Regressão quantílica na avaliação da adaptabilidade e estabilidade fenotípica

• Ano de defesa: 2014
• Autor(a) principal: Barroso, Laís Mayara Azevedo
• Orientador(a): Nascimento, Moysés
• Banca de defesa: Silva, Fabyano Fonseca e
• 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: Alfafa - Melhoramento genético; Interação genótipo x ambientes
• Palavras-chave em Inglês: Alfalfa - Breeding; Genotype x Environment interaction
• Área do conhecimento CNPq: CNPQ::CIENCIAS AGRARIAS
• Link de acesso: http://locus.ufv.br/handle/123456789/4076

In plant breeding, when the objective is to select or recommend genotypes to be planted, the study of the interaction between genotype and environment plays a important role. However, this kind of study does not provide detailed information on the behavior of each cultivar due to environmental variations. Thus become necessary to perform analyzes of stability and adaptability for identification and recommendation superior materials in different environments. Although the literature presents several methods for performing analysis of adaptability and stability, none of them take account of the presence of non-normal phenotype, in other words, phenotypic values asymmetric distributions or heavy tails. Thus, if there is the presence of such phenotypic values, the methods can be influenced and the recommendation may be mistaken, that is, the use of such methods cause inadequate estimates that do not reflect the true relationship between the variation environmental and phenotypic response. An interesting solution for treating this problem in a unified way, that is, the presence of outliers or asymmetry is to use the quantile regression (QR). Such methodology, besides the usual regression methods, using the conditional mean to explain the functional relationship between environmental variation and phenotypic response, makes use of conditional quantile functions. This way the QR possible to choose the quantile which best represents the functional relationship of interest in order to naturally cover the lack of normality cited above. Thus, this paper aims to present the methodology of quantile regression, through a detailed discussion of its theorical foundations, demonstrating it by concrete applications, its use in analysis of adaptability and stability, thus providing a easily and accessible material for readers interested in that subject, contributing researchers and those interested in this area. To the technic evaluation symmetric distributions phenotypic values, symmetric with outliers, right asymmetric, right asymmetric with outliers, left asymmetric and left asymmetric with outliers were simulated. Furthermore, we used data from an experiment on dry matter yield of 92 genotypes of alfalfa (Medicago sativa) evaluated in 20 environments. It is suggested that, for symmetrical phenotypic values should be determined if it has outlier, if it has a QR ( τ = 0,50 ) should be used, if not, should be used either Eberhart and Russell methodology (1966) or QR ( τ = 0,50 ). Since the phenotype is asymmetric, with or without the presence of outlier, it uses QR ( τ = 0,25 ) to right asymmetry and QR ( τ = 0,75) to the left asymmetry. According to the results the QR method was efficient for classifying alfalfa genotypes.