Fortified Proximal Causal Inference with many imperfect negative controls
Abstract: Causal inference from observational data often relies on the assumption of no unmeasured confounding, an assumption frequently violated in practice due to unobserved or poorly measured covariates. Proximal causal inference (PCI) offers a promising framework for addressing unmeasured confounding using a pair of outcome and treatment confounding proxies, commonly known as negative controls. However, existing PCI methods typically assume all specified proxies are genuine, which may be unrealistic and is untestable without extra assumptions. In this talk, after providing an overview of the PCI framework, we describe a new "fortified" PCI setting which is robust to many invalid treatment confounding proxies. Thus, we establish the somewhat surprising result of nonparametric identification of the population average treatment effect (ATE) under the assumption that a non-trivial lower bound on the number of genuine proxies is given to the analyst, without requiring specific knowledge of which proxies are genuine. For inference, we propose corresponding estimators which are multiply robust and locally semiparametric efficient for the ATE. These estimators are thus simultaneously robust to invalid proxies and partial model misspecification of nuisance parameters. The proposed methods are evaluated in an extensive simulation study, and applied to assess the effect of right heart catheterization in critically ill patients.
This is joint work with Myeonghun Yu and Xu Shi. Link to the paper: https://www.arxiv.org/
Eric J. Tchetgen Tchetgen is The University Professor, Professor of Biostatistics at the Perelman School of Medicine, and Professor of Statistics and Data Science at The Wharton School, University of Pennsylvania. He co-directs the Penn Center for Causal Inference, a leading initiative dedicated to advancing and disseminating causal inference methodologies in the health and social sciences. Professor Tchetgen Tchetgen is a leading expert in Causal Inference, Missing Data, and Semiparametric Theory, with impactful applications in HIV research, Genetic Epidemiology, Environmental Health, and Alzheimer’s Disease and related aging disorders. He is an Amazon scholar working with Amazon scientists, applying causal inference techniques to address complex challenges in the tech industry.
In recognition of his groundbreaking research, Professor Tchetgen Tchetgen was a co-recipient of the inaugural 2022 Rousseeuw Prize awarded to Causal Inference for pioneering research on causal inference with real-world applications in medicine and public health. In 2024, he was awarded the Marshall Joffe Epidemiologic Methods Research Award by the Society of Epidemiologic Research; and in 2025, he was awarded the inaugural David Cox Medal for Statistics, which celebrates exceptional mid-career researchers for their outstanding contributions to the field. Professor Tchetgen Tchetgen received this honor in recognition of his groundbreaking contributions to statistical theory and methodology, particularly in causal inference. His work has significantly advanced the discipline, most notably through the development of Proximal Causal Inference and instrumental variable methodology – two essential frameworks for addressing confounding in causal analysis.
This seminar will be held via Zoom only and will be recorded.
Time and date:
Thursday, October 23
9am - 10am AEDT
Join:
https://monash.zoom.us/j/84631593134?pwd=y8ZqSJIQHk6OFOAb2t0k4dKvNAugen.1
Or, go to https://monash.zoom.us/join and enter:
Meeting ID: 846 3159 3134
Passcode: 560452