Testing for structural Heterogeneity in vector autoregressive model by means of statistical classification methods

Abstract

The problems of analysis of complex systems, described by VARX model with structural heterogeneity, are studied in this article. It is supposed that struc- tural heterogeneity may be conditioned as a result of two reasons: 1) presence of different system state classes; 2) presence of structural changes. It is supposed that every state of the system is characterized by the stationary model, i.e. the general model is a segmented stationary VARX. The methods of discriminant analysis for VARX models are suggested for the classification of the complex system states. The results of the classification are afterwards used for the es- timation of the moments of single or multiple structural changes by means of statistical testing. The simulation results show the acceptable accuracy of the proposed methods.