Análise de trilha em dados de produção e tecnológicos da cana-de-açúcar

• Ano de defesa: 2010
• Autor(a) principal: Espósito, Deiciana Pagano
• Orientador(a): Peternelli, Luiz Alexandre
• Banca de defesa: Carneiro, Antônio Policarpo Souza , Cecon, Paulo Roberto
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Programa de Pós-Graduação: Mestrado em Estatística Aplicada e Biometria
• Departamento: Estatística Aplicada e Biometria
• País: BR
• Palavras-chave em Português: Correlação; Regressão; Seleção indireta; Análise multivariada; Componentes principais
• Palavras-chave em Inglês: Correlation; Regression; Indirect selection; Multivariate analysis; Principal components
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS
• Link de acesso: http://locus.ufv.br/handle/123456789/4030

In order to quantify the direct and indirect effect through path analysis using phenotypic and genotypic values of yield components - number of stalks per plot, average diameter of stalks and their average length - affecting tons of cane per hectare, data from two experiments of sugarcane were considered. Regarding the productivity of cane per hectare data from two different experiments were obtained from cane plants and ratoon canes, in the beginning of the selection program of sugarcane improvement in the state of Minas Gerais. The following characteristics were evaluated at plot level: the tons of cane per hectare (TCH), as the main variable, and its yield components, number of stalks (NS), mean diameter of stalks (DS) and average length of stalks (LS) as explicative variables. The coefficient of determination were high in all path analyses, which, in turn, indicates that the evaluated components explain, considerably, the variation in TCH. Through the analysis of the direct fenotypic and genotypic effects, NS was the variable that best correlated to TCH in both experiments and stages showing a possibility of obtaining significant gain through indirect selection to TCH by NS. The evaluation of the cause and effect relations among the production components of sugarcane helped to verify the variation across the experiments, which is probably related to the different origins of the families evaluated. In the trail analysis, the parameters are estimated from matrix correlations that may be ill-conditioned by the multicollinearity effect among the involved variables. Due to this fact, the data were evaluated by using ratoon canes obtained from the program of sugarcane improvement at the Federal University of Viçosa in order to compare the method based on ridge regression and the exclusion of variables for main components to estimate the path coefficients under the presence of multicollinearity. Ten plants per plot were used to carry out the analyses on explaining variables Brix (percentage soluble solids), Pol, pH (potential of Hydrogen), RS (reduction sugar), TRS (Total reduction sugar), Cu (copper), Al (aluminum), Mg (magnesium), Ca (Calcium), K (Potassium), aconitic acid, phenolic compounds and the main variable sugarcane juice color (ICUMSA color). The matrix containing correlation obtained from the data were submitted to different methods to have the multicollinearity diagnostic. Under severe multicollinearity, the methods based on ridge regression and in main components presented similar results in the estimation of the path coefficients, causing sensitive reduction in the magnitude of the variance inflation factor associated with the direct and indirect effects of the path analysis. Therefore, in this study, it was possible to identify the variables Al, K and phenolic compounds as the ones that explain the sugarcane juice color. However, the other characters must be taken into account due to their great correlation and low magnitude of the direct effect, making evident the necessity of simultaneous selections of characters, with emphasis on characters that have significant indirect effect. For purposes of improvement, the indirect selection for the ICUMSA color through index selection involving variables such as Brix, Pol, RS, TRS, pH, Cu, Al, Mg, Ca, K, phenolic compounds and aconitic acid is recommended.