Abstract
The EM algorithm is a popular method for maximizing a likelihood in the presence of incomplete data. When the likelihood has multiple local maxima, the parameter space can be partitioned into domains of convergence, one for each local maximum. In this paper we investigate these domains for the location family generated by the t-distribution. We show that, perhaps somewhat surprisingly, these domains need not be connected sets. As an extreme case we give an example of a domain which consists of an infinite union of disjoint open intervals. Thus the convergence behaviour of the EM algorithm can be quite sensitive to the starting point.
Similar content being viewed by others
References
Beaton, A. E. and Tukey, J. W. (1974) The fittinHg of power series, meaning polynomials illustrated on band-spectroscopic data, Technometrics, 16, 147–193.
Dempster, A. P., Laird, N. M. and Rubin, D. B. (1977) Maximum likelihood from incomplete data via the EM algorithm (with discussion), Journal of the Royal Statistical Society, B 39, 1–38.
Dempster, A. P., Laird, N. M. and Rubin, D. B. (1980) Iteratively reweighted least squares for linear regression where errors are normal/independent distributed, in Multivariate Analysis V (P. R. Krishnaiah, ed.), North-Holland, New York, pp. 35–37.
Hampel, F. R., Ronchetti, E. M., Rousseeuw, P. J., and Stahel, W. A. (1986) Robust Statistics: The Approach Based on Influence Functions. Wiley, New York.
Huber, P. J. (1981) Robust Statistcs. Wiley, New York.
Lange, K. L., Little, R. J. A., and Tylor, J. M. G. (1989) Robust and statistical modeling using the t-distribution, Journal of the American Statistical Association, 84, 881–896.
Little, R. J. A. and Rubin, D. B. (1987) Statistical Analysis with Missing Data. Wiley, New York.
Wu, C. F. J. (1983) On the convergence properties of the EM algorithms, Annals of Statistics, 11, 95–103.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Arslan, O., Constable, P.D.L. & Kent, J.T. Domains of convergence for the EM algorithm: a cautionary tale in a location estimation problem. Stat Comput 3, 103–108 (1993). https://doi.org/10.1007/BF00147772
Received:
Accepted:
Issue Date:
DOI: https://doi.org/10.1007/BF00147772