Spatial variance structures

Kriging models

Kriging models apply to points in an irregular spatial arrangement. They require coordinates of the points so that the distances can be calculated. For example
  • the distance between time points in a one-dimensional longitudinal analysis,
  • the spatial distance between plot coordinates in a two-dimensional field survey analysis.

    Information for determining distances is supplied by the Sortkey argument on the structure line.
  • For one dimensional cases, Sortkey may be
  • the name of a data field containing the coordinate values when it relates to an R structure
  • 0 in which case a vector of coordinates of length order must be supplied after all R and G structure lines.
  • fac(x) when it relates to model term fac(x).
  • In two directions ( IEXP, IGAU, IEUC, AEXP, AGAU, MATn), the Sortkey argument also depends on whether it relates to an R or G structure.
  • For an R structure, use the form rrcc where rr is the number of a data field containing the coordinates for the first dimension and cc is the number of a data field containing the coordinates for the second direction. For example, in the analysis of spatial data, if the x coordinate was in field 3 and the y coordinate was in field 4, the second argument would be 304.
  • fac(x,y) when it relates to model term fac(x,y). For example 
      y ~ mu ... !r ieucv(fac(x,y) .7 1.3)
  • Other Variance structures
  • Large Scale kriging with bisquare function

    Return to index