B0307
Title: Flexible modeling for spatial extremes with application to environmental studies
Authors: Yuan Tian - North Carolina State University (United States) [presenting]
Brian Reich - North Carolina State University (United States)
Abstract: Extreme events, such as unusually high temperature, major precipitation and life-threatening hurricanes, occur with an extremely small probability, but may have catastrophic consequences. It is therefore of great significance to make inferences and predictions about these rare events. Extreme value analysis plays a major role in modeling these rare events. One common approach is the max-stable process. A Bayesian hierarchical structure has been previously proposed that can be easily implemented and efficiently computed via MCMC algorithm to present the max-stable process. We extend the current RS model using nonparametric Bayesian modeling that relaxes the assumptions. Thus, we provide more flexibility in expanding the class of marginal distributions. In addition, we present a hybrid model that combines the strength of the original RS model and the nonparametric model. Tail behavior of the novel model is studied. The utility of the proposed model is evaluated in Monte Carlo simulation studies and is applied to precipitation data for the US.
