gp.realize {cmce}

Draw an unconditional GP realization.

Description

The list l.v is created by makedistlist() from a n x p design matrix on which to generate the unconditional GP realization. The correlation parameters take on values in (0,1) and use the Gaussian correlation model, calculated as rho^d(xi,xj)^2. This helper function is used to create the demonstration data used in the example for calibrate().

Usage

gp.realize(l.v,mu,lambda,rhos,lambdaf=Inf,eps=1e-10,from="")

Arguments

l.v

A list of difference matrices for the design as calculated by makedistlist()

mu

An n x 1 mean vector for the realization

lambda

A scalar quantity for the marginal precision of the drawn realization

rhos

A p x 1 vector of correlation parameters

lambdaf

A scalar quantity denoting the precision of the error (i.i.d.) component of the realization

eps

A fudge factor to help with inverting large correlation matrices

from

Internal use only

Value

An n x 1 vector of the GP realization calculated over the finite locations of the original design matrix.

See Also

getranges scaledesign makedistlist

Examples

library(cmce)

design=matrix(runif(10,1,5),ncol=2,nrow=5)
r=getranges(design)
design=scaledesign(design,r)
l.v=makedistlist(design)
rho=c(0.2,0.01)
muv=rep(0,nrow(design))
lambdav=1
surface=gp.realize(l.v,muv,lambdav,rho)