A1234
Title: Comparing two multivariate stochastic volatility models
Authors: Victor Pena - Universitat Politecnica de Catalunya (Spain) [presenting]
Abstract: Multivariate stochastic volatility models are useful for tracking time-varying patterns in covariance structures. Uhlig extended (UE) and beta-Bartlett (BB) processes are especially convenient for analyzing high-dimensional time series because they are conjugate with Wishart likelihoods. It is shown that UE and BB are closely related but not equivalent: their hyperparameters can be matched so that they have the same forward-filtered posteriors and one-step ahead forecasts but different joint (smoothed) posterior distributions. Under this circumstance, Bayes factors cannot discriminate the models, and alternative approaches to model comparison are needed. These issues are illustrated in a retrospective analysis of the volatilities of returns of foreign exchange rates.
