twCalcLogDenPar: twCalcLogDenPar

In twDEMC: parallel DEMC

Description

Invokes fLogDen with proposal in a parallel load balanced way.

Usage

twCalcLogDenPar(fLogDen, xProp, logDenCompX = NULL, intResComp = character(0), 
    argsFLogDen = list(), debugSequential = FALSE, remoteDumpfileBasename = NULL, 
    ...)

Arguments

fLogDen	the objective function
xProp	numeric matrix (nCases x nParm) of proposals
logDenCompX	all components of logDensity of xProp (result of fLogDen) colnames must contain intResCompNames rows: number of cases in xProp
intResComp	character vector: names of results components of fLogDen that are used for internal Metropolis decisions
argsFLogDen	arguments passed to fLogDen
debugSequential	see sfFArgsApplyLB
remoteDumpfileBasename	see sfRemoteWrapper
...	further arguments passed to fLogDen

Value

List with the following items

logDen

numeric vector: for each state: the sum of logDens over all components

logDenComp

numeric matrix: return components of fLogDen, one row for each state, columns: components

Author(s)

Thomas Wutzler

See Also

twDEMCBlockInt

Examples

data(twdemcEx1)
xProp <- stackChains(twdemcEx1)[,-1]

data(twLinreg1)
attach( twLinreg1 )
res <- twCalcLogDenPar(logDenGaussian,xProp
	,fModel=dummyTwDEMCModel		### the model function, which predicts the output based on theta 
	,obs=obs			### vector of data to compare with
	,invCovar=invCovar,		### the inverse of the Covariance of obs (its uncertainty)
	thetaPrior = thetaTrue,	### the prior estimate of the parameters
	invCovarTheta = invCovarTheta,	### the inverse of the Covariance of the prior parameter estimates
	xval=xval
)
str(res)

twDEMC documentation built on May 2, 2019, 5:38 p.m.