|
Abstract:
|
State-space models (SSM) with Markov switching offer a powerful framework for identifying multiple regimes in time series, analyzing interdependence and dynamics within regimes, and characterizing transitions between regimes. Markov-switching models present however great computational challenges due to the exponential number of possible regime sequences to account for. This paper proposes novel statistical methods for Markov-switching SSMs using maximum likelihood estimation, the EM algorithm, and bootstrapping. Considering three switching models, we develop solutions for initializing the EM, handling constraints on model parameters, accelerating convergence, and inferring parameters via parametric bootstrap. We assess the proposed methods by simulation and present applications to EEG studies of epilepsy and of motor imagery. All proposed methods are implemented as MATLAB functions available at https://github.com/ddegras/switch-ssm.
|
