A1100
Title: Semiparametric empirical Bayesian analysis of maxima and peaks over threshold
Authors: Stefano Rizzelli - University of Padova (Italy) [presenting]
Abstract: Predicting future observations is the central goal of several statistical applications concerning extreme-value data. Under mild assumptions, extreme value theory justifies modeling linearly normalized sample maxima by max-stable distributions and rescaled excesses of a large threshold by Generalized Pareto distributions. The Bayesian paradigm offers a direct approach to forecasting and uncertainty quantification. Various Bayesian procedures have been proposed in recent years, though they typically disregard the asymptotic bias inherent in the use of extreme value models, incorporating no information on the norming sequences in the prior specifications for location and scale parameters. Some recently proposed empirical Bayes approaches have been reviewed to suitably address this point via data-dependent priors. The resulting asymptotic posterior concentration properties are illustrated, and their implications for estimation and prediction of future extreme observations are pinpointed.
