Regional hypsometric relationship models assessed for clonal eucalyptus planting in a Cerrado area

• Autor(a) principal: Andrade, Valdir Carlos Lima de
• Data de Publicação: 2023
• Outros Autores: Schmitt, Thaís, Carvalho, Samuel de Pádua Chaves e, Binotti, Daniel Henrique Breda, Calegario, Natalino
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Ciência Florestal (Online)
• Texto Completo: https://periodicos.ufsm.br/cienciaflorestal/article/view/67995
• topic: Testes estatísticos; Análise de regressão; Modelos mistos; Biometria florestal; Statistical tests; Regression analysis; Mixed models; Forest biometrics

An alternative to reduce the time taken for height measurements of trees is the use of equations, usually obtained from the fitting of local hypsometric models, which require the fit of equations for each stratum that characterize the forest. Therefore, this work was developed with the objective of evaluating regional hypsometric models adjusted to clonal eucalyptus data. A total of 26 regional fixed-effect (FE) models were evaluated, adopting the following statistical criteria: lack of multicollinearity, significance in the estimation of regression coefficients, compliance with regression assumptions, graphical residual analysis and validation test with independent data adopting the mean squared of the prediction residuals, the sum of squares of the relative prediction residuals, the interquartile range between the 1st and 3rd quartiles and multiple linear correlation. After identifying the FE model that most stood out among the others, it was adjusted in the form of a mixed effect (ME) model, by including the random effect of the sampling unit. In this case, to compare with the respective FE model, in addition to the previous criteria, the following were adopted: Akaike information criterion, Bayesian information criterion and maximum likelihood ratio test. It was concluded that there is an inexorable need to consider the adjustment of models with ME, because it stands out above the respective model with FE.