Accuracy of Ranking Individuals in Field Tests of Different Designs: A Computer Simulation

A computer simulation was conducted to investigate the accuracy of ranking individual trees in field tests of different designs. A test population that consists of 900 trees from 45 full-sib families generated by three 6-parent disconnected half-diallels was considered. One incomplete block design with single-tree plots and four complete block designs with 1, 2, 4, and 10-tree row plots were examined. Various narrow-sense heritabilities, site variation patterns (patch sizes and gradient slopes), and two levels of dominant to additive genetic variance ratios (0 and 0.30) were evaluated. Results indicate that the accuracy of ranking depends more on the heritability of the trait and less on the magnitude of the dominant genetic variance, site variation patterns, and field designs. With patchy site variation, differences in ranking accuracy were observed for different designs, but became smaller with higher heritabilities. Impact of environmental gradient on the accuracy of individual ranking was negligible. Incomplete block design with single-tree plots (ICB1) provided the most accurate ranking when heritability was low while complete block design with 2-tree plots (RCB2) appeared to be the best when heritability was high. Large row plot designs were among the least effective in all the simulated scenarios. For traits with medium heritabilities, the statistical efficiencies of ICB1 and RCB with 1 and 2-tree plots are comparable.