Predictive zero-mean uniform discrimination

Abstract

A Bayesian solution for discriminating between two zero-mean normal populations with uniform co variance structures is derived and compared with the classical method of Barlett & Please (1963). In a particular application to some genotical data, the Bayesian and classical discriminants, though functionally different. yield similar results in the assignment of new observations. For classificatory purposes, they are incompatible only in a region of low density for both populations.