| Sumario: | A study was conducted to demonstrate the feasibility, flexibility, and simplicity of the eigenvector spatial filtering approach embedded in a semiparametric statistical framework. Data on cancer mortality for the 508 State Economic Areas in the U.S. were analyzed. Findings suggested that this framework allows visualizing the logical components of spatial processes, deals well with model misspecification, and can be used to perform spatial predictions and in-depth residual analysis. Findings indicated that the search strategy of minimizing the residual spatial autocorrelation offers an intuitively appealing objective function that results in suitable and more parsimonious models for the stochastic spatial signal.
|
|---|