MICROCOMPUTER BASED INTERACTIVE ANALYSIS OF UNIVARIATE AND MULTIVARIATE ARIMA MODELS.

Abstract

The Box-Jenkins methodology is a model-based approach to analyzing and forecasting time-series which has been especially successful in applications to short-term forecasting. Many other time-series methods, such as the various forms of exponential smoothing, are special cases of this model. The approach presupposes a form known as the ARIMA model, short for AutoRegressive Integrated Moving Average. In its forecasting form, the forecasted value of the time-series is a finite linear combination of previous observations of the series, and of previous observations of a series of shock terms. In this paper, the emphasis is mainly on the estimation of the coefficients of ARMA models by various nonlinear approximation algorithms. A nonlinear approximation method can be simply described as a trial and error approach to minimizing the error in a model. Three algorithms are presented with examples.