Testing for Homogeneity in Mixture Using Weighted Relative Entropy
Authors:
- 13 October 2008
- journal article
- research article
- Published by Taylor & Francis in Communications in Statistics - Simulation and Computation
- Vol. 37 (10) , 1981-1995
Abstract
In this article, we propose a test for homogeneity based on Kullback–Leibler information (also known as relative entropy). Though widely used in hypothesis testing problems, Kullback–Leibler information is not desirable to many researchers in the context of mixture because of its complicated form. In this article, a weighted relative entropy test (WE test), which has closed form expression in terms of the parameter estimators, is proposed. Theoretical results show that this test is consistent. Some simulation results demonstrate that the WE test is better than some leading tests when the mixture components come from normal distribution, and is competitive with them in the Poisson case. The usage of the test is illustrated in an example with data about acidity index of lakes.Keywords
This publication has 23 references indexed in Scilit:
- Testing homogeneity in discrete mixturesJournal of Statistical Planning and Inference, 2008
- Testing Homogeneity in a Mixture Distribution via theL2Distance Between Competing ModelsJournal of the American Statistical Association, 2004
- Likelihood ratio test for univariate Gaussian mixtureJournal of Statistical Planning and Inference, 2001
- A Modified Likelihood Ratio Test for Homogeneity in Finite Mixture ModelsJournal of the Royal Statistical Society Series B: Statistical Methodology, 2001
- How Many Clusters? Which Clustering Method? Answers Via Model-Based Cluster AnalysisThe Computer Journal, 1998
- Testing for the number of components in a mixture of normal distributions using moment estimatorsComputational Statistics & Data Analysis, 1994
- An Application of the Laplace Method to Finite Mixture DistributionsJournal of the American Statistical Association, 1994
- Modeling Lake-Chemistry Distributions: Approximate Bayesian Methods for Estimating a Finite-Mixture ModelTechnometrics, 1992
- Multi-sample cluster analysis using Akaike's Information CriterionAnnals of the Institute of Statistical Mathematics, 1984
- A new look at the statistical model identificationIEEE Transactions on Automatic Control, 1974