Selection strategies for regular vine copulae

Abstract

Regular vine (R-vine) copulae are a very flexible class of multivariate copulae, which have received increasing interest in finance and insurance. We will introduce these copulae, discuss their scope and parameter estimation. Since the class of R-vines is huge, model class selection is vital. Recently a top down and a bottom up approach for model selection have been developed. We will discuss these approaches and introduce some useful extensions based on using p-values of goodness-of-fit tests as selection weights. The use of R-vine copulae will be illustrated for a data set involving log concentrations of chemicals in water samples. The performance of these selection procedures are investigated through simulation.