Biometrika Advance Access originally published online on October 29, 2009
Biometrika 2009 96(4):975-982; doi:10.1093/biomet/asp056
Miscellanea |
Maximum likelihood estimation using composite likelihoods for closed exponential families
Department of Statistics, University of Leeds, Leeds, LS2 9JT, U.K. k.v.mardia{at}leeds.ac.uk j.t.kent{at}leeds.ac.uk
Novartis International AG, CH-4002 Basel, Switzerland ghughes{at}live.co.uk
Department of Statistics, University of Leeds, Leeds, LS2 9JT, U.K. c.c.taylor{at}leeds.ac.uk
Received for publication 1 April 2008. Revision received 1 May 2009.
In certain multivariate problems the full probability density has an awkward normalizing constant, but the conditional and/or marginal distributions may be much more tractable. In this paper we investigate the use of composite likelihoods instead of the full likelihood. For closed exponential families, both are shown to be maximized by the same parameter values for any number of observations. Examples include log-linear models and multivariate normal models. In other cases the parameter estimate obtained by maximizing a composite likelihood can be viewed as an approximation to the full maximum likelihood estimate. An application is given to an example in directional data based on a bivariate von Mises distribution.
Key Words: Bivariate von Mises distribution • Closed exponential family • Fisher information • Log-linear model • Maximum likelihood • Multivariate normal distribution • Pseudolikelihood
References
-
Arnold B. C., Castillo E., Sarabia J. M. Conditionally specified distributions: an introduction. Statist. Sci. (2001) 16:249–65.[CrossRef]
Arnold B. C., Strauss D. J. Pseudolikelihood estimation: some examples. Sankhy 
(1991a) B 53:233–43.
Arnold B. C., Strauss D. J. Bivariate distributions with conditionals in prescribed exponential families. J. R. Statist. Soc. (1991b) B 53:365–75.
Barndorff-Nielsen O. E., Cox D. R. Inference and Asymptotics (1994) London: Chapman and Hall.
Besag J. E. Spatial interaction and the statistical analysis of lattice systems (with discussion). J. R. Statist. Soc. (1974) B 34:192–236.
Besag J. E. Statistical analysis of non-lattice data. Statistician (1975) 24:179–95.[CrossRef][Web of Science]
Besag J. E. Efficiency of pseudolikelihood estimation for simple Gaussian fields. Biometrika (1977) 64:616–18.
Boomsma W., Mardia K. V., Taylor C. C., Ferkinghoff-Borg J., Krogh A., Hamelryck T. A generative, probabilistic model of local protein structure. Proc. Nat. Acad. Sci. (2008) 105:8932–37.
Brook D. On the distinction between the conditional probability and the joint probability approaches in the specification of nearest-neighbour systems. Biometrika (1964) 51:481–83.
Cox D. R., Reid N. A note on pseudolikelihood constructed from marginal densities. Biometrika (2004) 91:729–37.
Davison A. C. Statistical Models (2003) Cambridge: Cambridge University Press.
Godambe V. P. An optimum property of regular maximum likelihood equation. Ann. Math. Statist. (1960) 31:1208–11.[CrossRef]
Kent J. T. Robust properties of likelihood ratio tests. Biometrika (1982) 69:19–27.
Lindsay B. G. Composite likelihood methods. Contemp. Math. (1988) 80:221–39.
Mardia K. V., Hughes G., Taylor C. C., Singh H. Multivariate von Mises distribution with applications to bioinformatics. Can. J. Statist. (2008) 36:99–109.[CrossRef]
Mardia K. V., Taylor C. C., Subramaniam M. Protein bioinformatics and mixtures of bivariate von Mises distributions for angular data. Biometrics (2007) 63:505–12.[CrossRef][Web of Science][Medline]
Singh H., Hnizdo V., Demchuk E. Probabilistic model for two dependent circular variables. Biometrika (2002) 89:719–23.
Varin C. On composite marginal likelihoods. Adv. Statist. Anal. (2008) 92:1–28.[CrossRef]
Varin C., Vidoni P. A note on composite likelihood inference and model selection. Biometrika (2005) 92:519–28.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||