Publications

Selected publications

  1. Zeviani, W. M., Ribeiro, P. J., Jr, Bonat, W. H., Shimakura, S. E., & Muniz, J. A. (2014). The Gamma-count distribution in the analysis of experimental underdispersed data. Journal of Applied Statistics, 41(12), 2616–2626. https://doi.org/10.1080/02664763.2014.922168
  2. Carducci, C. E., Oliveira, G. C. de, Severiano, E. da C., & Zeviani, W. M. (2011). Modelagem da curva de retenção de água de Latossolos utilizando a Equação Duplo Van Genuchten. Revista Brasileira de Ciência do Solo, 35(1), 77–86. https://doi.org/10.1590/s0100-06832011000100007
  3. Zeviani, W. M., Silva, C. A., Carneiro, W. J. de O., & Muniz, J. A. (2012). Modelos não lineares para a liberação de potássio de estercos animais em latossolos. Ciência Rural, 42(10**, 1789–1796. https://doi.org/10.1590/s0103-84782012001000012

Complete CV

Please, visit my profile at any of the services below:




Doctor Thesis

Interpretable parametrizations in nonlinear models

Abstract: One of the advantages of the nonlinear regression models is to have interpretable parameters. In many instances, the parameters of interest, expressed as a function of the model parameters, are quantities subject to investigation. Then comes the concern of how to make inferences about them. For this, the delta method, the Monte Carlo simulation and bootstrap procedures are common alternatives. In addition, a reparametrization can be applied to the model in order to represent these parameters of interest into the model. In addition to improving the interpretation of the presence of the target parameter extends the possibilities regarding the specification of models and statistical inference. The aim of this work is to systematize the procedure to apply reparametrizations. Emphasis was given on nonlinear models considered in applications within the Agricultural Sciences. A list with 17 models reparametrized is provided. In the first case study, the threshold level of defoliation on cotton was evaluated with the following objectives: 1) to propose a model parameter that represents the level of economic damage, 2) evaluate alternative parameterizations through its properties, which considering measures of nonlinearity, 3) apply inference based on likelihood, 4) select a model to describe the relationship between yield and defoliation of cotton in each phenological stage. The reparametrized model showed better properties in phenological stages with pronounced nonlinear relationship. Otherwise the measures of curvature, the correlations of the estimators and likelihood profile plots indicated that a sub-model should be considered. In the second case study, the objective is to verify the e ff ect of sampling position and soil depth on the parameters I (infletion) and S (slope) of the soil water retention curve. For that 1) it was considered ANOVA and 2) weighted ANOVA in each experimental unit compared to 3) using nonlinear mixed effects on a parameterization developed. None of the alternative methods of analysis was superior to model nonlinear mixed effects in the parameterization developed, which had narrower confidence intervals for the parameters and pointed sampling position and depth effect.

Keywords: Likelihood. Delta method. Curvature measures. Mixed effects. van Genuchten.

Banca de Defesa (Aprovada em 13 de maio de 2013):

  • Prof. Dr. Joel Augusto Muniz (Orientador) - UFLA
  • Prof. Dr. Paulo Justiniano Ribeiro Jr (co-orientador) - UFPR
  • Prof. Dr. Júlio da Motta Singer - IME-USP
  • Prof. Dr. Júlio Silvio de Sousa Bueno Filho - UFLA
  • Prof. Dr. Augusto Ramalho de Morais - UFLA

Master Thesis

Evaluation nonlinear regression models on the potassuim kinetic release from organic residues

Abstract: Potassium (K) is required in great amounts by crops, but its availability in Brazilian soils is less than its demand. Its supply can be efficiently made from organic sources when we know the pattern of release. Nonlinear models are appropriate in these situations since then estimate quantities of practical interest, and they have goodness of fit. Although its inferential process is based on asymptotic arguments there are ways to know the nonlinearity intensity. In this work we evaluate the nonlinearity, through the curvatures of Bates & Watts, bias of Box and the least squares estimator sampling properties by simulation study of two nonlinear regression models. These models estimate K readily releasable content, their half-life time release and the release rate of K slow release. The data were obtained from the study over time of the K release from 11 organic waste in combination with 3 soils. The Exponential model gave more precise parameters estimates than the Ratio and closer relations between the asymptotic and simulation results. The Exponential model was more appropriate in terms of inferential and practical application, since by all measures showed lower non-linearity.

Keywords: Curvature of Bates e Watts. Bias of Box. Bootstrap. Half-life time.

Banca de Defesa (Aprovada em 18 de setembro de 2009):

  • Prof. Dr. Joel Augusto Muniz (Orientador) - UFLA
  • Prof. Dr. Carlos Alberto Silva (co-orientador) - UFLA
  • Profa. Dra. Taciana Villela Savian - UFLA
  • Prof. Dr. Luiz Alberto Beijo - UNIFAL-MG