The Second-Order Stationary Universal Kriging Model Revisited

摘要: Universal kriging originally was developed for problems of spatial interpolation if a drift seemed to be justified model the experimental data. But its use has been questioned in relation bias estimated underlying variogram (variogram residuals), and furthermore universal came considered an old-fashioned method after theory intrinsic random functions developed. In this paper is reexamined together with methods handling inference parameters. The efficiency covariance parameters shown terms bias, variance, mean square error sampling distribution obtained by Monte Carlo simulation three different estimators (maximum likelihood, corrected maximum restricted likelihood). It that unbiased estimates may but number samples small there can no guarantee ‘good’ (estimates close true value) because variance usually large. This problem not specific rather arises any where are inferred from validity evaluated statistically as risk function paper.