A Legendre multiwavelets approach to copula density estimation

Abstract

In this paper, a novel method for copula density estimation using Legendre multiwavelet is proposed. In general, copula density estimation methods based on the multiwavelet benefit from some useful properties, including they are symmetric, orthogonal and have compact support. In particular, the Legendre multiwavelet as a more general and vector-valued polynomial type of wavelets would results a more flexible and accurate approximation for the given copula density. In addition to high ability and nice properties of Legendre multiwavelet in approximation, its support is defined on unit interval, [0,1], as copulas that are normalized to have the support on the unit square and uniform marginal. We further make this approximation method more accurate by using multiresolution techniques. The comparative study reveals that the approximation proposed in this paper is more accurate than a scalar wavelet bases approximation. We eventually apply presented method to approximate multivariate distribution using pair-copula as a flexible multivariate copula to model a dataset of Norwegian financial data.