Economic forecasting. ARMA models as linear and time invariant (L.T.I.)

Abstract

The objective of this paper is to point out the relationship between the timeseries modelling used in Econometrics and the techniques frequently used in Digital Signal Processing. A time series is a discrete time signal. This relationship makes itpossible use the large body of knowledge and the high number of techniques of application in this last area in order to model, process, represent, predict and in general to extract information from the time series frecuently encountered in Economy.A clear example is the short time series modelling applied to prediction, an important problem in economy, and where well-known algorithms coming from the Digital Signal Processing area may be applied: for instance, methods based on the eigen structure of the signal (time series) autocorrelation matrix.