Purpose: To study large-sample tests of extreme-value dependence for multivariate copulas with max-stability characterization. Summary: Marginal cumulative distribution function of random variables forming a d-dimensional random vector are considered to obtain the cumulative distribution function of the overall d-dimension random vector. For this purpose, the terminology of copula is used and it is suggested that under the conditions of max-stable, the copula function becomes an extreme-value copula. It this is an unknown copula, then the problem is to test whether this unknown copula belongs to the class of extreme-value copulas. By empirically estimating the unknown copula form the given data, the test statistics are framed and used for testing the hypothesis. Further, the Monte Carlo experiments for data sets of dimension two to five are presented to study the finite-sample performance of a large number of candidate test statistics. The proposed test procedures are illustrated on bivariate financial data and trivariate geological data.
展开▼
