Efficient Adaptive Experimental Design for Average Treatment Effect Estimation

2020-02-13

AI Summary

This paper addresses the problem of efficiently estimating the average treatment effect (ATE) in causal inference. We propose an adaptive experimental design framework with three key contributions: (1) We formally define and dynamically learn the optimal treatment assignment probability that minimizes the semiparametric efficiency bound of the ATE estimator; (2) We introduce the A²IPW estimator, which achieves the theoretically optimal asymptotic variance in finite samples; and (3) We construct nonparametric confidence intervals that are valid at any stopping time, enabling rate-optimal sequential testing and early stopping. The method integrates adaptive randomization, semiparametric efficiency theory, and anytime-valid inference. It substantially reduces the sample size required to achieve a target statistical power while guaranteeing strict coverage probability—even in small-sample settings.

Abstract

We study how to efficiently estimate average treatment effects (ATEs) using adaptive experiments. In adaptive experiments, experimenters sequentially assign treatments to experimental units while updating treatment assignment probabilities based on past data. We start by defining the efficient treatment-assignment probability, which minimizes the semiparametric efficiency bound for ATE estimation. Our proposed experimental design estimates and uses the efficient treatment-assignment probability to assign treatments. At the end of the proposed design, the experimenter estimates the ATE using a newly proposed Adaptive Augmented Inverse Probability Weighting (A2IPW) estimator. We show that the asymptotic variance of the A2IPW estimator using data from the proposed design achieves the minimized semiparametric efficiency bound. We also analyze the estimator's finite-sample properties and develop nonparametric and nonasymptotic confidence intervals that are valid at any round of the proposed design. These anytime valid confidence intervals allow us to conduct rate-optimal sequential hypothesis testing, allowing for early stopping and reducing necessary sample size.

Problem

• Adaptive Experimental Design
• A2IPW Estimator
• Treatment Effect Evaluation

Innovation

• Adaptive Experimental Design
• A2IPW Estimator
• Precision-based Trial Termination