km1Nugget {DiceKriging} | R Documentation |
km1Nugget
is used to fit kriging models with an unknown nugget effect. This function should not be called directly, due to the environments defined in km
to avoid computing twice nxn
matrices. Call km
instead and specify the item nugget
.
km1Nugget(model, envir)
model |
an object of class km . |
envir |
an environment specifying where to assign intermediate values for future gradient calculations. |
An object of class km
.
When a nugget effect is to be estimated, optimization if performed with respect to the correlation parameters and alpha = sigma^2/(sigma^2 + delta^2)
, with sigma^2
the variance of unnoisy part, and delta^2
the noise variance. Thus, alpha
is the proportion of variance explained by the unnoisy part of the process. It can easily be shown that all other parameters ML estimators are function of the correlation parameters and alpha
estimators.
O. Roustant, David Ginsbourger, Ecole des Mines de St-Etienne.
km1Nugget.init
, km
, kmNoNugget
, kmNuggets