摘要：This article focuses on the modeling of agricultural yield data using hierarchical Bayesian models. In recovering the generating process of these data, we consider the temporal, spatial and spatio-temporal relationships pertinent to the prediction and pricing of insurance contracts based on regional crop yields. A county-average yield data set was analyzed for the State of Paraná, Brazil for the period of 1990 through 2002. The choice of the best model from among the several non-nested models considered was based on the posterior predictive criterion. The methodology used in this article proposes improvements in the statistical and actuarial methods often applied to the calculation of insurance premium rates. These improvements are especially relevant to situations of limited data. These conditions are frequently encountered, especially at the individual level.
