The University of Liverpool
Department of Mathematical Sciences
Division of Statistics and Probability
SEMINAR
Multivariate Stochastic Volatility with Bayesian Dynamic Linear Models
Dr K. Triantafyllopoulos
Wednesday, 26th September, 2pm
The Whittaker Room (211)
Abstract:
In this talk we discuss volatility estimation for a wide class of matrix-variate state space models, known also as matrix-variate dynamic linear models. Considering that the time series is observed at roughly equal intervals of time, we develop in closed form a sequential Bayesian algorithm that is suitable for estimation and forecasting. Our approach models the volatility matrix via a stochastic evolution law, which is defined upon the foundations of singular multivariate beta distributions. To complete the stochastic volatility law we use a diagonal matrix of discount factors, which aim is to discount the different volatilities at different rates and thus to provide a more pragmatic modelling approach. We discuss measures of goodness of fit, such as likelihood based, mean of squared standardized forecast errors and value at risk. The methods are illustrated with spot prices of aluminium from the London metal exchange. We find that our approach provides a flexible and pragmatic modelling toolkit that is applicable to a wide range of multivariate time series data and it is fast and easily implementable.
Following the talk, tea and biscuits will be available in Room 304
ALL WELCOME
*************************************************
Ingrid Harper
Division of Statistics and Probability
Department of Mathematical Sciences
University of Liverpool
Mathematical Sciences Building
Peach Street
Liverpool L69 7ZL
Tel: 0151 794 4751
Fax: 0151 794 4754
*************************************************
|