Assessing the Relative Efficiency of Two Experimental Designs in Soybean Selection.

Intra-cultivar variation in soybean is underexploited due to the general assumption that genotypes in advanced generations are fairly homogeneous. The aim of this study was to investigate the efficiency of completely randomized design (CRD) at ultra-low plant density (12,000 plants ha⁻¹) and randomized block design (RBD) at a seeding rate of 340,000 plants ha⁻¹ in uncoveringthe presence of genetic variation for seed yield potential. To this end, we evaluatedfive soybean cultivars (Saluki 4910, Saluki 4411, Davis, Blackhawk, and Pharaoh)in two environments at Carbondale, IL and assessed the relative efficiency of the two experimental designs based on the coefficient of variation (CV), p-value, experimental power, and mean squared error. The results showed thatthe CV in CRD was 60.59%, almost double than that in RBD (32.67%). The analysis of variance revealed that CRD under ultra-low plant density(p< 0.001) was more effective in revealing significant variation for seed yield compared with RBD (p = 0.4879). Additionally, the power of CRD was 1.000 (σ = 45.211; δ = 38.679), whereas that of RBD was as low as 0.2874 (σ = 438.897; δ = 169.065). Overall, our data suggested that CRD at ultra-low plant density might be more effective in identifying lines with advanced seed yield potential within elite soybean cultivars.