This project implements methods from the paper Stronger Neyman Regret Guarantees for Adaptive Experimental Design. It is built to test and compare adaptive A/B testing techniques. We compare our adaptive, strongly convex, no-variance-regret average treatment effect (ATE) estimation algorithms with the adaptive no-variance-regret algorithm from Dai et al (2023).
- abtester/: Main library with optimizers and utility functions.
- scripts/: Scripts for data preprocessing, running experiments, and analysis.
-
Clone the repository.
-
Navigate to the project directory and install dependencies using pip or Poetry.
-
Preprocess data:
python scripts/preprocess.py
-
Run experiments:
python -m scripts.run_experiments