A0737
Title: Bayesian estimation of R-vine Copula with Gaussian-mixture GARCH margins
Authors: Rewat Khanthaporn - Health New Zealand Te Whatu Ora (New Zealand)
Nuttanan Wichitaksorn - Auckland University of Technology (New Zealand) [presenting]
Abstract: The purpose is to show the Bayesian estimation of multivariate regular vine (R-vine) copula models with the generalized autoregressive conditional heteroskedasticity (GARCH) margins having the Gaussian-mixture distributions. The Bayesian estimation consists of Markov chain Monte Carlo (MCMC) and variational Bayes (VB) with data augmentation. Due to expensive computation, R-vines have been of limited use while this issue is overcome through parallel computation. To illustrate this, thirteen bivariate copula functions are used for an R-vine pair structure with a large number of marginal distributions where the exponential-type GARCH margins are modelled with the intertemporal capital asset pricing specification through the mixture of Gaussian and generalized Pareto distributions. Results from a simulation study indicate that the proposed models and methods outperform the competing ones. With 100 financial returns in an empirical study, favourable results are still obtained.
