A0212
Title: Dirichlet distribution the superhero leading to robust innovations
Authors: Andriette Bekker - University of Pretoria (South Africa) [presenting]
Mohammad Arashi - Ferdowsi University of Mashhad (Iran)
Abstract: The Dirichlet distribution is a well-known candidate for modeling compositional data sets. However, in the presence of outliers, this distribution fails to provide a robust model, adequate for outlying data sets, and these challenging issues require continuous exploration of alternative approaches. Such a drawback can be overcome by resorting to the beta-generating technique, which is a well-known mechanism in developing flexible models. The Kummer-Dirichlet distribution and the gamma distribution are coupled, and the development results in the proposal of the Kummer-Dirichlet gamma distribution, which has great flexibility in modeling. The method of maximum likelihood is applied in the estimation of the parameters. A model testing technique will be briefly reviewed to evaluate the performance. The usefulness of this newly proposed Dirichlet model is demonstrated through the application of synthetic and real data sets, where outliers are present. Finally, this candidate is briefly introduced as a model of prior under a Bayesian framework.
