multipca

A Python package for Multi-way Principal Component Analysis (Multiway PCA) with confidence intervals.

Citation

If you use this package in your research, please cite:

@article{ouyang2023multiway,
  title={On the Multiway Principal Component Analysis},
  author={Ouyang, Jialin and Yuan, Ming},
  journal={Annals of Statistics (to appear)},
  year={2023},
  url={https://arxiv.org/abs/2302.07216}
}

Installation

Requires Python ≥ 3.9. We recommend using a virtual environment:

# Create and activate virtual environment (optional but recommended)
python -m venv .venv
source .venv/bin/activate  # On Windows, use `.venv\Scripts\activate`

# Install package
pip install multipca

Quick Start

The package includes an example Citibike dataset that demonstrates multiway PCA analysis of bike sharing patterns in NYC. Here's how to get started:

import multipca
import matplotlib.pyplot as plt

# Load the example Citibike data (shape: 24 hours × 35 stations × 522 days)
array = multipca.load_citibike_example()

# MPCA with confidence intervals
results = multipca.mpca_CI(array, r=4)

# Plot the first component's daily pattern with confidence bands
multipca.CI_band_plot(results, rank=1, mode=1)
plt.show()

# If you only need components without confidence intervals
components = multipca.mpca(array, r=4)

For more examples including World Bank economic indicators analysis, check out examples/citibike_worldbank.ipynb in the GitHub repository.

Documentation

Full documentation is available in the GitHub repository.

Examples

Check out the example notebook for:

• Detailed usage examples
• Plotting functions
• Real-world applications with Citibike and World Bank data

Features

• Multiway PCA implementation
• Confidence intervals for components
• Visualization tools:
  • Component plots with confidence bands
  • Loading plots

License

This project is licensed under the MIT License - see the LICENSE file for details.