Computationally exploitable structure of covariance matrices and generalized convariance matrices in spatial models

Many spatial analyses based on random field models, as varied as kriging, likelihood-based estimation of autocovariance functions, and optimal design of spatial experiments, require the repeated evaluation of a covariance matrix V or a kth-order generalized covariance matrix K and its subsequent inversion. This is generally a formidable computing problem for moderate and large data sets. In this article, however, it is shown that under certain model assumptions, V (or K) possesses one of several types of patterned structure that, if exploited, can significantly reduce the computational burden of the analysis. These patterned structures are characterized, and their implications for matrix evaluation and inversion are considered. The usefulness of the results is illustrated with a soil pH data set

[1]  D. Cook,et al.  Multiple regression in geographical mortality studies, with allowance for spatially correlated errors. , 1983, Biometrics.

[2]  G. Matheron.  Principles of geostatistics , 1963 .

[3]  W. F. Trench.  An Algorithm for the Inversion of Finite Toeplitz Matrices , 1964 .

[4]  N. Cressie.  Kriging Nonstationary Data , 1986 .

[5]  Michael L. Stein,et al.  Minimum norm quadratic estimation of spatial variograms , 1987 .

[6]  Peter K. Kitanidis,et al.  Statistical estimation of polynomial generalized covariance functions and hydrologic applications , 1983 .

[7]  Alex B. McBratney,et al.  Comparison of several spatial prediction methods for soil pH , 1987 .

[8]  G. Matheron.  The intrinsic random functions and their applications , 1973, Advances in Applied Probability.

[9]  Shalhav Zohar,et al.  Toeplitz Matrix Inversion: The Algorithm of W. F. Trench , 1969, JACM.

[10]  Noel A Cressie,et al.  A nonparametric view of generalized covariances for kriging , 1987 .

[11]  K. Ord.  Estimation Methods for Models of Spatial Interaction , 1975 .

[12]  N. Cressie.  Fitting variogram models by weighted least squares , 1985 .

[13]  T. Kailath,et al.  Efficient inversion of Toeplitz-block Toeplitz matrix , 1983 .

[14]  H. Akaike.  Block Toeplitz Matrix Inversion , 1973 .

[15]  P. Delfiner,et al.  Linear Estimation of non Stationary Spatial Phenomena , 1976 .

[16]  K. Mardia,et al.  Maximum likelihood estimation of models for residual covariance in spatial regression , 1984 .

[17]  Julian Besag,et al.  Statistical Analysis of Field Experiments Using Neighbouring Plots , 1986 .

[18]  K. Mardia.  Minimum norm quadratic estimation of components of spatial covariance , 1985 .

[19]  D. Zimmerman.  A random field approach to spatial experiments , 1986 .

[20]  Asymptotic Distributions of Minimum Norm Quadratic Estimators of the Covariance Function of a Gaussian Random Field , 1989 .

[21]  I. J. Good,et al.  The Inverse of a Centrosymmetric Matrix , 1970 .

[22]  Peter K. Kitanidis,et al.  Minimum-variance unbiased quadratic estimation of covariances of regionalized variables , 1985 .

[23]  R. J. Martin.  On the design of experiments under spatial correlation , 1986 .

[24]  Noel A Cressie,et al.  A Graphical Procedure for Determining Nonstationarity in Time Series , 1988 .