Marginal Sensitivity Model

The Marginal Sensitivity Model (MSM) addresses the challenge of causal inference from observational data when unobserved confounding is present, aiming to provide robust estimates of causal effects despite this limitation. Current research focuses on extending MSM's applicability to various settings, including continuous treatments and complex causal structures, often employing neural networks and kernel methods to improve scalability and achieve sharp bounds on causal effect estimates. This work is significant for enhancing the reliability of causal inferences across diverse fields, from healthcare and economics to climate science, by providing principled methods for quantifying uncertainty due to unobserved confounders.