LOOCV: Leave-one-our-cross-validation statistics.

In geoRExtended: An extension of the geoR package.

Description

A function which get some common summary stats of a leave-one-out-cross-validation.

Usage

LOOCV(Xvalid)

Arguments

Xvalid

The results from either the xvalid function of geoR or the krige.cv function of the gstat package.

Details

A quick function to to summarise the results of LOOCV.

Value

RMSE	Root mean square error of the LOOCV.
ME	Mean error of the LOOCV.
thetaMean	mean theta.
thetaMedian	median theta.

Author(s)

Jason Lessels [e-mail]

Examples

reml <- likfitRcpp(s100, ini=c(0.5, 0.5), fix.nug = FALSE, lik.met = "REML")
# Perform cross validation using gstat and geoR methods.
krige_data <- data.frame(data=s100$data, x=s100$coords[,1], y=s100$coords[,2])
model. <- vgm(psill=reml$sigmasq, "Exp", range=reml$phi, reml$nugget)
system.time(gstatXV <- krige.cv(formula=data~1, locations=~x+y, data=krige_data,model=model.))
system.time(geoRXV <- xvalid(s100, model=reml))
# Getting a quick summary of the cross validation results
LOOCV(geoRXV)
LOOCV(gstatXV)

geoRExtended documentation built on May 2, 2019, 6:14 p.m.