A1362
Title: A Bayesian Lasso for tail index regression
Authors: Miguel de Carvalho - University of Edinburgh and Universidade de Aveiro (Portugal) [presenting]
Daniel Paulin - University of Edinburgh (United Kingdom)
Abstract: Extreme events, such as heatwaves, wildfires, and widespread flooding, can have devastating impacts on communities and ecosystems. The purpose is to introduce a novel regression model for heavy-tailed phenomena, leveraging Bayesian regularization within a generalized additive framework. This approach is centered on a conditional Pareto-type specification, enhanced by Bayesian Lasso shrinkage priors and refined with low-rank thin plate splines basis expansion. The effectiveness of the proposed methods is demonstrated using both artificial and real data, with a focus on extreme wildfire events in Portugal and the key factors driving them.
