A0699
Title: Adaptive thresholding and tail index estimation under normal and extreme regimes
Authors: Omid Ardakani - Georgia Southern University (United States) [presenting]
Abstract: Heavy-tailed and high-dimensional data present significant challenges for modeling and risk assessment. The application of Bayesian nonparametric methods and generalized Pareto distributions are extended in the context of extreme value theory and high-dimensional settings. First, a Dirichlet process mixture model is developed to estimate thresholds and tail indices. Subsequently, the challenges in high-dimensional models are addressed, focusing on extremal dependence by incorporating a tail adaptation mechanism and copula modeling. The empirical study examines the behavior of macroeconomic variables under normal and extreme regimes using these techniques to offer insight into the tail risks associated with macroeconomic outcomes.
