Tamanho de parcela e número de repetições para avaliação de produtividade de grãos em híbridos e densidades de semeadura de milho

Abstract

Corn (Zea mays L.) is the main cereal grown in the world, mainly due to its high grain productivity and versatility of use in agroindustrial chains. In the face of global climate change and the rising cost of inputs for its production, several breeding programs have allocated resources to the creation of hybrids with higher productivity per plant, responsiveness to agricultural inputs and tolerance to biotic and abiotic stresses. Thus, the grain yield of the crop has been frequently used to discriminate genotypes and to adapt cropping systems. However, the great variability of plot size and number of repetitions used in experiments with these purposes may result in a disparity of results, mainly due to the great influence of these factors on the experimental precision. In this sense, the objective was to determine the appropriate plot size and number of repetitions to determine the grain yield in hybrids of corn sowing densities through resampling with replacement in several scenarios. The experiment was conducted in randomized block design with ten repetitions and using plots of three meters long and five sowing lines spaced at 0,45 m. Thirteen commercial corn hybrids (AS 1555 PRO3, B2401 PWU, B2418 VYHR, B2612 PWU, BG 7061 YHR, DKB 235 PRO3, NS 45 VIP3, NS 73 VIP3, NS 80 VIP3, P2501, P3016 VYHR, P3565 PWU, 30F53 VYHR) were used under the densities of 66667 and 88889 plants ha-1 . During harvest, plots were divided into fifteen linear meters, with each linear meter harvested individually to determine grain yield. We generated 135 scenarios of combinations between plot size (1 to 15 linear meters) and number of repetitions (2 to 10) to perform the statistical analysis. For each scenario, 3000 resamples with replacement were performed and, in each sample, the variance analysis of grain yield was performed. The DMS of the Tukey test was calculated, at 5% probability, for the main effects of hybrids and densities and the unfolding of the interactions between them. The minimum, percentile 2.5, mean, percentile 97.5 and maximum values of each statistic were extracted from the 3000 resamples and the range of the 95% confidence interval was calculated by the difference between the percentiles 97.5 and 2.5. The 3000 resamples performed resulted in little variation in grain yield means among the 135 scenarios, however, reductions in the width of the grain yield confidence interval were obtained with increasing number of repetitions and plot size. Likewise, the increase of repetitions and plot size promoted reductions in the DMS and respective AIC95%, which were more significant when using larger numbers of repetitions. The use of eight repetitions and a plot size equal to four linear meters (14.40 m2 ) allows the evaluation of the grain yield of hybrids at sowing densities with high experimental precision.