Análise de sobrevivência aplicada no melhoramento genético do cafeeiro
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidade Federal de Viçosa
Abstract
Este estudo aplicou técnicas de análise de sobrevivência a dados do genótipo resultante do cruzamento entre o Híbrido de Timor MG 0357 e o Tupi Amarelo IAC 5162, com foco na produtividade, resistência a doenças e qualidade da bebida, bem como na influência de covariáveis sobre a produção. O evento de interesse foi definido com base na produção média anual das testemunhas Paraíso MG H419-1 e Catuaí Vermelho IAC 144. Foram utilizados o estimador de Kaplan-Meier e os testes de log-rank e Wilcoxon para a comparação entre grupos, enquanto modelos de regressão de Cox, com aproximação da função de verossimilhança parcial, e modelos para dados agrupados avaliaram a relação entre as covariáveis e o tempo até a primeira produção. A adequação dos modelos foi verificada por meio dos resíduos de Schoenfeld e dos resíduos de Cox-Snell; utilizou-se ainda a curva ROC (Receiver Operating Characteristic) para a validação dos modelos para dados agrupados. Os resultados indicaram resistência do cafeeiro à ferrugem, bem como uma associação entre baixo grau de incidência das doenças e maior produtividade. Os testes de log-rank e Wilcoxon detectaram diferenças significativas entre os grupos relacionados a todas as covariáveis analisadas, com destaque para a covariável ciclo de maturação, na qual foram identificados diferentes ciclos produtivos com elevadas capacidades de produção. Os modelos de Cox, ajustados com aproximações da função de verossimilhança parcial, revelaram violações nas pressuposições de riscos proporcionais. Nos modelos discretos, as covariáveis altura da planta, vigor vegetativo, temperatura, precipitação e número de nós nos ramos plagiotrópicos mostraram-se significativas na relação com a produtividade. A curva ROC indicou um desempenho excelente desses modelos. A análise de sobrevivência demonstrou-se eficaz na identificação de fatores associados à produtividade do cafeeiro, destacando-se a aplicação das curvas de Kaplan-Meier e dos testes log-rank e Wilcoxon. Embora os modelos discretos impliquem certa perda de informação, mostraram-se eficientes para a identificação das covariáveis importantes para a produção. Palavras-chave: empates; censura; dados agrupados
This study applied survival analysis techniques to data from the genotype resulting from the cross between the Timor Hybrid MG 0357 and the Yellow Tupi IAC 5162, focusing on productivity, disease resistance and beverage quality, as well as the influence of covariates on production. The event of interest was defined based on the average annual production of the controls Paraíso MG H419-1 and Catuaí Vermelho IAC 144. The Kaplan-Meier estimator and the log-rank and Wilcoxon tests were used for comparison between groups, while Cox regression models, with approximation of the partial likelihood function, and models for grouped data evaluated the relationship between the covariates and the time until the first production. The adequacy of the models was verified by means of Schoenfeld residuals and Cox-Snell residuals; the ROC (Receiver Operating Characteristic) curve was also used to validate the models for grouped data. The results indicated resistance of coffee plants to rust, as well as an association between low incidence of diseases and higher productivity. The log- rank and Wilcoxon tests detected significant differences between the groups related to all the covariates analyzed, with emphasis on the covariate maturation cycle, in which different production cycles with high production capacities were identified. The Cox models, adjusted with approximations of the partial likelihood function, revealed violations in the assumptions of proportional hazards. In the discrete models, the covariates plant height, vegetative vigor, temperature, precipitation and number of nodes in plagiotropic branches were shown to be significant in the relationship with productivity. The ROC curve indicated excellent performance of these models. The survival analysis proved to be effective in identifying factors associated with coffee productivity, with emphasis on the application of the Kaplan-Meier curves and the log-rank and Wilcoxon tests. Although the discrete models imply a certain loss of information, they proved to be efficient in identifying the covariates important for production. Keywords: ties; censorship; grouped data
This study applied survival analysis techniques to data from the genotype resulting from the cross between the Timor Hybrid MG 0357 and the Yellow Tupi IAC 5162, focusing on productivity, disease resistance and beverage quality, as well as the influence of covariates on production. The event of interest was defined based on the average annual production of the controls Paraíso MG H419-1 and Catuaí Vermelho IAC 144. The Kaplan-Meier estimator and the log-rank and Wilcoxon tests were used for comparison between groups, while Cox regression models, with approximation of the partial likelihood function, and models for grouped data evaluated the relationship between the covariates and the time until the first production. The adequacy of the models was verified by means of Schoenfeld residuals and Cox-Snell residuals; the ROC (Receiver Operating Characteristic) curve was also used to validate the models for grouped data. The results indicated resistance of coffee plants to rust, as well as an association between low incidence of diseases and higher productivity. The log- rank and Wilcoxon tests detected significant differences between the groups related to all the covariates analyzed, with emphasis on the covariate maturation cycle, in which different production cycles with high production capacities were identified. The Cox models, adjusted with approximations of the partial likelihood function, revealed violations in the assumptions of proportional hazards. In the discrete models, the covariates plant height, vegetative vigor, temperature, precipitation and number of nodes in plagiotropic branches were shown to be significant in the relationship with productivity. The ROC curve indicated excellent performance of these models. The survival analysis proved to be effective in identifying factors associated with coffee productivity, with emphasis on the application of the Kaplan-Meier curves and the log-rank and Wilcoxon tests. Although the discrete models imply a certain loss of information, they proved to be efficient in identifying the covariates important for production. Keywords: ties; censorship; grouped data
Description
Citation
BELO, Lucas Pereira. Análise de sobrevivência aplicada no melhoramento genético do cafeeiro. 2025. 127 f. Dissertação (Mestrado em Estatística Aplicada e Biometria) - Universidade Federal de Viçosa, Viçosa. 2025.
