Adaptive ABC model choice and geometric summary statistics for hidden Gibbs random fields
- 285 Downloads
Selecting between different dependency structures of hidden Markov random field can be very challenging, due to the intractable normalizing constant in the likelihood. We answer this question with approximate Bayesian computation (ABC) which provides a model choice method in the Bayesian paradigm. This comes after the work of Grelaud et al. (Bayesian Anal, 4(2):317–336, 2009) who exhibited sufficient statistics on directly observed Gibbs random fields. But when the random field is latent, the sufficiency falls and we complement the set with geometric summary statistics. The general approach to construct these intuitive statistics relies on a clustering analysis of the sites based on the observed colors and plausible latent graphs. The efficiency of ABC model choice based on these statistics is evaluated via a local error rate which may be of independent interest. As a byproduct we derived an ABC algorithm that adapts the dimension of the summary statistics to the dataset without distorting the model selection.
KeywordsApproximate Bayesian computation Model choice Hidden Gibbs random fields Summary statistics Misclassification rate \(k\)-nearest neighbors
The three author were financially supported by the Labex NUMEV. We are grateful to Jean-Michel Marin for his constant feedback and support. Some part of the present work was presented at MCMSki 4 in January 2014 and benefited greatly from discussions with the participants during the poster session. We would like to thank Stéphanie Allassonnière and Nathalie Peyrard for fruitful comments and finally the anonymous referrees and the Editors whose valuable comments and insightful suggestions led to an improved version of the paper.
- Biau, G., Cérou, F., Guyader, A.: New insights into Approximate Bayesian Computation. Annales de l’Institut Henri Poincaré (B) Probabilits et Statistiques, in press (2013)Google Scholar
- Estoup, A., Lombaert, E., Marin, J.-M., Robert, C., Guillemaud, T., Pudlo, P., Cornuet, J.-M.: Estimation of demo-genetic model probabilities with approximate Bayesian computation using linear discriminant analysis on summary statistics. Mol. Ecol. Ressour. 12(5), 846–855 (2012)CrossRefGoogle Scholar
- Everitt, R.G.: Bayesian parameter estimation for latent Markov random fields and social networks. J. Comput. Gr. Stat. 21(4), 940–960 (2012)Google Scholar
- Friel, N.: Bayesian inference for Gibbs random fields using composite likelihoods. In Proceedings of the 2012 Winter Simulation Conference (WSC), pp. 1–8 (2012)Google Scholar
- Hurn, M.A., Husby, O.K., Rue, H.: A Tutorial on Image Analysis. In Spatial Statistics and Computational Methods, volume 173 of Lecture Notes in Statistics pages 87–141. Springer, New York, ISBN 978-0-387-00136-4 (2003)Google Scholar
- Marin, J.-M., Pillai, N.S., Robert, C.P., Rousseau, J.: Relevant statistics for Bayesian model choice. J. R. Stat. Soc. 73(2), 173 (2013)Google Scholar
- Moores, M.T., Hargrave, C.E., Harden, F., Mengersen, K.: Segmentation of cone-beam CT using a hidden Markov random field with informative priors. J. Phys. 489, 012076 (2014)Google Scholar
- Moores, M.T., Mengersen, K., Robert, C.P.: Pre-processing for approximate Bayesian computation in image analysis. ArXiv e-prints March (2014)Google Scholar
- Prangle, D., Fearnhead, P., Cox, M.P., Biggs, P.J., French, N.P.: Semi-automatic selection of summary statistics for ABC model choice. Stat. Appl. Genet. Mol. Biol. 7, 1–16 (2013)Google Scholar
- Tavaré, S., Balding, D.J., Griffiths, R.C., Donnelly, P.: Inferring coalescence times from DNA sequence data. Genetics 145(2), 505–518 (1997)Google Scholar