spm2 (version 1.1.3)

krigecv: Cross validation, n-fold and leave-one-out for kriging methods ('krige')

Description

This function is a cross validation function for kriging methods ('krige') in 'gstat'.

Usage

krigecv(
  longlat,
  trainy,
  trainpredx = NULL,
  validation = "CV",
  cv.fold = 10,
  nmax = 12,
  transformation = "none",
  delta = 1,
  formula = var1 ~ 1,
  vgm.args = ("Sph"),
  anis = c(0, 1),
  alpha = 0,
  block = 0,
  beta,
  predacc = "VEcv",
  ...
)

Value

A list with the following components: me, rme, mae, rmae, mse, rmse, rrmse, vecv and e1; or vecv only

Arguments

longlat

a dataframe contains longitude and latitude of point samples.

trainy

a vector of response, must have length equal to the number of rows in trainx.

trainpredx

a dataframe contains predictive variables of point samples. If longitude and latitude are going to be used as predictive variables, they should also be included but they should be named in names other than 'long' and 'lat'.

validation

validation methods, include 'LOO': leave-one-out, and 'CV': cross-validation.

cv.fold

integer; number of folds in the cross-validation. if > 1, then apply n-fold cross validation; the default is 10, i.e., 10-fold cross validation that is recommended.

nmax

for local kriging: the number of nearest observations that should be used for a kriging prediction or simulation, where nearest is defined in terms of the space of the spatial locations. By default, 12 observations are used.

transformation

transform response variable to normalise the data; can be "sqrt" for square root, "arcsine" for arcsine, "log" or "none" for non transformation. By default, "none" is used.

delta

numeric; to avoid 'log(0)' in "log" transformation. The default is 1.

formula

formula defining response vector and (possible) regressor. an object (i.e., 'variogram.formula') for 'variogram' or a formula for 'krige'. see 'variogram' and 'krige' in the 'gstat' package for details.

vgm.args

arguments for 'vgm', e.g. variogram model of response variable and anisotropy parameters. see 'vgm' in the 'gstat' package for details. By default, "Sph" is used.

anis

anisotropy parameters: see notes 'vgm' in the 'gstat' package for details.

alpha

direction in plane (x,y). see variogram in the 'gstat' package for details.

block

block size. see 'krige' in the 'gstat' package for details.

beta

for simple kriging. see 'krige' in the 'gstat' package for details.

predacc

can be either "VEcv" for vecv or "ALL" for all measures in function pred.acc.

...

other arguments passed on to the function 'gstat'.

Author

Jin Li

References

Li, J., 2013. Predictive Modelling Using Random Forest and Its Hybrid Methods with Geostatistical Techniques in Marine Environmental Geosciences, In: Christen, P., Kennedy, P., Liu, L., Ong, K.-L., Stranieri, A., Zhao, Y. (Eds.), The proceedings of the Eleventh Australasian Data Mining Conference (AusDM 2013), Canberra, Australia, 13-15 November 2013. Conferences in Research and Practice in Information Technology, Vol. 146.

A. Liaw and M. Wiener (2002). Classification and Regression by randomForest. R News 2(3), 18-22.

Pebesma, E.J., 2004. Multivariable geostatistics in S: the gstat package. Computers & Geosciences, 30: 683-691.

Examples

# \donttest{
library(sp)
library(spm)
data(swmud)
data(petrel)

set.seed(1234)
okcv1 <- krigecv(longlat = swmud[, c(1,2)], trainy = swmud[, 3], nmax = 7, transformation =
"arcsine", vgm.args = ("Sph"), predacc = "VEcv")
okcv1

set.seed(1234)
skcv1 <- krigecv(longlat = swmud[, c(1,2)], trainy = swmud[, 3], nmax = 7, transformation =
"arcsine", vgm.args = ("Sph"), predacc = "VEcv", beta = mean(swmud[, 3]))
skcv1

set.seed(1234)
ukcv1 <- krigecv(longlat = swmud[, c(1,2)], trainy = swmud[, 3], nmax = 7, transformation =
"arcsine", formula = var1 ~ long + lat, vgm.args = ("Sph"), predacc = "VEcv")
ukcv1

set.seed(1234)
okcv2 <- krigecv(longlat = swmud[, c(1,2)], trainy = swmud[, 3], validation = "LOO", nmax = 7,
transformation = "arcsine", vgm.args = ("Sph"), predacc = "ALL")
okcv2

set.seed(1234)
n <- 20 # number of iterations,60 to 100 is recommended.
VEcv <- NULL
for (i in 1:n) {
okcv1 <- krigecv(longlat = petrel[, c(1,2)], trainy = petrel[, 5], nmax = 12,
transformation = "arcsine", predacc = "VEcv")
VEcv [i] <- okcv1
}
plot(VEcv ~ c(1:n), xlab = "Iteration for OK", ylab = "VEcv (%)")
points(cumsum(VEcv) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(VEcv), col = 'blue', lwd = 2)

set.seed(1234)
n <- 20 # number of iterations, 60 to 100 is recommended.
measures <- NULL
for (i in 1:n) {
okcv1 <- krigecv(longlat = petrel[, c(1,2)], trainy = petrel[, 3], nmax = 12, transformation =
"arcsine", predacc = "ALL")
measures <- rbind(measures, okcv1$vecv)
}
plot(measures ~ c(1:n), xlab = "Iteration for OK", ylab = "VEcv (%)")
points(cumsum(measures) / c(1:n) ~ c(1:n), col = 2)
abline(h = mean(measures), col = 'blue', lwd = 2)
# }