Funções de ligação na determinação da longevidade de sementes de soja
Carregando...
Data
2018-12-17
Autores
Santos, Amanda Rithieli Pereira dos
Título da Revista
ISSN da Revista
Título de Volume
Editor
Universidade Estadual Paulista (Unesp)
Resumo
O estudo da longevidade busca determinar o tempo que as sementes são capazes de manter sua viabilidade quando armazenadas em condições ideais para cada espécie. O parâmetro usual definido como P50 avalia o tempo em que as amostras ou lotes perdem 50% de sua viabilidade. As respostas obtidas nesses estudos são binárias (sementes mortas ou vivas), e devem ser avaliadas por uma distribuição acumulada no tempo, sendo usualmente utilizada a transformação ou função de ligação, Probit. Para espécies com resultados não normais e/ou comportamento não cumulativo negativo o modelo de Probit não possibilitou a obtenção do P50. Assim outras funções simétricas e assimétricas, que já foram aplicadas nas áreas da física e matemática, podem ser uma alternativa. Nesse contexto o objetivo neste trabalho consistiu em avaliar as diferentes funções de ligação simétricas Probit, Logit e Cauchy e assimétricas Aranda-Ordaz, Box-Cox, Complemento Log-Log, Logaritmo, Potência Stukel e Weibull na determinação da longevidade de sementes de soja. Em 10 amostras de sementes de soja obtidas no estado de Goiás, safra de 2016/2017, foram realizados testes de germinação, teor de água, longevidade e teor de óleo. Os dados de porcentagem de protrusão (longevidade) foram transformados pelas funções simétricas e assimétricas com ajustes realizados por regressão linear, os quais propiciaram a determinação do P50 para cada amostra. A qualidade dos ajustes das foram avaliadas pelo coeficiente de determinação corrigido e a porcentagem de P50 incluso no intervalo de interesse. Para avaliar a normalidade dos dados foi utilizado o teste de Komogorov-Smirnov antes e após as transformações. Conclui-se que a funções simétricas de Cauchy-SSF (Capítulo 1) e as funções assimétricas de Aranda-Ordaz, Stukel, Potência e Weibull (Capítulo 2) são mais robustas que Probit na determinação da longevidade de sementes de soja com alta viabilidade inicial, e que as funções de Aranda-Ordaz e Stukel são alternativas para dados com viabilidade inicial abaixo de 80%, principalmente quando os dados não apresentam normalidade.
The study of longevity seeks to determine the time the seeds are able to maintain their viability when stored under ideal conditions for each species. The usual parameter set to P50 evaluates the time the samples or lots lose 50% of their viability. The responses obtained in these studies are binary (dead or living seeds), and should be evaluated by a cumulative distribution in time, being usually used the transformation or binding function, Probit. For species with non-normal results and / or negative non-cumulative behavior, the Probit model did not allow obtaining the P50. Thus other symmetric and asymmetric functions, which have already been applied in the areas of physics and mathematics, may be an alternative. In this context, the objective of this work was to evaluate the different Probit, Logit and Cauchy and asymmetric Aranda-Ordaz, Box-Cox, Log-Log Complement, Logarithm, Stukel and Weibull symmetric binding functions in determining the longevity of soybean seeds. In 10 samples of soybean seeds obtained in the state of Goiás, harvest of 2016/2017, tests of germination, water content, longevity and oil content were performed. Percentage protrusion data (longevity) were transformed by the symmetric and asymmetric functions with adjustments made by linear regression, which allowed the determination of the P50 for each sample. The quality of the adjustments were evaluated by the corrected determination coefficient and the percentage of P50 included in the range of interest. To evaluate the normality of the data, the Komogorov-Smirnov test was used before and after the transformations. It is concluded that the symmetric functions of Cauchy-SSF (Chapter 1) and the asymmetric functions of Aranda-Ordaz, Stukel, Potency and Weibull (Chapter 2) are more robust than Probit in determining the longevity of soybean seeds with high initial viability , and that the Aranda-Ordaz and Stukel functions are alternatives for data with initial viability below 80%, especially when the data are not normal.
The study of longevity seeks to determine the time the seeds are able to maintain their viability when stored under ideal conditions for each species. The usual parameter set to P50 evaluates the time the samples or lots lose 50% of their viability. The responses obtained in these studies are binary (dead or living seeds), and should be evaluated by a cumulative distribution in time, being usually used the transformation or binding function, Probit. For species with non-normal results and / or negative non-cumulative behavior, the Probit model did not allow obtaining the P50. Thus other symmetric and asymmetric functions, which have already been applied in the areas of physics and mathematics, may be an alternative. In this context, the objective of this work was to evaluate the different Probit, Logit and Cauchy and asymmetric Aranda-Ordaz, Box-Cox, Log-Log Complement, Logarithm, Stukel and Weibull symmetric binding functions in determining the longevity of soybean seeds. In 10 samples of soybean seeds obtained in the state of Goiás, harvest of 2016/2017, tests of germination, water content, longevity and oil content were performed. Percentage protrusion data (longevity) were transformed by the symmetric and asymmetric functions with adjustments made by linear regression, which allowed the determination of the P50 for each sample. The quality of the adjustments were evaluated by the corrected determination coefficient and the percentage of P50 included in the range of interest. To evaluate the normality of the data, the Komogorov-Smirnov test was used before and after the transformations. It is concluded that the symmetric functions of Cauchy-SSF (Chapter 1) and the asymmetric functions of Aranda-Ordaz, Stukel, Potency and Weibull (Chapter 2) are more robust than Probit in determining the longevity of soybean seeds with high initial viability , and that the Aranda-Ordaz and Stukel functions are alternatives for data with initial viability below 80%, especially when the data are not normal.
Descrição
Palavras-chave
Glycine max, Funções simétricas, Funções assimétricas, Probit, P50, Viabilidade