Title: Model selection tests in the Markov switching models
Authors: Anna Czapkiewicz - AGH University of Science and Technology (Poland) [presenting]
Antoni Leon Dawidowicz - Jagiellonian University (Poland)
Abstract: Markov switching models are discussed in two cases: models with constant transition matrix and models with time-varying probabilities in the transition matrix. We can not use the classical tests to compare these models since some regular conditions are not fulfilled. We propose approaches of comparing discussed models in two cases. Firstly, we compare models with different number of regimes. Secondly, we compare a model with a constant transition matrix and a model with time-varying transition matrix. For testing purposes we study asymptotic properties of maximum likelihood estimators.