Abstract
In two-component mixtures of exponential distributions, different strategies for starting the likelihood maximization algorithm converge to different types of maxima. The power of an LR test of homogeneity against such a mixture strongly depends on the considered strategy, and global maximization need not result in the largest power. An explanation is given on basis of a systematic investigation of the likelihood function in a large number of simulations, using a variety of diagnostic tools. Thereby, we also gain a deeper insight into the properties of the samples that generate particular types of solutions of the likelihood equation. In particular, “spurious solutions” often occur; these are mainly responsible for the fact that global maximization may not result in a statistically meaningful estimator. Removing the smallest elements of a sample may drastically increase the power of previously inferior strategies.
Similar content being viewed by others
References
Biernacki, C., Celeux, G. and Govaert, G. (2003). Strategies for getting the highest likelihoods in mixture models, Special Issue on Recent Developments in Mixture Models (Guest Editors: D. Böhning and W. Seidel),Computational Statistics & Data Analysis,41, 561–575.
Böhning, D. (2000).Computer-Assisted Analysis of Mixtures and Applications, Chapman & Hall/CRC, Boca Raton.
Böhning, D. and Schlattmann, P. (1992). Computer-assisted analysis of mixtures (C.A.MAN): Statistical algorithms,Biometrics,48, 283–303.
Böhning, D., Dietz, E., Schaub, R., Schlattmann, P. and Lindsay, B. (1994). The distribution of the likelihood ratio for mixtures of densities from the one-parameter exponential family,Annals of the Institute of Statistical Mathematics,46, 373–388.
Dacunha-Castelle, D. and Gassiat, E. (1999). Testing the order of a model using locally conic parametrization: Population mixtures and stationary ARMA processes,The Annals of Statistics,27, 1178–1209.
Ghosh, J. K. and Sen, P. K. (1985). On the asymptotic performance of the log likelihood ratio statistic for the mixture model and related results,Proceedings of the Berkeley Conference in Honor of Jerzy Neyman and Jack Kiefer (eds. L. M. Le Cam and R. A. Olshen),2, 789–806, Wadsworth, Belmont, California.
Hartigan, J. A. (1985). A failure of likelihood asymptotics for normal mixtures,Proceedings of the Berkeley Conference in Honor of Jerzy Neyman and Jack Kiefer (eds. L. M. Le Cam and R. A. Olshen),2, 807–810, Wadsworth, Belmont, California.
Karlis, D. (2001). A cautionary note about the EM algorithm for finite exponential mixtures, Tech. Report, No. 150, Department of Statistics, Athens University of Economics.
Karlis, D. and Xekalaki, E. (2003). Choosing initial values for the EM algorithm for finite mixtures, Special Issue on Recent Developments in Mixture Models (Guest Editors: D. Böhning and W. Seidel),Computational Statistics & Data Analysis,41, 577–590.
Lesperance, M. and Kalbfleisch, J. (1992). An algorithm for computing the nonparametric MLE of a mixing distribution,Journal of the American Statistical Association,87, 120–126.
Lindsay, B. G. (1995).Mixture Models: Theory, Geometry and Applications, Institute of Mathematical Statistics, Hayward, California.
Liu, X. and Shao, Y. (2003). Asymptotics for likelihood ratio tests under loss of identifiability,The Annals of Statistics,31, 807–832.
McLachlan, G. and Peel, D. (2000).Finite Mixture Models, Wiley, New York.
Seidel, W. and Ševčíková, H. (2002a). Efficient calculation of the NPMLE of a mixing distribution for mixtures of exponentials,Discussion Papers in Statistics and Quantitative Economics,96, Universität der Bundeswehr Hamburg.
Seidel, W. and Ševčíková, H. (2002b). Tools for analyzing and maximizing likelihood functions in mixture models,Discussion Papers in Statistics and Quantitative Economics,104, Universität der Bundeswehr Hamburg.
Seidel, W. and Ševčíková, H. (2003). A detailed investigation of likelihood maxima in two-component exponential mixture models and their implication on LR tests,Discussion Papers in Statistics and Quantitative Economics,106, Universität der Bundeswehr Hamburg.
Seidel, W., Mosler, K. and Alker, M. (2000a) A cautionary note on likelihood ratio tests in mixture models,Annals of the Institute of Statistical Mathematics,52, 481–487.
Seidel, W., Mosler, K. and Alker, M. (2000b). Likelihood ratio tests based on subglobal optimization: A power comparison in exponential mixture models,Statistical Papers,41, 85–98.
Seidel, W., Ševčíková, H. and Alker, M. (2000c). On the power of different versions of the likelihood ratio test for homogeneity in an exponential mixture model,Discussion Papers in Statistics and Quantitative Economics,92, Universität der Bundeswehr Hamburg.
Author information
Authors and Affiliations
Additional information
This research has been supported by a grant from the Deutsche Forschungsgemeinschaft.
About this article
Cite this article
Seidel, W., Ševčíková, H. Types of likelihood maxima in mixture models and their implication on the performance of tests. Ann Inst Stat Math 56, 631–654 (2004). https://doi.org/10.1007/BF02506480
Received:
Revised:
Issue Date:
DOI: https://doi.org/10.1007/BF02506480