Accuracy analysis of different components evaluation of time series

Abstract

Mathematical models for decomposing and forecasting of time series are analysed in this paper. Sometimes we have the situations when we don't get good result describing time series by standard components (trend, cyclical, seasonal). It is often happened with economic indicators. For example, the method of evaluation is changed or economic crisis has happened. Inclusion of new variables – intervention variables – can help with these problems. There is investigation of model accuracy depending on accuracy of evaluation of different components. Time series of different structure were simulated for imitation of real situations. Estimation of time series models with intervention variables and traditional components was done by program Demetra.