Selection Criterion of Working Correlation Structure for Spatially Correlated Data

To obtain regression parameter estimates in generalized estimation equation modeling, whether longitudinal or spatially correlated data, it is necessary to specify the structure of working correlation matrix. The can be affected by choice this Within spatial statistics, matrix also influences how variability modeled. Therefore, study proposes a new method for selecting matrix, based on conditioning variance-covariance naive. performance evaluated an extensive simulation study, using marginal distributions normal, Poisson, and gamma data. specification semivariogram models, Wendland, Matérn, spherical model families. results reveal that regarding hit rates true simulated proposed criterion resulted better than competing criteria: quasi-likelihood under independence QIC, information CIC, Rotnizky–Jewell RJC. application appropriate selection was shown first-semester average rainfall data 2021 state Pernambuco, Brazil.