Considerations on experimental planning and suitability of using statistical tests in Agricultural Sciences

• Autor(a) principal: Cavalcante, Marcelo
• Data de Publicação: 2021
• Outros Autores: Costa, João Gomes da
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Diversitas Journal
• Texto Completo: https://diversitasjournal.com.br/diversitas_journal/article/view/1931

In scientific research, planning and carrying out experiments are common practices in Agricultural Sciences. Although the experimental methods are widely known, it is clear that their assumptions are often not verified, reflecting the low precision and quality of the results. This research aims to: 1) discuss basic concepts of experimental statistics; 2) present the main statistical models and methods used in Agricultural Sciences; 3) evaluate the suitability of using statistical tests in scientific articles. In planning, the choice of design should consider environmental heterogeneity. Hypotheses are tested by the F test, and Anova's assumptions must be met. Data transformation is an efficient tool that is often necessary but little explored. When testing qualitative treatments, a comparison/grouping of means test is used, with the Scott-Knott test being the most robust. When quantitative, linear regression. When successive evaluations are carried out in the same experimental unit, repeated measures analysis is used. The coefficient of variation indicates experimental precision and varies with treatments, variables and environment. The repeatability coefficient estimates the number of evaluations and experiments to be carried out, with a high degree of precision. As for suitability, 32 articles were analyzed and all presented insufficient information on the methodology, compromising the reproducibility of the research. Therefore, experimental planning allows conducting research with savings in time, costs and labor, without compromising the accuracy of the conclusions