Multisite Data Simulation

Overview

uniharmony.datasets.make_multisite_classfication helps researchers generate synthetic datasets that mimic real-world multi-center studies. Whether you’re testing machine learning algorithms, developing statistical methods, or teaching data science concepts, this simulator creates realistic data with built-in complexities that mirror actual research scenarios.

Usage

from uniharmony import make_multisite_classification

# Generate data with 3 sites, 500 samples total
X, y, sites = make_multisite_classification(
    n_sites=3,
    n_samples=500,
    n_features=20,
    random_state=42  # For reproducibility
)
# X contains your features
# y has labels
# sites tells you which site each sample came from

Behind the Scenes

Each sample’s features combine three components that can be adapted:

Feature = Signal + Noise + Site Effect

Flexible Configuration

You can choose how many sites you want, set the total number of samples, control how many features and control the class balance. Work with binary classification (yes/no) or multi-class problems.