fastrerandomize

An R package for hardware-accelerated rerandomization and exact randomization testing in experimental design.

Features

• 🚀 GPU/XLA-accelerated acceptable randomization generation
• 🔢 Supports both exact enumeration and Monte Carlo sampling
• 📉 Built-in balance metrics (Hotelling's T²) and custom threshold functions
• 📈 Randomization-based inference with fiducial intervals
• 💾 Memory-efficient, key-based batched processing for large experiments

Installation

# Install from CRAN
install.packages("fastrerandomize")

# Build Python backend (requires conda)
library(fastrerandomize)
build_backend(conda_env = "fastrerandomize")

About

Package by Rebecca Goldstein, Connor Jerzak, Aniket Kamat, and Fucheng Warren Zhu