Comparison of methods of computing the expected classification error

The paper is concerned with expansions of Okamoto, and Deev and the direct Sitgreaves formula for the expected probability of misclassification by a linear discriminant function. The advantages of Deev's expansion over that of Okamoto is proved.