criterio.cv: Cross-validation summaries

In geospt: Geostatistical Analysis and Design of Optimal Spatial Sampling Networks

Description

Generate a data frame of statistical values associated with cross-validation

Usage

criterio.cv(m.cv)

Arguments

m.cv

data frame containing: the coordinates of data, prediction columns, prediction variance of cross-validation data points, observed values, residuals, zscore (residual divided by kriging standard error), and fold. If the rbf.tcv function is used, the prediction variance and zscore (residual divided by standard error) will have NA's

Value

data frame containing: mean prediction errors (MPE), average kriging standard error (ASEPE), root-mean-square prediction errors (RMSPE), mean standardized prediction errors (MSPE), root-mean-square standardized prediction errors (RMSSPE), mean absolute percentage prediction errors (MAPPE), coefficient of correlation of the prediction errors (CCPE), coefficient of determination (R2) and squared coefficient of correlation of the prediction errors (pseudoR2)

Examples

library(gstat)
data(meuse) 
coordinates(meuse) <- ~x+y 
m <- vgm(.59, "Sph", 874, .04) 

# leave-one-out cross validation: 
out <- krige.cv(log(zinc)~1, meuse, m, nmax = 40) 
criterio.cv(out)

# multiquadratic function
data(preci)
coordinates(preci)~x+y

# predefined eta
tab <- rbf.tcv(prec~x+y,preci,eta=1.488733, rho=0, n.neigh=9, func="M") 
criterio.cv(tab)

geospt documentation built on May 31, 2017, 2:21 a.m.