Abstract
This study compares the error rates between the linear discriminant function (LDF) and the quadratic discriminant function (QDF) in the case of univariate normal distributions with unequal variances. Asymptotic expansions for the distributions of two discriminant functions are presented. By using these expansions, we obtain the values of the expected error rates and discuss the performance of two discriminant functions. It is shown that the QDF dose not always have better performance than the LDF even if variances are not equal.
