Consideration of the appropriate variation sources of the statistical model and their impacts on plant breeding

• Autor(a) principal: Melo, Rita Carolina de
• Data de Publicação: 2018
• Outros Autores: Coimbra, Jefferson Luís Meirelles, Trevisani, Nicole, Guidolin, Altamir Frederico, Kavalco, Sydney Antonio Frehner
• Tipo de documento: Artigo
• Idioma: eng
• Título da fonte: Acta Scientiarum. Agronomy (Online)
• Texto Completo: http://www.periodicos.uem.br/ojs/index.php/ActaSciAgron/article/view/39791
• topic: VCU trials; variance analysis; mathematical expectations.; Melhoramento Vegetal

The present work has aimed to assess the consideration of the appropriate variation sources of the statistical model and their impacts on the conclusions plant breeding. The Value for Cultivation and Use test was conducted to assess three common locations (Lages, Ponte Serrada, and Canoinhas) and four non-common locations (Chapecó, Guatambu, Urussanga, and Campos Novos). The grain yields of six bean genotypes were evaluated in order to represent the imbalance between the common and non-common locations. The statistical analysis considered two situations: i) union of the location factors and cultivation years, with a single variation source called environment and ii) decomposition of the mean square values of the two factors, location and year. According to the simplified analysis (environmental variation source), the F test for the genotype factor was highly significant (p = 0.0006). On the other hand, the hypothesis test for the genotype factor was not significant (p = 0.7370) when the decomposition of mean squares was used. The simplified analysis presents some erroneous points, such as the use of a mean residue to test the hypothesis of the genotype factor, since this factor is composed of several sources of variation, and there is no exact F test. However, approximate F tests can be obtained by constructing linear combinations of average squares. This fact notes the relevance of considering the appropriate sources of variation within the statistical model, with a direct impact on the conclusions and recommendations of cultivars with superior performance.