Estimates of variance components and heritability for seed yield in two soybean populations from an incomplete vs. complete block design

Block designs are used to control heterogeneity within an experimental area. Subsequently when conducting yield trials with several genotypes, each block of the randomized complete block design can become heterogeneous. This study was conducted to compare a RGB design with an incomplete block (ICB) design imposed ex post facto. Estimates of the variance components and heritability were compared across three statistical models for two soybean populations. Population 1 contained forty-five F_4:8:9 soybean lines, and Population 2 contained fifty F_4:8:9 soybean lines. Both populations were grown in replicated yield trials at six locations in 1985 and 1986. The ICB design would have been advantageous in twenty-one of forty-two (50%) tests because of the large within block heterogeneity. In 36% of the analyses, the block variance decreased with the incomplete block design, and was equal for the two designs in the remaining 1A%. Additionally, the ICB design resulted, at worst, in an error variance that was not more than 2% greater than the RGB design and, in the best case, was 67% lower than the RCB estimate. Consequently, the error ratios (RCB/ICB) were greater than 100% in 64% of the analyses, less than 100% in 12% of the analyses and equal to 100% in 24% of the analyses. The ICB estimates of heritability were greater than the RCB estimates in 36% of the analyses, equal to the RCB estimates in 40% of the analyses and smaller than the RCB estimates in 24% of the analyses. The ICB design would have resulted in greater precision and accuracy in the estimation of variance components across all three statistical models. Diallel phenotypic correlation coefficients were used to place the six locations into two sets of three locations. One set of three locations would have been adequate for selecting the superior soybean genotypes using either design.