Title: Robust extreme event analysis
Authors: Henry Lam - University of Michigan (United States) [presenting]
Abstract: A robust optimization approach is investigated to estimate and quantify the uncertainty of extremal measures of interest from limited data. The approach relies on finding worst-case tail distributions under geometric assumptions and other calibrated auxiliary constraints. We will present some structural results and solution procedures. We will illustrate how the approach balances statistical accuracy with conservatism and connects to conventional extreme value theory.