Predição de dados de área referentes ao uso de sementes melhoradas de milho em Moçambique

• Ano de defesa: 2022
• Autor(a) principal: Chipenete, Cláudio Francisco
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Lavras; Programa de Pós-Graduação em Estatística e Experimentação Agropecuária; UFLA; brasil; Departamento de Estatística
• Palavras-chave em Português: Modelos autorregressivos espaciais; Matriz de ponderação espacial; Krigagem; Milho - Melhoramento genético; Melhoramento de sementes; Spatial autoregressive models; Spatial weighing matrix; Kriging; Corn - Genetic improvement; Seed improvement; Estatística
• Link de acesso: http://repositorio.ufla.br/jspui/handle/1/53333

In spatial statistical analysis for area or lattice data, some of the purposes are to evaluate the presence or not of clusters in the study of a given phenomenon, to adjust a linear regression model to the data and to predict unobserved values in some evaluated areas. However, one of the challenges is to define correctly the neighbourhood specifications, which is represented in an spatial weighting matrix, generally called W matrix. The W matrix is a component that always is present in this kind of analysis. However, since there are different criteria in its specification, a non-assertive choice may affect the final results. For example, a wrong choice of this matrix can lead to overestimate or underestimate the parameters of the spatial lag autoregressive model (SAR), which can compromise the prediction. An alternative to solve this prediction problem is to use the kriging predictors, which is used in Geostatistical analysis. Some studies have shown that by using the kriging predictor, it was obtained better results in the prediction. The objectives of this thesis were to evaluate the efficiency to predict missing values by considering the kriging and the SAR models with different spatial weighting matrix. In this study it was used simulated data and maize production data from farmers which used improved maize seeds in Mozambique in 2012. The efficiency of the different prediction methods was evaluated by using the root mean square error (RMSE) and relative efficiency (RE). The results showed us that the proposed kriging predictor presented highest efficiency and the different W matrices had an effect on the predictors efficiency.