Comparisons of Prediction Formulas for Maize Hybrids1

Abstract

Three-way and double-cross maize (Zea mays L.) hybrid means were predicted with various formulas involving linear functions of general and specific effects estimated from single-cross diallels. When the efficiencies of these predictors were compared with the correlations of observed and predicted means as the criteria of comparison, the “optimumw eight” predictor was most efficient for hybrids from unselected lines derived from the variety ‘BSK.’ The “optimum weight” predictor did not give the highest correlation for hybrids from a highly selected set of 10 lines, indicating a substantial bias due to epistasis. Even with hybrids from the selected set of lines, the genotype by environmental interaction was a greater potential source of bias than was epistasis unless the trials were grown in a large number of environments. The prediction formula with nonparental single-cross means (Jenkins' method B) was nearly as efficient as the “optimum weight” predictor. We xecommendth at the appropriate single crosses be grown in a preliminary trial at several locations and that the three-way and double-cross performance be predicted from the nonparental singlecross means. A reduced number of selected hybrids per se should then be evaluated over enough environments to enable the breeder to select the superior hybrids for commercial release. Copyright © . .