Expectation-maximization estimators for incompletely observed data

• Main Author: Vasić Vladimir V.
• Format: Article
• Language: English
• Published: Faculty of Economics, Belgrade 2004-01-01
• Series: Ekonomski Anali
• Subjects: expectation-maximization method; maximum likelihood method; statistical estimators; censored exponential distribution model; mixed distribution model
• Online Access: http://www.doiserbia.nb.rs/img/doi/0013-3264/2004/0013-32640461165V.pdf
• ISSN: 0013-3264

Expectation-maximization is a broadly applicable approach to the iterative computation of maximum likelihood estimates. Each iteration of expectation-maximization method consists of two steps: the expectation step and the maximization step. Expectation-maximization method is useful in a variety of problems where the maximum likelihood estimates are very difficult to find. The basic idea of expectation-maximization method is to relate incomplete data problems to complete data problems where estimation by maximum likelihood method is much simpler.