Multivariate Structural Time Series Models - (Now published in 'System Dynamics in Economic and Financial Models', CHeij, H Schumacher, B Hanzon and C Praagman (eds.) John Wiley & Sons, Chichester (1997), pp.269-298.)
Much of economic analysis presupposes that certain economic time series can be decomposed into trends and cycles. Structural time series models are explicitly set up in terms of such unobserved components. This paper sets up various multivariate structural time series models, shows how they can model the data parsimoniously and how they can aid the analysis of the interrelationships between time series. he computations are carried out using the new STAMP 5.0 package. The graphics and diagnostics play a key role in the development of a model selection methodology.