Testes de multicolinearidade em variáveis morfológicas e produtivas de tomateiro

• Ano de defesa: 2016
• Autor(a) principal: Haesbaert, Fernando Machado
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Santa Maria; BR; Agronomia; UFSM; Programa de Pós-Graduação em Agronomia
• Palavras-chave em Português: Análise multivariada; Correlação; Pressupostos; Olericultura; Multivariate analysis; Correlation; Assumptions; Olericulture; CNPQ::CIENCIAS AGRARIAS::AGRONOMIA
• Link de acesso: http://repositorio.ufsm.br/handle/1/3259

This work presents a comparative study of multicollinearity identification methodologies in multivariate analyzes. Multicollinearity is caused by intense linear relationship between the study variables and can interfere with the interpretation of the results of various multivariate statistical techniques. The objectives of this study were to compare multicollinearity identification methodologies in different settings number of variables, sample size and degree of correlation between variables and identify the most appropriate techniques for the identification of multicollinearity. Morphological and productive variables of an experiment with tomato data were used to generate random samples with multivariate normal distribution in scenario variables numbers and sample sizes in three levels of correlation between variables (low, medium and high). For each scenario were obtained in 1000 multivariate samples and quantified the percentage of presence of multicollinearity statement by the criteria of determining the correlation matrix, condition number and factor inflation variance and the test Farrar and Glauber and Haitovsky. The criteria and the evaluation tests multicollinearity have different results are amended as the number of variable sample size and the degree of correlation between variables. Sample size slightly higher than the number of variables increases the occurrence of multicollinearity. The criteria of the condition number and variance inflation factor is effective in identifying multicollinearity among tomato variables.