A constrained em algorithm for univariate normal mixtures
- 1 January 1986
- journal article
- research article
- Published by Taylor & Francis in Journal of Statistical Computation and Simulation
- Vol. 23 (3) , 211-230
- https://doi.org/10.1080/00949658608810872
Abstract
The EM algorithm is widely used to calculate maximum-likelihood estimates corresponding to a mixture of normal distributions. The algorithm is altered using simple constraints which increase robustness against poor initial guesses while maintaining a low work requirement per iteration. The modified algorithm is described and the results of some numerical tests are given.Keywords
This publication has 6 references indexed in Scilit:
- Mixture Densities, Maximum Likelihood and the EM AlgorithmSIAM Review, 1984
- Some Methods for Studying the Mixture of Two Normal (Lognormal) DistributionsJournal of the American Statistical Association, 1979
- Estimating Mixtures of Normal Distributions and Switching RegressionsJournal of the American Statistical Association, 1978
- Discrete Parameter Variation: Efficient Estimation of a Switching Regression ModelEconometrica, 1978
- Maximum Likelihood from Incomplete Data Via the EM AlgorithmJournal of the Royal Statistical Society Series B: Statistical Methodology, 1977
- Estimating the components of a mixture of normal distributionsBiometrika, 1969