Title: Parsimonious inverted Wishart processes for multivariate stochastic volatility
Authors: Joshua Chan - Purdue University (United States)
Roberto Leon-Gonzalez - GRIPS (Japan) [presenting]
Rodney Strachan - The University of Queensland (Australia)
Abstract: Efficient algorithms are developed for Bayesian inference in inverted Wishart multivariate stochastic volatility models that are invariant to the ordering of variables. Furthermore, the algorithm searches for factor-type restrictions that reduce the dimension of the latent states. We combine particle Gibbs with an efficient proposal density, Hamiltonian Monte Carlo and Parameter-Expansion-Data-Augmentation, to produce an algorithm that can handle moderately large dimensions and sample sizes. We illustrate the methods with an application to several macroeconomic and financial datasets.