Model selection in hidden Markov models : a simulation study

Costa, Michele ; De Angelis, Luca (2010) Model selection in hidden Markov models : a simulation study. Bologna, IT: Dipartimento di Scienze Statistiche "Paolo Fortunati", Alma Mater Studiorum Università di Bologna, p. 15. DOI 10.6092/unibo/amsacta/2909. In: Quaderni di Dipartimento. Serie Ricerche ISSN 1973-9346.
Full text available as:
Download (155kB) | Preview


A review of model selection procedures in hidden Markov models reveals contrasting evidence about the reliability and the precision of the most commonly used methods. In order to evaluate and compare existing proposals, we develop a Monte Carlo experiment which allows a powerful insight on the behaviour of the most widespread model selection methods. We find that the number of observations, the conditional state-dependent probabilities, and the latent transition matrix are the main factors influencing information criteria and likelihood ratio test results. We also find evidence that, for shorter univariate time series, AIC strongly outperforms BIC.

Document type
Monograph (Working Paper)
Costa, Michele
De Angelis, Luca
Model selection procedure, Hidden Markov model, Monte Carlo experiment, information criteria, likelihood ratio test. Selezione del modello, Modello markoviano latente, Esperimento Monte Carlo, Criterio di informazione, Test del rapporto di verosimiglianza.
Deposit date
15 Dec 2010 13:15
Last modified
16 May 2011 12:15

Other metadata

This work may be freely consulted and used, may be reproduced on a permanent basis in a digital format (i.e. saving) and may be printed on paper with own personal equipment (without availing of third -parties services), for strictly and exclusively personal, research or teaching purposes, with express exclusion of any direct or indirect commercial use, unless otherwise expressly agreed between the user and the author or the right holder. All other rights are reserved. In particular, it is not allowed to retransmit it via telecommunication network, to distribute or send it in any form, including the personal redirection (e-mail).



Staff only: View the document