Metropolis: Metropolis Markov chain sample

In baccione-eawag/EawagSchoolTools: Miscellaneous Tools for Eawag Summer School in Environmental Systems Analysis

Description

Calculates a Markov Chain of a probability distribution of a vector of continuous random variables, using the Metropolis algorithm with a normal jump distribution with given standard deviations and correlation structure.

Usage

Markovchain.Metropolis(log.pdf,z.ini,prop.sd,prop.cor=0,
                       sampsize,thin=1,...)

Arguments

log.pdf	function log.pdf(z,...) that calculates the log of the probability density function of the vector of random variables that are to be sampled. log.pdf must be given a single vector z at which the density is to be evaluated. Additional arguments from the call to markovchain.metropolis will be passed on. If the probability density is zero, NA must be returned.
z.ini	vector of values at which the chain is to be started
prop.sd	vector of standard deviations of the proposal distribution
prop.cor	correlation matrix of proposal distribution, or NA if all correlations are zero
sampsize	sample size (length of the chain)
thin	factor with which to thin storage of results (thin=n: only each nth result is returned; this saves memory)
...	additional parameters passed to log.pdf

Details

The log of the probability density of the distribution must be specified as a function log.pdf(z,...), where z is the vector of values for which the density should be evaluated.

Value

Returns a list containing:

z	sample as a matrix, with sample points in rows
log.pdf	vector with log pdf values of the sample
reject.freq	rejection frequency of the jumps
error	error message (empty string if no error occurred)

Author(s)

Peter Reichert <[e-mail]>

baccione-eawag/EawagSchoolTools documentation built on Dec. 19, 2021, 6:38 a.m.