AvaliaÃÃo de edidas de influÃncia local em modelos lineares espaciais t-Student / Evaluation of local influence measure in t-student linear spacial models

AUTOR(ES)

Naimara Vieira do Prado

FONTE

IBICT - Instituto Brasileiro de Informação em Ciência e Tecnologia

DATA DE PUBLICAÇÃO

28/02/2012

RESUMO

The system of precision agriculture considers that agricultural areas are not homogeneous, so, fertilizers and agricultural lime must be applied at different rates, according to chemical analyzes and physical properties of soil. The use of geostatistics with precision agriculture allows establishing spatial dependence relations among the sampled points. It is possible to construct thematic maps for each soil characteristic by modeling spatial dependence and estimating their parameters using kriging as interpolation method. The presence of outliers can influence the drawing and interpretation of maps and leads to inappropriate rates of application. The t-Student distribution tried to reduce the influence of outliers during the estimation of spatial dependence parameters. The detection of influential points in the studied area provides greater reliability in the use of drawnmaps through local influence analysis and diagnostic techniques as well as an efficient application of inputs. Thus, this study aimed at analyzing the local influence in spatially referenced data, according to t-Student distribution, using the additive perturbation on the variable response (Y ! = Y + ω) and perturbation on the variable response with matrix scale (Y ! = Y +−1/2ω). The estimation of parameters that define spatial variability structure was obtained by the maximum likelihood method with BFGS algorithms and Newton-Raphson support. Sets of simulated and real data were used with contents of phosphorus and potassium in soil, in the univariate case. Soybean yield was used according to the average height of plants and number of pods per plant for the multivariate study. The applied techniques were effective to detect the influential points in the studied area, from a stochastic process with t-Student distribution

ASSUNTO(S)

diagnÃsticos mÃxima verossimilhanÃa variabilidade espacial diagnostics maximum likelihood spatial variability engenharia agricola

Documentos Relacionados

• InfluÃncia local em modelos geoestatÃsticos T-Student com aplicaÃÃes a dados agrÃcolas
• AvaliaÃÃo da deficiÃncia do HormÃnio de crescimento: influÃncia do estado nutricional
• Influência local em um modelo espacial linear da produtividade da soja utilizando distribuição t-Student
• Modelos lineares com erros slash-elÃpticos: uma abordagem em influÃncia local
• Função de acoplamento t-Student assimetrica : modelagem de dependencia assimetrica