Primary Submission Category: Design of Experiments
Towards Efficient Statistical Inference and Optimal Design in Adaptive Experiments
Authors: Wenxin Zhang, Mark van der Laan,
Presenting Author: Wenxin Zhang*
Adaptive experiments play a crucial role in clinical trials and online A/B testing. Unlike static, non-adaptive trial designs, adaptive experimental designs dynamically adjust treatment randomization probabilities and other key design elements in response to data collected sequentially during the experiment. These designs are useful for achieving different objectives, such as reducing uncertainty in causal estimand estimation or improving benefit of participants within the experiment. Despite their advantages, the adaptive nature of these designs and the time-dependent nature of the data introduce significant challenges in making unbiased statistical inferences from non-i.i.d. data.
Building upon the Targeted Maximum Likelihood Estimator (TMLE) literature that has provided valid statistical inference tailored to adaptive experimental settings using inverse weighting strategies tailored for adaptive experiment settings, we propose a new TMLE that further improves the efficiency for estimating causal estimands under adaptive designs.
Beyond efficient statistical inference, we further introduce a general framework for implementing optimal adaptive designs, customized to achieve various objectives efficiently. The performance of our proposed estimators and adaptive designs is demonstrated through theoretical analysis and extensive simulations.