Spatial Cross-Sectional Credibility Models With General Dependence Structure Among Risks
Credibility models with general dependence structure among risks and conditional spatial cross-sectional dependence are studied in this project. Predictors of future losses for a Buhlmann-type credibility model under both types of dependence are derived by minimizing the quadratic loss function, and this is further extended to Buhlmann-Straub and regression credibility model formulations. Non-parametric estimators of structural parameters of various models under a spatial statistics context are also considered especially for the case of equal unconditional means. An example with crop insurance losses is studied to illustrate the use of predictors and estimators proposed in this project. Finally, the performance of the predictors and estimators are evaluated in a simulation study.