Modifications are proposed for a method of discriminant analysis with missing values described by Chan, Gilman, and Dunn (1976). The modified methods are special cases of techniques for handling missing values in multivariate linear regression. Less stringent conditions for when missing values are missing at random are suggested.