Approximations for bivariate extreme values
Let \((X,Y)\) denote a random vector with distribution function (d.f.) \(F\) and suppose that both \(X\) and \(Y\) are standardized to have \[ \text{Pr}(X\leq x)=\text{Pr}(Y\leq x)=\exp(-x^{-1}),\quad x>0. \] In recent years a number of statistical models have been proposed for extreme values of \((X,Y).\) The basis for these models is the assumption that \(F\) belongs to the domain of attraction of a bivariate extreme value df, \(G\), for instance. One of several equivalent limit laws in extreme value theory for this assumption to hold is that as \(t\to\infty\) the joint tail of \(F\) is asymptotically the same as the joint tail of \(G.\) All the models proposed for bivariate extreme values are based on zero order approximations of limit laws. The aim of this paper is to provide certain general approximations about the tail behavior of \(F\) that could form a basis for improved modeling of bivariate extreme values. The tail behavior of distributions in the domain of attraction of bivariate extreme value distributions is studied. Results on finite approximations of the tail behavior and its analytical shape are provided. The results could form a basis to improve current statistical modeling of bivariate extreme values.
- On Pickands coordinates in arbitrary dimensions
- On the distribution of Pickands coordinates in bivariate EV and GP models
- Expansions for bivariate extreme value distributions
- A polynomial model for bivariate extreme value distributions
- Peaks-over-threshold stability of multivariate generalized Pareto distributions
- scientific article; zbMATH DE number 952758 (Why is no real title available?)
This page was built for publication: Approximations for bivariate extreme values
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5936311)
