Optimal Designs in Plant Breeding Experiments: A Simulation Study Comparing Grid-Plot and Partially Replicated (p-Rep) Design

In studies on plant genetic improvement, it is common to use multi-environmental tests that aim to evaluate and select different lines under different environmental conditions. Another objective is to measure the main quantitative characteristics, such as grain yield, in order to select the best lines under test (test lines). To achieve these and other goals, good experimental design and adequate statistical analysis are essential. In this perspective, the present work evaluated the use of simulated data from spatially optimized designs in multi-environmental tests to compare different partially replicated (p-rep) designs with grid-plot designs considering gain and quality of the genetic material selection. For p-rep, the percentages of repetitions of test lines and the number of standard varieties (checks) were varied. Joint and individual analyses were conducted using linear mixed models with spatial variation in plot errors. Results indicated that p-rep designs were superior to grid-plot designs on relative realized genetic gain, accuracy, and selection probability. When comparing different p-rep designs, it was observed that those without the presence of checks presented the best results in relation to genetic gain. Furthermore, it was observed that for each set of parameters, in which only genetic correlation varied, the greater the genetic correlation, the higher the gain and selection quality. In all locations considered, greater precision was observed in the estimates of variance components when obtained through joint analysis, as expected.