llikGP: Calculate a GP log likelihood

Description

Calculate a Gaussian process (GP) log likelihood or posterior probability with reference to a C-side GP object

Usage

llikGP(gpi, dab = c(0, 0), gab = c(0, 0))
llikGPsep(gpsepi, dab = c(0, 0), gab = c(0, 0))

Arguments

gpi

a C-side GP object identifier (positive integer); e.g., as returned by newGP

gpsepi

similar to gpi but indicating a separable GP object

dab

ab for the lengthscale parameter, see Details

gab

ab for the nugget parameter, see Details

Value

A real-valued scalar is returned.

Details

An “ab” parameter is a non-negative 2-vector describing shape and rate parameters to a Gamma prior; a zero-setting for either value results in no-prior being used in which case a log likelihood is returned. If both ab parameters are specified, then the value returned can be interpreted as a log posterior density. See darg for more information about ab

Examples

Run this code

# NOT RUN {
## partly following the example in mleGP
library(MASS)

## motorcycle data and predictive locations
X <- matrix(mcycle[,1], ncol=1)
Z <- mcycle[,2]

## get sensible ranges
d <- darg(NULL, X)
g <- garg(list(mle=TRUE), Z)

## initialize the model
gpi <- newGP(X, Z, d=d$start, g=g$start)

## calculate log likelihood
llikGP(gpi)
## calculate posterior probability
llikGP(gpi, d$ab, g$ab)

## clean up
deleteGP(gpi)
# }