Gibbs.sample.coeff.gp: Samples new coeffients via Gibbs sampling in a spectral GP...
In spectralGP: Approximate Gaussian Processes Using the Fourier Basis

Description Usage Arguments Details Value Author(s) References See Also

Samples new coefficients via Gibbs sampling in a spectral GP object following the Gibbs sampling scheme of Wikle (2002), which involves an extra variance component (sig2e and a noisy version of the process (z).

1
2
3

## S3 method for class 'gp'
Gibbs.sample.coeff(object, z, sig2e, meanVal=0,
sdVal=1,returnHastings=FALSE, ...)

`object`	A GP object, created by `gp`.
`z`	Vector of values for `z`, the noisy version of the process.
`sig2e`	Noise variance component that distorts `z` as a version of the process.
`meanVal`	Optional mean value for `z`.
`sdVal`	Optional standard deviation value for `z`.
`returnHastings`	Optional argument telling whether to return the logdensity of the proposal for use in a Metropolis-Hastings correction calculation.
`...`	Other arguments.

This function can be used in an MCMC context to take Gibbs samples of the process coefficients, as part of the algorithm of Wikle (2002). The function modifies the GP object, updating the coeff and process components.

The function modifies the GP object, which is essentially a pointer (an R environment in this case), so NULL is returned, unless returnHastings=TRUE.

Christopher Paciorek paciorek@alumni.cmu.edu

Type 'citation("spectralGP")' for references.

gp, propose.coeff.gp, updateprocess.gp

spectralGP documentation built on May 2, 2019, 2:40 a.m.