updateSigmaByGibbsSamplingInvchisq: updateSigmaByGibbsSamplingInvchisq

In twDEMC: parallel DEMC

Description

update Variance by sampling from a scaled inverse Chi-square distribution with using prior information on sigma

Usage

updateSigmaByGibbsSamplingInvchisq(theta, argsFUpdateBlock = NULL, 
    fResid, nu0 = 0, sigma20, ...)

Arguments

theta	numeric vector: current state that is used in density function of the block
argsFUpdateBlock	list with possible component intermediate that will be provided to fResid
fResid	function to calculate residuals between prediction and observations, must return a numeric vector
nu0	weight of prior: equivalent number of observations, defaults to no prior (zero obs)
sigma20	prior estimate of variance, needs to be specified only when giving nu0 different from zero
...	further arguemnts to fResid

Details

Gelman et al., 2003, p.51: Estimating variance with known mean.

See the vignette on unknown observation uncertainty.

Value

list with components

accepted	boolean scalar: if step was accepted
xC	numeric vector: components of position in parameter space that are being updated
intermediate	new intermediate state, if provided by fResid

Author(s)

Thomas Wutzler

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