Effects of Mis-Specification in Bivariate Extreme Value Problems |
| |
Authors: | Debbie J Dupuis Jonathan A Tawn |
| |
Institution: | (1) Department of Statistical and Actuarial Sciences, University of Western Ontario, London, Ontario, N6A 5B7, Canada;(2) Department of Mathematics and Statistics, Lancaster University, Lancaster, LA1 4YF, U.K. |
| |
Abstract: | The need to incorporate the structure of complex problems in extreme value analyzes, and the requirement to exploit all the limited information that is available, has led to the increased use of advanced dependence models. When they are appropriate, these dependence models can lead to substantial benefits over simpler univariate extreme value methods. Here we explore some inference problems for the marginal and conditional distributions caused by model mis-specification. We find distinct differences in estimation characteristics when the dependence structure is asymptotically dependent or asymptotically independent, and that conditional models can be substantially improved if the variables are standardized to have common marginal distributions. |
| |
Keywords: | asymptotic independence bivariate extreme value distribution conditional distribution covariates dependence generalized extreme-value distribution |
本文献已被 SpringerLink 等数据库收录! |
|