Inferência e diagnóstico no modelo espacial linear t-Student reparametrizado: aplicações a dados agrícolas

• Ano de defesa: 2021
• Autor(a) principal: Schemmer, Rosangela Carline
• Orientador(a): Opazo, Miguel Angel Uribe
• Banca de defesa: Opazo, Miguel Angel Uribe , De Bastiani, Fernanda , Galea Rojas, Manuel
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Estadual do Oeste do Paraná; Cascavel
• Programa de Pós-Graduação: Programa de Pós-Graduação em Engenharia Agrícola
• Departamento: Centro de Ciências Exatas e Tecnológicas
• País: Brasil
• Palavras-chave em Português: Algoritmo iterativo; Análise de influência; Geoestatística; Modelagem robusta; Robust modeling
• Palavras-chave em Inglês: Iterative algorithm; Influence analysis; Geostatistics
• Área do conhecimento CNPq: Sistemas biológicos e Agroindustriais
• Link de acesso: http://tede.unioeste.br/handle/tede/5451

This research aims at studying and developing the inference and diagnostic techniques for reparametrized t-Student linear spatial model, with and without replication, applied to data agricultural. Initially, a t-Student distribution reparametrization was carried out, assuming the existence of the second finite moment, with some recurring properties. Analytical expressions were tested for the score function and Fisher matrix of reparameterized distribution. The approach occurred to estimate some parameters, based on the development of an Iterative algorithm. Some criteria were shown to choose the best model by the shape parameter η. A diagnostic analysis was developed to detect the presence of influential observations and possible outliers. These procedures were developed without and with replication. Three papers were written due to the development of this research: diagnosis of influence in a reparameterized t-Student spatial linear model, statistical inference in reparameterized t-Student spatial linear model, reparameterized t-Student spatial linear model with replication. Finally, it can be concluded that reparametrized t-Student linear spatial modeling allows a more robust modeling in the presence of influential observations.