Structural Time Series Models and the Kalman Filter: A Concise Review
FEUNL Working Paper No. 541
30 Pages Posted: 20 Nov 2009
Date Written: June 19, 2009
Abstract
The continued increase in availability of economic data in recent years and, more importantly, the possibility to construct larger frequency time series, have fostered the use (and development) of statistical and econometric techniques to treat them more accurately. This paper presents an exposition of structural time series models by which a time series can be decomposed as the sum of a trend, seasonal and irregular components. In addition to a detailled analysis of univariate speci cations we also address the SUTSE multivariate case and the issue of cointegration. Finally, the recursive estimation and smoothing by means of the Kalman lter algorithm is described taking into account its different stages, from initialisation to parameter's estimation.
Keywords: SUTSE, cointegration, ARIMA, smoothing, likelihood
JEL Classification: C10, C22, C32
Suggested Citation: Suggested Citation