Quantile-Based Estimation of the Finite Cauchy Mixture Model
Open Access
Abstract
Heterogeneity and outliers are two aspects which add considerable complexity to the analysis of data. The Cauchy mixture model is an attractive device to deal with both issues simultaneously. This paper develops an Expectation-Maximization-type algorithm to estimate the Cauchy mixture parameters. The main ingredient of the algorithm are appropriately weighted component-wise quantiles which can be efficiently computed. The effectiveness of the method is demonstrated through a simulation study, and the techniques are illustrated by real data from the fields of psychology, engineering and computer vision.Keywords
Funding Information
- King Abdullah University of Science and Technology (D-257-363-1439)
This publication has 26 references indexed in Scilit:
- Centrality Metrics’ Performance Comparisons on Stock Market DatasetsSymmetry, 2019
- Asymmetric impact of gold, oil prices and their volatilities on stock prices of emerging marketsResources Policy, 2016
- A NEW FAMILY OF GENERALIZED GAMMA DISTRIBUTION AND ITS APPLICATIONJournal of Mathematics and Statistics, 2014
- Mixture models with an improper componentJournal of Applied Statistics, 2011
- Rayleigh Mixture DistributionJournal of Applied Mathematics, 2011
- A finite mixture of two Weibull distributions for modeling the diameter distributions of rotated-sigmoid, uneven-aged standsCanadian Journal of Forest Research, 2001
- Robust mixture modelling using the t distributionStatistics and Computing, 2000
- Asymptotic Number of Roots of Cauchy Location Likelihood EquationsThe Annals of Statistics, 1985
- The Study of Outliers: Purpose and ModelJournal of the Royal Statistical Society Series C: Applied Statistics, 1978
- On the Estimation of Mixing DistributionsThe Annals of Mathematical Statistics, 1966