Nonparametric estimation in semi-parametric univariate mixture models. (English) Zbl 1060.62041

Cruz-Medina, I. R.; Hettmansperger, T. P.

J. Stat. Comput. Simulation 74, No. 7, 513-524 (2004).

Summary: We introduce new estimates of the mixing proportions, locations, and variances of the components of a finite univariate mixture model. We assume that the components are symmetric and differ only in the locations. No parametric model assumptions are imposed on the components. Further, when there is additional information available in the form of training samples that contain information concerning the mixing proportion, the new methods are robust to the symmetry assumption.

MSC:

62G05

Nonparametric estimation

Keywords:

tetranomial approach; sextinomial approach; simulations; EM algorithm; Finite mixture models; Training samples