Gaussian spatial linear models with repetitions: An application to soybean productivity

Soybean yield in the south region of Brazil was analyzed by using Gaussian spatial linear models to account for the spatial dependence between the observations. Some special cases of Gaussian spatio-temporal linear models were also considered. The data set was collected over five years and we regarded four explanatory variables. Each year was considered to be a realization of the process. We checked the goodness of fit and since we noted the presence of an atypical year with lower productivity than in the other years, we conducted influence diagnostics analysis. Moreover, we considered two appropriate perturbation schemes in the response variable and case weight perturbation. The results showed that the methods were effective for detecting influential observations.