A0576
Title: Order selection for heterogeneous semiparametric hidden Markov models
Authors: Yudan Zou - The Chinese University of Hong Kong (Hong Kong) [presenting]
Abstract: Hidden Markov models (HMMs), which can characterize dynamic heterogeneity, are useful instruments for analyzing longitudinal data. The order of HMMs, or the number of hidden states, is typically assumed to be known or predetermined by criterion-based techniques in conventional analysis. Considering pairwise comparisons under criterion-based methods become computationally expensive as the model space expands, a few studies have conducted order selection and parameter estimation simultaneously. However, because they only considered homogeneous parametric instances, they cannot account for circumstances in which non-parametric forms or heterogeneity appear. The aim is to propose a Bayesian double-penalized (BDP) procedure for simultaneous order selection and parameter estimation for heterogeneous semi-parametric HMMs. To overcome the difficulties in updating the order, a brand-new Markov chain Monte Carlo algorithm and an effective adjust-bound reversible jump strategy are created. Simulation results reveal that the proposed BDP procedure performs well in estimation and works noticeably better than the standard criterion-based approaches. Application of the suggested method to the Alzheimer's Disease Neuroimaging Initiative research further supports its usefulness.
