UCPH Statistics Seminar: Juraj Bodik

Title: Estimating Cross-World and Extreme Counterfactuals

Speaker: Juraj Bodik from University of Lausanne

Abstract: In many applications, what matters is not the average benefit but the risk of rare, worst-case harm; we study the tails of the individual treatment effect (ITE) distribution. While average and quantile treatment effects (ATE and QTE) are well understood in randomized trials, the distribution of ITE remains only partially identified because joint potential outcomes are unobservable. The missing unobserved quantity is cross-world dependence between the potential outcomes. In this talk, we discuss how we can estimate it from data and move towards more reliable and informative estimation of extreme quantiles of ITEs.