Dynamic Dependence Networks: Multiregression Dynamic Models for Financial Time Series and Portfolio Decisions

• Activity Number: 337
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Business and Economic Statistics Section
• Author(s): Zoey Yi Zhao*+ and Mike West
• Companies: Duke University and Duke University
• Keywords: Multiregression dynamic models ; Sparse conditional dependence network ; Cholesky stochastic volatility ; Bayesian model averaging ; Parallel algorithm ; Portfolio optimization

Abstract

We discuss novel developments in Bayesian analytics for the class of multiregression dynamic models. Our core focus is on increasingly high-dimensional multivariate time series; we aim to enable analysis that can be decomposed into sets of univariate models processed in parallel yet coupled for forecasting and decisions. Our innovation involves sparse dynamic conditional dependence networks (DCDNs), idiosyncrasies in time-varying auto-regressive lag structures, and flexibility of discounting methods for stochastic volatilities. DCDNs provide an inherently parallelizable framework under which we develop on-line forecasting and decisions, broadening the current theory and methods of Bayesian dynamic models. DCDNs require model structure search and uncertainty analysis; we propose an innovative method to address this by transforming the model uncertainties from multivariate to univariate context using Bayesian model averaging and power discounting techniques. The examples show that this can succeed in effectively capturing time-varying model uncertainties on various model parameters, while also identifying practically superior predictive and lucrative models in financial application.