Nearest-neighbor-adjusted means as a selection criterion within two soybean populations
When the nearest-neighbor adjustment (NNA) reduces the magnitude of the residual mean square, plant breeders have the option of selecting genotypes on the basis of the NNA or unadjusted (UNADJ) means. The actual gain from selection for a specific set of experiments can be compared when selection is based on each criterion. Our objective was to compare the yields of lines selected with the NNA and UNADJ criteria. Three hundred soybean [Glycine max (L.) Merr.] experimental lines were evaluated in six environments. Each environment was considered a selection environment, and the actual yield advance was measured in the other five environments. In 11 out of 12 cases, the lines selected by the NNA and UNADJ criteria were equal in yield when compared in testing environments. Interlocation correlations were similar for both models. Predicted genetic gain was overestimated more often when using the NNA than the UNADJ model. Key words:Glycine max, heritability