Efficiency of augmented designs for selection
Each of two populations of 529 F(,3)-derived lines of oats (Avena sativa L.) in the F(,5') one (W2) developed by interspecific (A. sativa('3) x A. sterilis) hydridization and the other (C2) developed by cultivar hybridization, and five oat cultivars used as check entries, were tested in each of four environments in Iowa. They were evaluated in an augmented randomized complete block (ARCBD), a randomized complete block (RCBD), and a simple lattice design (LD) with a field arrangement such that one set of plots in a replication accommodated all three designs. Each of two replications for a population tested in an environment also was used as the basis for studying the efficiency of three unreplicated experimental designs for plant breeding. These were the NOADJ, in which unadjusted phenotypic values for grain yield (GYLD) were used to judge the worth of the oat lines, and the ARCBD and NOCH unreplicated designs in which the phenotypic value of a line was adjusted by subtracting the corresponding block effect estimated from check and entry means, respectively;To compare the efficiencies of the unreplicated and replicated experimental designs for selection for GYLD, the genotypic yielding abilities of the lines were evaluated by calculating means (called genotypic values) for GYLD from data collected from these lines in previous experiments;Among the replicated experimental designs, LD was the most efficient for controlling intra-site error variance and ARCBD was the least efficient. This superiority of LD, however, was not very great, and there was little difference among the replicated designs when efficiency of selection for GYLD was the judgment criterion. The use of experimental designs, either unreplicated or replicated, that used plot yield adjustments, in general, was not better for selection than were NOADJ or RCBD, respectively, neither of which entailed plot adjustment. All replicated and unreplicated designs were successful for selection according to every criterion used, but making a choice among either unreplicated or among replicated designs for efficiency for selection was relatively unimportant. All criteria used to evaluate replicated and unreplicated designs indicated that success from selection was highly dependent upon the genetic material being used.