Importance of joint interpretation of the coefficient of variation and determination in agricultural trials: A case study with bean yield

Agriculture & Silviculture

Abstract

Verifying the goodness of an agricultural experiment in analysis of variance (ANOVA) models only through the coefficient of variation (CV) of its variables can lead us to make decisions about imprecise trials (error II) or even take us to the extreme. more complicated to recommend alternatives that in practice do not have significant effects (error I). A statistician that must accompany the CV to assess this goodness and that is not affected by the degrees of freedom of the experimental error as in the latter, is the coefficient of determination (r2). Given this situation, it was proposed to demonstratively detail a practical example to increase the value judgment of a researcher when accepting the results of an experiment, considering for this purpose what is established by the CV and the r2 simultaneously. Performance was analyzed as a variable in an experiment with 24 data, considering a randomized complete block design with six treatments and four repetitions. The assumptions of normality, independence and homoscedasticity of the residuals were checked; the ANOVA was performed and the CV, r2, was calculated, in addition to the repeatability index of the trial. The apparent goodness of the experiment was demonstrated by obtaining a CV of 10.88%, well below the maximum limit given by the literature (30%), although the r2 barely presented a value of 0.47 far from the minimum limit of 0.60; which highlights the importance of looking at CV alongside r2 when examining the robustness of an experimental result.