Dynamic Linear Models for Extreme Values

Authors

  • Elvis Rafael Arrazola Acosta Departamento de Estadística Matemática, Universidad Nacional Autónoma de Honduras. https://orcid.org/0009-0008-6807-0896
  • Izhar Asael Alonzo Matamoros Departamento de Estadística Matemática, Universidad Nacional Autónoma de Honduras.
  • Cristian Andrés Cruz Torres Departamento de Estadística Matemática, Universidad Nacional Autónoma de Honduras https://orcid.org/0000-0002-2185-5783

DOI:

https://doi.org/10.5377/pc.v1i19.18700

Keywords:

Extreme Values, Dynamic Linear Models, Markov Chain Monte Carlo, Convolution Processes, Precipitation

Abstract

Currently, climate change is one of the phenomena that worries the world's population, which is why we propose an approach to model measured extreme values of rainfall, drought, etc. First, observations follow a Generalized Extreme Value (GEV) distribution for which location, scale, or shape parameters define the spatiotemporal structure. The generalized distribution of extreme values is extended to incorporate time dependence using a state space representation where state variables are measured through a Dynamic Linear Model (DLM). The spatial element is imposed through the evolution matrix of the DLM where we adopt a form of convolution process. We show how to produce temporal and spatial estimates of our model through a custom Markov Chain Monte Carlo (MCMC) simulation. The methodology is illustrated using extreme data yields through daily measurements of precipitation levels produced daily in Washington State, USA.

Downloads

Download data is not yet available.
Abstract
68
PDF (Español (España)) 49

Published

2024-10-28

How to Cite

Arrazola Acosta, E. R., Alonzo Matamoros, I. A., & Cruz Torres, C. A. (2024). Dynamic Linear Models for Extreme Values. Portal De La Ciencia, 1(19), 37–50. https://doi.org/10.5377/pc.v1i19.18700

Issue

Section

Central Theme