A new parametric family of high-dimensional non-exchangeable extreme-value copulas is presented. The construction is based on the Lévy-frailty model introduced by Mai and Scherer. This model offers great flexibility and a tractable approach for constructing multivariate probability distributions in large dimensions. Dependence is induced by Lévy subordinators. In order to ensure non-exchangeability inhomogeneous rate parameters are chosen for the exponentially distributed trigger variables. The thus derived family of one-factor Lévy-frailty copulas (LFC) is studied in detail. Furthermore, an algorithm for random sampling of the copula is provided. In a second step an estimator for the parameters of the established LFC is developed. The estimation idea consists in minimizing the mean squared error of the underlying Bernstein function (Laplace exponent of the subordinator) and certain estimates thereof. The estimates of the Bernstein function are shown to be strongly consistent. Additionally, the performance of the estimator is tested via Monte Carlo simulation. The results turn out to be quite precise, in most cases already for a sample size of 50.
«