Some fundamentals and procedures of parametric, distribution-free and discrete discriminant analysis

The concepts of local, global and hybrid classification rules, which allow a syste¬matic description of the most important parametric and nonparametric classification rules and of their interconnexions and properties, .are presented. Some new procedures are devel¬oped. These procedures include the maximum likelihood ratio rule, empirical globally optimal linear rules, hybrid rules, some of them also for discrete cases