leaveOneOutFun {DiceKriging}

Leave-one-out least square criterion of a km object

Description

Returns the mean of the squared leave-one-out errors, computed with Dubrule's formula.

Usage

leaveOneOutFun(param, model, envir = NULL)

Arguments

param

a vector containing the optimization variables.

model

an object of class km.

envir

an optional environment specifying where to assign intermediate values for future gradient calculations. Default is NULL.

Value

The mean of the squared leave-one-out errors.

Note

At this stage, only the standard case has been implemented: no nugget effect, no observation noise.

Author(s)

O. Roustant, Ecole des Mines de St-Etienne

References

F. Bachoc (2013), Cross Validation and Maximum Likelihood estimations of hyper-parameters of Gaussian processes with model misspecification. Computational Statistics and Data Analysis, 66, 55-69. http://www.lpma.math.upmc.fr/pageperso/bachoc/publications.html

O. Dubrule (1983), Cross validation of Kriging in a unique neighborhood. Mathematical Geology, 15, 687-699.

See Also

leaveOneOut.km, leaveOneOutGrad