Linear discrimination with symmetrical models (Q1091065)

This paper gives an alternative method to Fisher's method of linear discrimination of an observation into one of two populations. The alternative method is derived by using a new separability measure and the perception criterion function. Just as the Bayesian optimality of the Fisher method is connected with assumptions about normality of the populations, the optimality of the alternative method is connected with some symmetrical models. Analysis of these optimality conditions is discussed and examples of probability models which satisfy the conditions of optimality are given.