Abstract
An alternative derivation of the best linear unbiased predictor for the general linear model (in which the errors are nonspherical) is provided. This derivation is simpler than the original version due to Goldberger [3], and may be preferable for instructional purposes.