Estratégia experimental para avaliação de híbridos de milho

Abstract

In order to obtain genetic progress for corn breeding it is necessary to conduct the experiments as precisely as possible. Among the factors that affect experimental precision, the size of the plots is one of the most important. As corn hybrid architecture is constantly modified, it is important to frequently identify the optimum size of the plots. This experiment was conducted with the objective of proposing the number of plants for evaluating simple corn hybrids, as well as the need of borders in the experiment, and if the distinct plot lines affect hybrid performance. To do this, we used data obtained from an experiment in which we evaluated six simple commercial hybrids, in a randomized block design, with three replicates and four lines of 40 plants each. Harvesting was conducted per georeferenced plant in each line/plot. Grain productivity data (g/plant) were submitted to variance analysis, considering all 160 plants per plot. We also considered the effect of the border in relation to both centrals – useful area – and each line within the plots. Resampling was conducted in order to evaluate the different number of plants within the plot, ranging from two to 159 plants. For each plot size, we conducted 1000 resamplings, performing variance analyses and estimating the accuracy and coincidence in identifying the two best hybrids from the standard plot (160 plants). We conclude that, in the experiments with simple corn hybrids, there is no need for borders, and hybrid performance does not vary in function of plot lines. The accuracy and coincidence estimates for identifying the two best hybrids are large in relation to the employment of 160 plants, based on plots with 50 plants. In general, the variance within the plots is always inferior to that observed between plots. Therefore, to reduce and improve the efficiency of corn breeding programs, the plots for evaluating simple hybrids have no need for more than 50 plants.