Index-mixed copulas (2306.10663v2)

Abstract: The class of index-mixed copulas is introduced and its properties are investigated. Index-mixed copulas are constructed from given base copulas and a random index vector, and show a rather remarkable degree of analytical tractability. The analytical form of the copula and, if it exists, its density are derived. As the construction is based on a stochastic representation, sampling algorithms can be given. Properties investigated include bivariate and trivariate margins, mixtures of index-mixed copulas, symmetries such as radial symmetry and exchangeability, tail dependence, measures of concordance such as Blomqvist's beta, Spearman's rho or Kendall's tau and concordance orderings. Examples and illustrations are provided, and applications to the distribution of sums of dependent random variables as well as the stress testing of general dependence structures are given. A particularly interesting feature of index-mixed copulas is that they allow one to provide a revealing interpretation of the well-known family of Eyraud-Farlie-Gumbel-Morgenstern (EFGM) copulas. Through the lens of index-mixing, one can explain why EFGM copulas can only model a limited range of concordance and are tail independent, for example. Index-mixed copulas do not suffer from such restrictions while remaining analytically tractable.