Mixture models and atypical values

  • N. A. Campbell
Article

Abstract

The resolution of a mixture of two or more populations into its component distributions may be markedly influenced by one or a few atypical values. The maximum likelihood solution effectively assigns (part of) each observation to one or another of the components via the posterior probabilities, even though the observation may be widely discrepant from all components. This paper presents a modification of the mixture problem, in which the typicality of each observation is considered, as well as the posterior probabilities, with the contribution of atypical observations being downweighted. The extension to the multivariate case is discussed.

Key words

mixture models maximum likelihood solution atypical values multivariate case 

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. Aitchison, J., Habbema, J. D. F., and Kay, J. W., 1977, A critical comparison of two methods of statistical discrimination: Appl. Stat., v. 26, no. 1, p. 15–25.Google Scholar
  2. Campbell, N. A., 1980, Robust procedures in multivariate analysis, I, Robust covariance estimation: Appl. Stat., v. 29, no. 3, p. 231–237.Google Scholar
  3. Clarke, M. R. B., 1971, Updating the sample mean and dispersion matrix: Appl. Stat., v. 20, no. 2, p. 206–209.Google Scholar
  4. Collins, J. R., 1976, Robust estimation of a location parameter in the presence of asymmetry: Ann. Stat., v. 4, no. 1, p. 68–85.Google Scholar
  5. Day, N. E., 1969, Estimating the components of a mixture of normal distributions: Biometrika, v. 56, no. 3, p. 463–474.Google Scholar
  6. Dempster, A. P., Laird, N. M., and Rubin, D. B., 1977, Maximum likelihood from incomplete data via the EM algorithm (with Discussion): Jour. Roy. Stat. Soc. Ser. B., v. 39, no. 1, p. 1–38.Google Scholar
  7. Fowlkes, E. B., 1979, Some methods for studying the mixture of two normal (lognormal) distributions: J. Amer. Stat. Assoc., v. 74, no. 367, p. 561–575.Google Scholar
  8. Hampel, F. R., 1973, Robust estimation: A condensed partial survey: Z. Wahr. verw. Geb., v. 27, p. 87–104.Google Scholar
  9. Hampel, F. R., 1974, The influence curve and its role in robust estimation: J. Amer. Stat. Assoc., v. 69, no. 346, p. 383–393.Google Scholar
  10. Hasselblad, V., 1966, Estimation of parameters for a mixture of normal distributions: Technometrics, v. 8, no. 3, p. 431–446.Google Scholar
  11. Hogg, R. V., 1977, An introduction to robust procedures: Commun. Statist.—Theor. Meth., v. A6, no. 9, p. 789–794.Google Scholar
  12. Holgersson, M. and Jorner, U., 1978, Decomposition of a mixture into normal components: A review: J. Bio-Med. Comp., v. 9, p. 367–392.Google Scholar
  13. Holland, P. W. and Welsch, R. E., 1977, Robust regression using iteratively reweighted leastsquares: Commun. Stat.—Theor. Meth., v. A6, no. 9, p. 813–827.Google Scholar
  14. Huber, P. J., 1964, Robust estimation for a location parameter: Ann. Math. Stat., v. 35, no. 1, p. 73–101.Google Scholar
  15. Sinclair, A. J., 1976, Applications of Probability Graphs in Mineral Exploration. Assoc. of Exploration Geochemists, Rexdale, Ontario, 95 pp.Google Scholar

Copyright information

© Plenum Publishing Corporation 1984

Authors and Affiliations

  • N. A. Campbell
    • 1
  1. 1.Division of Mathematics and StatisticsCSIROWembleyAustralia

Personalised recommendations