Computationally intensive techniques for a fully Bayesian, decision theoretic approach to financial forecasting and portfolio selection

This paper considers the problem of modelling and forecasting for multivariate financial time series. The use of Dynamic Linear State Space models and Stochastic Volatility models with Kalman filtering techniques to address this problem is considered in the context of providing a modular software implementation. The combination of these two approaches is presented with an illustrative example. We also show how a marginal posterior forecast distribution may be used in order to implement a fully Bayesian decision theoretic approach to portfolio selection.