predict.CRNGP: Gaussian process predictions using a GP object for correlated...

In hetGP: Heteroskedastic Gaussian Process Modeling and Design under Replication

Gaussian process predictions using a GP object for correlated noise (of class CRNGP)

Description

Gaussian process predictions using a GP object for correlated noise (of class CRNGP)

Usage

## S3 method for class 'CRNGP'
predict(object, x, xprime = NULL, t0 = NULL, ...)

Arguments

object	an object of class CRNGP; e.g., as returned by mleCRNGP
x	matrix of designs locations to predict at (one point per row). Last column is for the integer valued seed. If trajectories are considered, i.e., with time, the prediction will occur at the same times as the training data unless t0 is provided.
xprime	optional second matrix of predictive locations to obtain the predictive covariance matrix between x and xprime
t0	single column matrix of times to predict at, if trajectories are considered. By default the prediction is at the same times as the training data.
...	no other argument for this method

Details

The full predictive variance corresponds to the sum of sd2 and nugs. See mleHomGP for examples.

Value

list with elements

• mean: kriging mean;

• sd2: kriging variance (filtered, e.g. without the nugget value)

• cov: predictive covariance matrix between x and xprime

• nugs: nugget value at each prediction location, for consistency with mleHomGP.

hetGP documentation built on Oct. 3, 2023, 1:07 a.m.