vgram calculates an empirical variogram. Note that, by convention, the empirical variogram actually estimates the semivariogram, not the theoretical variogram (which is twice the semivariogram).
formula: A formula describing the relationship between the response and any covariates of interest, e.g., response ~ 1. The variogram is computed for the residuals of the linear model lm(formula, data).
data: A data.frame, SpatialPointsDataFrame, SpatialPixelsDataFrame, or SpatialGridDataFrame object.
coordnames: The columns of data containing the spatial coordinates, provided as a formula (e.g., ~ x + y), column numbers (e.g., c(1, 2)), or column names (e.g., c("x", "y")). The default is NULL, but this must be specified if data is of class data.frame.
nbins: The number of bins (tolerance regions) to use when estimating the sample semivariogram.
maxd: The maximum distance used when calculating the semivariogram. Default is NULL, in which case half the maximum distance between coordinates is used.
angle: A single value (in degrees) indicating the starting direction for a directional variogram. The default is 0.
ndir: The number of directions for which to calculate a sample semivariogram. The default is 1, meaning calculate an omnidirection semivariogram.
type: The name of the estimator to use in the estimation process. The default is "standard", the typical method-of-moments estimator. Other options include "cressie" for the robust Cressie-Hawkins estimator, and "cloud" for a semivariogram cloud based on the standard estimator. If "cloud" is specified, the nbins argument is ignored.
npmin: The minimum number of pairs of points to use in the semivariogram estimator. For any bins with fewer points, the estimate for that bin is dropped.
longlat: A logical indicating whether Euclidean (FALSE) or Great Circle distance (WGS84 ellipsoid) (longlat = TRUE) should be used. Default is FALSE.
verbose: Print computation information. Default is TRUE.
coords: A deprecated argument.
Returns
Returns an evgram object with components:
Details
Note that the directions may be different from other packages (e.g., gstat or geoR packages) because those packages calculate angles clockwise from the y-axis, which is a convention frequently seen in geostatistics (e.g., the GSLIB software library).
Additionally, note that calculating the empirical variogram for the residuals of lm(response ~ 1) will produce identical results to simply computing the sample semivariogram from the original response. However, if a trend is specified (the righthand side of ~ has non-trival covariates), then the empirical variogram of the residuals will differ from that of the original response. A trend should be specified when the mean is non-stationary over the spatial domain.