B1029
Title: Order selection in mixed hidden Markov model
Authors: Yiqi Lin - The Chinese University of Hong Kong (Hong Kong) [presenting]
Abstract: In the recent decades, Hidden Markov model (HMM) is widely used in many research fields. However, traditional HMMs frequently assume that the number of hidden states (order of HMM) is a constant and should be specified prior to analysis. This assumption is apparently unrealistic and too restrictive in many applications. We consider HMMs by allowing the number of hidden states to be unknown and determined by the data. We propose a novel likelihood-based penalized method, along with an efficient Monte Carlo expectation conditional maximization (MCECM) algorithm, to simultaneously perform order selection and parameter estimation in the context of HMMs. Simulation studies are conducted to evaluate the performance of the proposed method. An application of the proposed methodology to a real-life study is presented.
