On structural approximating multivariate discrete probability distributions
Jǐŕı Grim · Czech digital mathematics library · 1984
The purpose of structural approximating is to develop a flexible parametric model applicable to estimating various types of probability distributions.The approach combines the idea of product approximating with the concept of finite mixtures.To optimize mixtures of product approximations the maximum-likelihood principle is used.An efficient numerical solution is enabled by a convergent iterative procedure.A numerical example previously used by other authors is included to compare different structural approximations.