When
Where
Sriram Somanchi
Assistant Professor of Business Analytics, Mendoza College of Business, University of Notre Dame
Efficient Discovery of Heterogeneous Treatment Effects in Randomized Experiments via Anomalous Pattern Detection
Abstract: In the recent literature on estimating heterogeneous treatment effects, each proposed method makes its own set of restrictive assumptions about the intervention’s effects and which subpopulations to explicitly estimate. Moreover, the majority of the literature provides no mechanism to identify which subpopulations are the most affected–beyond manual inspection–and provides little guarantee on the correctness of the identified subpopulations. Therefore, we propose Treatment Effect Subset Scan (TESS), a new method for discovering which subpopulation in a randomized experiment is most significantly affected by a treatment. We frame this challenge as a pattern detection problem where we efficiently maximize a nonparametric scan statistic over subpopulations. Furthermore, we identify the subpopulation which experiences the largest distributional change as a result of the intervention, while making minimal assumptions about the intervention’s effects or the underlying data generating process. In addition to the algorithm, we demonstrate that the asymptotic Type I and II error can be controlled, and provide sufficient conditions for detection consistency–i.e., exact identification of the affected subpopulation. Finally, we validate the efficacy of the method by discovering heterogeneous treatment effects in simulations and in real-world data from a well-known program evaluation study.
Bio: Dr. Sriram Somanchi is an Assistant Professor of Business Analytics at Mendoza College of Business. His research harnesses the power of large-scale data and machine learning to discover subgroups that are statistically robust and theoretically grounded. Subgroup discovery is crucial to address contemporary business and management challenges, as we navigate away from generic ‘average’ solutions towards an era marked by increasingly customized solutions. His primary domain of application is in healthcare, where his methods contribute to the promise of personalization, improve the efficiency of healthcare delivery, and enrich the clinical and operational aspects of healthcare management. Additionally, he showcases the wide applicability of subgroup discovery methods to address important issues in digital experimentation, crowdsourcing, behavioral economics, and service operations. To solve these intricate problems, his research draws on a rich foundation in social science and statistical machine learning to develop and deploy methods that bridge these related, but distinct disciplines, thus breaking new ground in a nascent academic landscape. His research has been published in the Journal of Machine Learning Research (JMLR), the Journal of Computational and Graphical Statistics (JCGS), ACM Transactions of Information Systems (ACM TOIS), Manufacturing and Service Operations Management (M&SOM), Production and Operations Management (POM), Journal of American Medical Association (JAMA) Network Open, Statistics and Medicine, as well as leading conferences. Dr. Somanchi has a Ph.D. in Information Systems and Management from Heinz College at Carnegie Mellon University. He is a graduate of the Machine Learning Department at CMU and earned an M.E. in computer science from the Indian Institute of Science, Bangalore, India.