O coeficiente de variação como critério de avaliação em experimentos de milho e feijão

Abstract

The objective of this work was to verify if the Coefficient of Variation (CV) is the most adequate criterion to discard experiment of cultivars evaluation what is the most appropriated value for the CV, and if there are alternatives to increase efficiency. Monte Carlo simulation was chosen. To know what are the most important variables and the limits for simulation, some experimental data from bean and corn crops were considered initially. The following data were considered: number of cultivars, number of blocks, genetic variance, average yield, and the coefficient of variation. All information was processed using Delphi program. Field tests were simulated considering Randomized Complete Blocks Design. Each simulation corresponded to one experiment and the simulated data followed the approximation methodology of the normal distribution function. The number of experiments with coefficient of variation superior and inferior or equal to 20% also was identified, for each simulated configuration, in the 2,000 experiments. As a conclusion, the Coefficient of Variation was not considered a reliable predictor to evaluate the efficiency of experiments for cultivar recommendation. It must be associated to other parameters to make the recommendation of a cultivar more reliable. The repeatability is the parameter that will make possible to determine the criteria to discard experiments for cultivars recommendations.