Statistical power of the North carolina experiment III design in determining the likelihood of success of pedigree breeding programs in selfing plants

In plant pedigree breeding programs, predictions of the likelihood of program success with respect to improvement of a given quantitative character can be made on the basis of the phenotypic difference between parental lines, 2[d], and the additive genetic component, D, of the genetic variance of the quantitative character in the F∞ inbred products of the cross between them. Estimations of D are subject to uncertainty, and as a consequence, breeding program recommendations may result in decision-making errors. The power of the North Carolina Experiment III design for estimating D with regard to such errors was considered as a function of [d] and true D, and of the size and family structure of the experiment. The optimal number of F_2 families and of replicates per family for the estimation of D with high power was determined under various sets of genetic conditions. The sensitivity of the procedure to the underlying true value of D and to the interval between critical decision-making thresholds was examined. Although the North Carolina Experiment III design proved robust under a wide range of genetic conditions, large experiments are required at borderline decision-situations when the difference in mean phenotypic expression between the parental lines crossed is small, while the likelihood of an equivocal decision is high when the difference between the parental lines crossed is large.