MCMCMetropolisGibbs: MCMC using Metroplis Hastings within Gibbs

In OakleyJ/MUCM: Gaussian process emulator methods based on the MUCM toolkit

Description

Generate a Markov Chain of the parameters in the correlation function Using Metropolis-Hastings within Gibbs

Usage

MCMCMetropolisGibbs(inputs, outputs, fn, H, MCMC.iterations, starting.values,
  proposal.sd = 0.1, cor.function, MC.plot = TRUE, ...)

Arguments

inputs	A data frame, matrix or vector containing the input values of the training data.
outputs	A data frame, matrix or vector containing the output values of the training data.
fn	A function used to maximise the negetive log likelihood
H	A matrix of prior mean regressors from the training data
MCMC.iterations	The number of iterations that MCMC should be run for
starting.values	the starting values for which the MCMC can start running
proposal.sd	is the standard deviation of the random walk proposal (default 0.1)
cor.function	Specifies a correlation function used as part of the prior information for the emulator. This package has options of: corGaussian, corMatern2.5, corGaussianPeriodic, corGaussianPeriodic2 and corCombined. One can also specify a user defined function.
MC.plot	If TRUE, produces a trace plot of the MCMC output of log likelihood against the number of iterations. (default=TRUE)
...	additional arguments to be passed on to correlation functions (see corGaussian)

Value

The function returns a list containting the following components:

density.sample	The negetive log likelihood of the MCMC output at the starting value and at each iteration
theta.sample	A matrix of the theta sample at the starting value and at each iteration

Note

Note that this function first calculates the negetive log likelihood of the starting values so returns MCMC.iterations + 1 values.

Author(s)

Originally written by Jeremy Oakley. Modified by Sajni Malde

OakleyJ/MUCM documentation built on May 7, 2019, 9:01 p.m.