Cross-Validation With Small Samples: An Algorithm for Computing Gollob's Estimator

Several years ago, Gollob (1967) presented a novel method for empirically estimating the population cross-validation coefficient of a sample regression equation. The method has been rarely used and has not been very thoroughly investigated. The principal reasons for apparent lack of interest in the method seem to be that investigators were unacquainted with it or that like Campbell (1974) they considered it to be too cumbersome to calculate. The purpose of the present paper was, therefore, twofold: to acquaint (or reacquaint) readers with Gollob's method and to provide a reasonably efficient algorithm for computation.