MH: Run a Metropolis-Hastings chain

In nickpoison/nltsa: Nonlinear Time Series Analysis

Description

This function implements a Metropolis-Hastings MCMC algorithm.

Usage

  MH(initial.state, target.logpdf, n.mcmc,
    kernel.rnd = random.walk,
    kernel.logpdf = function(from, to) 0,
    .progress = progress_text)

Arguments

initial.state	Vector of starting point
target.logpdf	Evaluate the log-pdf of the target distribution
n.mcmc	Number of MCMC iterations
kernel.rnd	Sampling from proposal kernel on the parameter space, default to random.walk
.progress	Progress bar to use, from plyr package, default to progress_text

Value

A list with the following components:

trace	Matrix (n.mcmc, length(initial.state)) of accepted states
logpdf	Matrix (n.mcmc, 1) of the log-pdf of the accepted states
proposed	Matrix (n.mcmc, length(initial.state)) of the proposed states
proposedpdf	Matrix (n.mcmc, 1) of the log-pdf of the proposed states
logmhratio	Matrix (n.mcmc, 1) of the logarithm of Metropolish acceptance ratio

See Also

MH, progress_text, random_walk

nickpoison/nltsa documentation built on May 23, 2019, 4:48 p.m.