Summary
In this paper, I apply Rissanen’s minimum description length (MDL) principle to select the number of components in a mixture of Gaussian distributions and estimate the parameters of those distributions. Wolfe’s (1970) and Day’s (1969) maximum likelihood approaches to this problem do not apply directly to the case of unequal component covariance matrices, because the likelihood function diverges. The MDL approach successfully extends maximum likelihood to cover such cases, though. I apply the MDL method to three data sets from the literature. In general, the MDL approach selects simpler models than classical approaches do.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Preview
Unable to display preview. Download preview PDF.
References
Akaike, H. (1973): Information theory and an extension of the maximum likelihood principle. In: B. N. Petrov and F. Csaki, (eds.): Proc. 2nd International Symp. Inf. Theory. Akademia Kiedo, Budapest, 267–281.
Banfield, J., and Raftery, A. E. (1993): Model-based Gaussian and non- Gaussian clustering. Biometrics, 803–821.
Bozdogan, H. (1993): Choosing the number of component clusters in the mixture-model using a new informational complexity criterion of the inverse Fisher information matrix. In: O. Opitz, B. Lausen and R. Klar (eds.): Information and Classification: Concepts, Methods and Applications. Proc. Annual Conf., Gesellschaft fiir Klassifikation, Springer-Verlag, Berlin, 40–54.
Bryant, P. G. (1994): Selecting models using the minimum description length principle. University of Colorado at Denver, College of Business, Faculty Working Paper 1994–14.
Celeux, G. and Govaert, G. (1995): Gaussian parsimonious clustering. Pattern Recognition, 28(5), 781–793.
Day, N. E. (1969): Estimating the components of a mixture of normal distributions. Biometrika, 56, 463–474.
Fisher, R. A. (1922): On the mathematical foundations of theoretical statistics. Philosophical Transactions of the Royal Society of London, Series A, 222, 309–368.
Fisher, R. A. (1936): The use of multiple measurements in taxonomic problems. Annals of Eugenics, VII, Part II, 179–188.
Mclachlan, G. and Basford, K. (1988): Mixture models. Marcel Dekker, Inc., New York.
Miller, R. G. and Halpern, J. W. (1985): Chemical and overt diabetes. In: D. F. Andrews and A. M. Herzberg (eds.): Data. Springer-Verlag, New York, Chapter 36, 215–220.
Rissanen, J. (1987): Stochastic complexity. Journal of the Royal Statistical Society, Series B, 493 223–265.
Rissanen, J. (1989): Stochastic Complexity in Statistical Inquiry. World Scientific Publishing Company, Singapore.
Rissanen, J. (1994): Shannon - Wiener information and stochastic complexity. Address to the N. Wiener Centenary Congress, Michigan State University., East Lansing, Michigan, December 3, 1994. To appear in: Proc. N. Wiener Centenary Congress.
Windham, M. P. and Cutler, A. (1992): Information ratios for validating cluster analyses. Journal of the American Statistical Association, 87, 1188–1192.
Wolfe, J. H. (1970): Pattern clustering by multivariate mixture analysis. Multivariate Behavioral Research, 5, 329–350.
Wright, R. M. and Switzer, P. (1971): Numerical classification applied to certain Jamaican Eocene nummulitids. Mathematical Geology, 33, 297–311.
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 1996 Springer-Verlag Berlin · Heidelberg
About this paper
Cite this paper
Bryant, P.G. (1996). MDL for Mixtures of Normal Distributions. In: Bock, HH., Polasek, W. (eds) Data Analysis and Information Systems. Studies in Classification, Data Analysis, and Knowledge Organization. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-80098-6_1
Download citation
DOI: https://doi.org/10.1007/978-3-642-80098-6_1
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-60774-8
Online ISBN: 978-3-642-80098-6
eBook Packages: Springer Book Archive