You are not currently logged in.
Access JSTOR through your library or other institution:
Computation of Spatial Covariance Matrices
Journal of Computational and Graphical Statistics
Vol. 14, No. 1 (Mar., 2005), pp. 155-167
Published by: Taylor & Francis, Ltd. on behalf of the American Statistical Association, Institute of Mathematical Statistics, and Interface Foundation of America
Stable URL: http://www.jstor.org/stable/27594102
Page Count: 13
You can always find the topics here!Topics: Covariance, Spatial models, Matrices, Rectangles, Mathematical functions, Analytics, Crop reports, Computer software, Expected values, Modeling
Were these topics helpful?See somethings inaccurate? Let us know!
Select the topics that are inaccurate.
Preview not available
In an earlier article, Ghosh derived the density for the distance between two points uniformly and independently distributed in a rectangle. This article extends that work to include the case where the two points lie in two different rectangles in a lattice. This density allows one to find the expected value of certain functions of this distance between rectangles analytically or by one-dimensional numerical integration. In the case of isotropic spatial models or spatial models with geometric anisotropy terms for agricultural experiments one can use these theoretical results to compute the covariance between the yields in different rectangular plots. As the numerical integration is one-dimensional these results are computed quickly and accurately. The types of covariance functions used come from the Matérn and power families of processes. Analytic results are derived for the de Wijs process, a member of both families and for the power models also. Software in R is available. Examples of the code are given for fitting spatial models to the Fairfield Smith data. Other methods for the estimation of the covariance matrices are discussed and their pros and cons are outlined.
Journal of Computational and Graphical Statistics © 2005 American Statistical Association