Decomposition of a mixture into normal components: a review.

The problem of estimating the parameters of a finite mixture of normal distributions has been considered by several researchers in different fields of application. This paper consists of a brief historical background and a comprehensive account of the theoretical aspects behind the principal lines of approach of different techniques of estimation. These include moment- and maximum likelihood estimation, estimation by minimum variance decomposition and graphical techniques. Further, three different methods of estimation are compared on medical data.