Predictor-Dependent Modeling for Bivariate Extremes

• Activity Number: 341
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Risk Analysis
• Author(s): Daniela A. Castro*+ and Miguel de Carvalho and Jennifer L. Wadsworth
• Companies: Pontificia Universidad Católica de Chile and Pontificia Universidad Católica de Chile and University of Cambridge
• Keywords: Bivariate extremes values ; Predictor-dependent probability measures ; Spectral measure ; Statistics of extremes

Abstract

The spectral density plays a key role in bivariate extreme value modeling. In this talk I will discuss a regression model for the spectral density of a bivariate extreme value distribution, which allows us to assess how extreme value dependence can evolve over a certain predictor. For estimating our model we propose a kernel-based estimator which produces mean-constrained on the unit simplex, and hence valid spectral densities. Simulated and real data case studies will be used throughout to illustrate our methods.