Forecasting, structural time series models and the Kalman filter. From an initial draft of the introduction.

There are two reasons for wishing to model a univariate time series. The first one is to provide a description of the series in terms of its components of interest. The other motive is the prediction of future observations. A structural time series model is one which is set up in terms of components which have a direct interpretation. The principle structural time series model are regression models in which the explanatory variables are functions of time and the parameters are time- varying.