Date: Wed. 22th October at 3PM
Title: Extremile scalar-on-function regression with application to climate scenarios
Abstract: Understanding and modeling the risk of climatic extreme events is an increasingly imperative task. Addressing this challenge requires robust statistical tools, and since climate variables like temperature and wind speed are intrinsically continuous phenomena, a functional data approach is particularly suitable. To this end, this talk introduces an extremile regression model—a method more robust than quantile regression that has an intrinsic link to extreme value theory—tailored specifically for such functional covariates. The presentation will establish the theoretical framework and introduce the concept of conditional extremities, followed by a demonstration of the model's applicability through the analysis of the CH2018 dataset to quantify future climate behaviour.