offers are sent to random customer segments
- Creating development, validation, and test sets with custom ratios
Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
dataframerequired | | — | The input pandas DataFrame containing the records to be partitioned. |
number_of_groups | | 2 | The number of discrete groups to create. Records are distributed as evenly as possible across these groups. |
random_seed | | 412 | The seed value for the random number generator to ensure reproducibility. |
Returns
The shuffled DataFrame with an additional group column (integer values) indicating
the assigned group number.
Example
python
from analysistoolbox.data_processing import CreateRandomSampleGroups
import pandas as pd
# Assign customers to Treatment (1) and Control (2) groups
customers = pd.DataFrame({
'customer_id': range(1001, 1007),
'name': ['Alice', 'Bob', 'Charlie', 'David', 'Eve', 'Frank']
})
experimental_groups = CreateRandomSampleGroups(
customers,
number_of_groups=2,
random_seed=42
)