Efeito do uso da transformação de box e cox em modelos volumétricos de Pinus Taeda L.

• Autor(a) principal: Monteiro, Kássio Vilela
• Data de Publicação: 2019
• Tipo de documento: Trabalho de conclusão de curso
• Idioma: por
• Título da fonte: Repositório Institucional da UFU
• Texto Completo: https://repositorio.ufu.br/handle/123456789/27947
• topic: Cubagem do tronco; Modelos de regressão; Cubation of the trunk; Regression models; Método de box e cox; Mensuração volumétrica; Box and cox method; Volumetric measurement; CNPQ::CIENCIAS AGRARIAS::RECURSOS FLORESTAIS E ENGENHARIA FLORESTAL

From the knowledge about the application of wood volume in a forest planting, which makes possible the correct planning and ordering of the production. In this sense, no study selects the single and double input volumetric equations for Pinus plants, in order to compare with an application of the Box and Cox method (1964). For an analysis, a total of 158 trees were used, distributed in five municipalities of Planalto Catarinense. The trees were subjected to rigorous cubage of the trunk by the Smalian method. Based on literature searches, eight volumetric models were selected and adjusted. The effect on the accuracy and precision of volumetric variations was compared before and after the application of the Box and Cox (1964) method, used when data normalization is desired. The adjusted performance models were calculated as percentage adjusted coefficient of determination (R2aj.%), Root mean sqaure error (RMSE), Mean absolute percentage error (MAPE), corrected Akaike information criterion (AICc), Weighted Akaike [wi (AIC)] information and graphical analysis of the residues. An evaluation of the models, such as the effectiveness of Box-Cox transformation, was made by the comparative method with the results. The chosen models were single input or nonlinear Husch model and double input or nonlinear Schumacher and Hall model, such models obtained with a higher performance in data predators, both with and without the application of the Box-Cox method. Both models had a weighted Akaike statistic as a great facilitator in choosing.