A nonparametric method for producing isolines of bivariate exceedance probabilities (Q127498)

Let \(\mathbf{X}=\left(X_{1},X_{2} \right)^{T} \in \mathbb{R}^{2} \) be a random vector. The isolines indicating regions of equal joint exceedance probability for bivariate data are given by \[l_{\mathbf{X}}\left(p \right) := \left\lbrace \mathbf{x} \in \mathbb{R}^{2} : \bar{F}_{\mathbf{X}} \left(\mathbf{x}\right) = p \right\rbrace , \] where \( \bar{F}_{\mathbf{X}} \left(\mathbf{x} \right) = P \left( \mathbf{X} > \mathbf{x} \right) =P\left(X_{1} > x_{1},X_{2} > x_{2} \right) \) and \(p\) are exceedance probabilities. The authors assume that the random vector \(\mathbf{X}\) is regularly varying in the sense used in multivariate extreme value theory and propose a method for estimating isolines \(l_{\mathbf{X}}\left(p \right)\). The procedures of the isoline estimate are presented in two cases: asymptotic dependence and asymptotic independence. Simulation studies and applications to real data are also given.