Noise-Indicator Arma Model with Application in Fitting Physically-Based Time Series

Abstract

In this paper we propose modification of a linear autoregressive moving-average (ARMA) model by using the so-called Noise-Indicator time series. The obtained model, named NIN–ARMA model, is nonlinear threshold autoregressive one. The basic stochastic properties of the NIN–ARMA model have been analyzed and the Empirical Characteristic Function (ECF) method has been used for parameters estimation. Finally, the NIN–ARMA model has been applied in fitting of two actual, real-based physical time series.