getAIESamples: getAIESamples

In petedodd/MCIR: Monte Carlo Inference Routines: A Collection of Generic Algorithms for Bayesian Inference

Description

Function to obtain affine-invariant ensemble samples

Usage

getAIESamples(halfN, nosteps, lnprob, x0, a = 2, eps = 0.01, CPUs = 1,
  dynlib = NULL, echo = FALSE, lgf = NULL)

Arguments

halfN	half the number of walkers in the ensemble
nosteps	the number of steps taken (chainlengths)
lnprob	the loglikelihood function
x0	a vector of initial parameter values
a	alg tuning: todo
eps	alg tuning: todo
CPUs	number of CPUs to use (experimental; not working)
dynlib	if a compiled library is need to evaluate lnprob
echo	if TRUE then more verbose output
lgf	if present, the filename for a logfile to dump information into

Details

todo: provenance and more on arguments

Value

list(chains=list(R1=R2,R2=R2),acc=acc,accv=accv,a=a, llike=list(LL1=LL1,LL2=LL2),lnprob=lnprob) - returned in this form to allow continuation with continueAIESamples. The chains R1 and R2 are matrices with columns for parameteres and rows for iterations.

Author(s)

Pete Dodd

Examples

#Rosenrbock banana function
rosen <- function(x) -(1-x[1])^2 - 100*(x[2] - x[1]^2)^2

#50 walkers for 200 steps
S <- getAIESamples(50,200,arosen,runif(2))
plotAIEchains(S)
SP <- processAIESamples(S,burnin=50)
corplot(SP)

#continue for another 100 steps
S <- continueAIESamples(Y=S,nosteps=100)

petedodd/MCIR documentation built on Jan. 9, 2020, 9:18 a.m.