fitsvar.sb.iso {npsp}

Fit an isotropic Shapiro-Botha variogram model

Description

Fits a ‘nonparametric’ isotropic Shapiro-Botha variogram model by WLS through quadratic programming. Following Gorsich and Genton (2004), the nodes are selected as the scaled roots of Bessel functions (see disc.sb).

Usage

fitsvar.sb.iso(
  esv,
  dk = 4 * ncol(esv$data$x),
  nx = NULL,
  rmax = esv$grid$max,
  min.contrib = 10,
  method = c("cressie", "equal", "npairs", "linear"),
  iter = 10,
  tol = sqrt(.Machine$double.eps)
)

Arguments

esv

pilot semivariogram estimate, a np.svar-class (or svar.bin) object. Typically an output of the function np.svariso.

dk

dimension of the kappa function (dk == 0 corresponds to a model valid in any dimension; if dk > 0, it should be greater than or equal to the dimension of the spatial process ncol(esv$data$x)).

nx

number of discretization nodes. Defaults to min(nesv - 1, 50), where nesv is the number of semivariogram estimates.

rmax

maximum lag considered in the discretization (range of the fitted variogram on output).

min.contrib

minimum number of (equivalent) contributing pairs (pilot estimates with a lower number are ignored, with a warning).

method

string indicating the WLS fitting method to be used (e.g. method = "cressie"). See "Details" below.

iter

maximum number of interations of the WLS algorithm (used only if method == "cressie").

tol

absolute convergence tolerance (used only if method == "cressie").

Details

The fit is done using a (possibly iterated) weighted least squares criterion, minimizing:

WLS(\theta) = \sum_i w_i[(\hat{\gamma}(h_i)) - \gamma(\theta; h_i)]^2.

The different options for the argument method define the WLS algorithm used:

"cressie"

The default method. The procedure is iterative, with w_i = 1 (OLS) used for the first step and with the weights recalculated at each iteration, following Cressie (1985), until convergence:

w_i = N(h_i)/\gamma(\hat{\theta}; h_i)^2,

where N(h_i) is the (equivalent) number of contributing pairs in the estimation at lag h_i.

"equal"

Ordinary least squares: w_i = 1.

"npairs"

w_i = N(h_i).

"linear"

w_i = N(h_i)/h_i^2 (default fitting method in gstat package).

Function solve.QP of quadprog package is used to solve a strictly convex quadratic program. To avoid problems, the Cholesky decomposition of the matrix corresponding to the original problem is computed using chol with pivot = TRUE. If this matrix is only positive semi-definite (non-strictly convex QP), the number of discretization nodes will be less than nx.

Value

Returns the fitted variogram model, an object of class fitsvar. A svarmod object with additional components esv (pilot semivariogram estimate) and fit containing:

u

vector of lags/distances.

sv

vector of pilot semivariogram estimates.

fitted.sv

vector of fitted semivariances.

w

vector of (least squares) weights.

wls

value of the objective function.

method

string indicating the WLS fitting method used.

iter

number of WLS iterations (if method == "cressie").

References

Ball, J.S. (2000) Automatic computation of zeros of Bessel functions and other special functions. SIAM Journal on Scientific Computing, 21, 1458-1464.

Cressie, N. (1985) Fitting variogram models by weighted least squares. Mathematical Geology, 17, 563-586.

Cressie, N. (1993) Statistics for Spatial Data. New York. Wiley.

Fernandez Casal R., Gonzalez Manteiga W. and Febrero Bande M. (2003) Flexible Spatio-Temporal Stationary Variogram Models, Statistics and Computing, 13, 127-136.

Gorsich, D.J. and Genton, M.G. (2004) On the discretization of nonparametric covariogram estimators. Statistics and Computing, 14, 99-108.

Shapiro, A. and Botha, J.D. (1991) Variogram fitting with a general class of conditionally non-negative definite functions. Computational Statistics and Data Analysis, 11, 87-96.

See Also

svarmod.sb.iso, disc.sb, plot.fitsvar.

Examples

# Trend estimation
lp <- locpol(aquifer[,1:2], aquifer$head, nbin = c(41,41),
             h = diag(100, 2), hat.bin = TRUE)
                               # 'np.svariso.corr()' requires a 'lp$locpol$hat' component

# Variogram estimation
esvar <- np.svariso.corr(lp, maxlag = 150, nlags = 60, h = 60, plot = FALSE)

# Variogram fitting
svm2 <- fitsvar.sb.iso(esvar)  # dk = 2
svm3 <- fitsvar.sb.iso(esvar, dk = 0) # To avoid negative covariances...
svm4 <- fitsvar.sb.iso(esvar, dk = 10) # To improve fit...

plot(svm4, main = "Nonparametric bias-corrected semivariogram and fitted models", legend = FALSE)
plot(svm3, add = TRUE)
plot(svm2, add = TRUE, lty = 3)
legend("bottomright", legend = c("NP estimates", "fitted model (dk = 10)", "dk = 0", "dk = 2"),
            lty = c(NA, 1, 1, 3), pch = c(1, NA, NA, NA), lwd = c(1, 2, 1, 1))