ANALYSIS OF MULTIVARIATE STATISTICAL MODELS WITH HETEROGENEOUS STRUCTURE IN THE CASE OF HIDDEN MARKOV DEPENDENCE OF THE STATE

Abstract

For vector autoregressive model with heterogeneous endogenous-exogenous structure and Markov switching states we propose the EM-algorithm for mixture decomposition, as well as discriminant analysis algorithm for classification of new observations. Accuracy of the algorithms is examined by means of computer simulation experiments.