Author(s): Panagiota Galiatsatou; Panayotis Prinos
Linked Author(s): Panagiotis Prinos
Keywords: Generalized Pareto Distribution (GPD); Ξ-linkage assumption; Pseudo-likelihood; Parametric bootstrapping; Efficiency gains; Return level estimation
Abstract: In the present paper a constant shape parameter over space is used for the Dutch coast on the North Sea to extract results of return levels for a process of daily storm surges, taking account of the degree of spatial dependence among sites. Marginal inference under spatial linkage is conducted for all stations with a ξ-linkage approach, using an algorithm proposed by Butler, (2005). Return levels and their standard errors are extracted under the assumption of spatial linkage and compared to the ones when no parameter linkage is considered. Variance and efficiency gains in the model parameters are estimated and discussed. To calculate these measures, a simple parametric bootstrap method for the case of the ξ-linkage is used.
Year: 2007