Kernel density estimation for compositional data. (English) Zbl 0585.62069

Aitchison, J.; Lauder, I. J.

J. R. Stat. Soc., Ser. C 34, 129-137 (1985).

Summary: Although rich parametric families of distributions over the simplex now exist for describing patterns of variability of compositional data, there remain problems where such descriptions fail. For such cases this paper suggests two main kernel methods of density estimation and compares their performance on real and simulated data sets.

MSC:

62G05

Nonparametric estimation

Keywords:

Dirichlet distribution; logistic-normal distribution; simplex; compositional data; kernel methods; density estimation