Dimensionamento de parcelas experimentais: proposição de método de estimação e aplicação

Abstract

The use of mathematical models to estimate the size of the experimental plot is a practice that can help in choosing the best method of exploration in order to optimize the resources available to the researcher. The most used mathematical models are the nonlinear functions that make it possible to synthesize the curve of variability, coefficient of variation and experimental precision coefficient, product of the grouping of the simulated plots from the uniformity test in a small set of parameters that allow estimating the ideal size of the experimental plot. Among the methods most used for this purpose are the methods of maximum modified curvature and those of bissegmented regression with plateau response. New methods are being proposed to achieve two basic needs: to estimate coherent results from a practical point of view and to improve the quality of fit. Thus, the objective of the research is to propose the use of the modified logarithmic model with plateau response for estimation of the optimal plot size in rice crop experiment and compare it with the modified maximum curvature methods, the linear model with plateau response and The quadratic model with plateau response, as well as propose the incorporation of the coefficient of experimental precision as a measure of variability in the experimental plot design. The objective of this study was to propose the application of the modified logarithmic model with plateau response in the estimation of sample size for different soybean cultivars. The results show that the use of the experimental precision coefficient in the estimation of the optimum plot size improves the quality of adjustment of the models used. The modified logarithmic model with plateau response can be used to estimate the optimal size of experimental plots. The modified logarithmic model with proposed plateau response is shown to be adequate for estimating the minimum sample size in soybean cultivars. In general, it is suggested to use at least 59 plants per plot to evaluate the yield per plant in experimental plots in the soybean crop.