Abstract
Many modern tech companies, such as Google, Uber, and Didi, utilize online experiments (alsoknown as AB testing) to evaluate new policies against existing ones. Analyzing the causarelationship between platform policies and outcomes of interest is of great importance to improvekey platform metrics. This study focuses on capturing dynamic treatment effects in complextemporal/spatial experiments and designing informative experiments. We propose atemporallspatio-temporal varying coefficient decision process (VCDP) model to characterizedynamic treatment effects. Average treatment effects are decomposed into direct and indirecieffects (DE and lE) with estimation and inference procedures developed for both. Meanwhile, weestablish a framework for calculating conditional guantile treatment effects (CQTE) based onindependent characteristics. Notably, we demonstrate that dynamic CQTE equals the sum otindividual CQTEs across time under specific model assumptions. Additionally, we propose threeoptimal allocation strategies for seguential treatments in dynamic settings to minimize variance inIreatment effect estimation. Estimation procedures based on off-policy evaluation (OPE) methodsare developed. Theoretical properties of the proposed methods are established, including weakconvergence, asymptotic power, and optimality of the proposed treatment allocation designExtensive simulations and real data analyses support the usefulness of the proposed methods.
Speaker Intro
Hongtu Zhu is a tenured professor of biostatistics, statistics, computer science, and genetics atUniversity of North Carolina at Chapel Hil. He was DiDi Fellow and Chief Scientist of Statistics atDiDi Chuxing between 2018 and 2020 and was Endowed Bao-Shan Jing Professorship inDiagnostic lmaging at MD Anderson Cancer Center between 2016 and 2018. He is aninternationally recognized expert in statistical learning, medical image analysis, precision medicine,biostatistics, artificial intelligence, and big data analytics. He has been an elected Fellow ofAmerican Statistical Association and Institute of Mathematical Statistics since 2011. He received anestablished investigator award from Cancer Prevention Research Institute of Texas in 2016 andreceived the lNFORMS Daniel H. Wagner Prize for Excellence in Operations Research Practice in2019. He has published more than 320+ papers in top journals including Nature, Science, Cell.Nature Genetics, PNAS, AOS, JASA, and JRsSB, as well as 55+ conference papers in topconferences including NeUrIPS, AAAI, KDD, ICDM, MICCAl, and IPMI.
