adapt_tmcmc_metrop: Simulate adaptive TMCMC algorithm (SCAM, Atchade and RAMA...

In kkdey/tmcmcR: tmcmcR : a R/C++ package for Transformation based Markov Chain Monte Carlo

Description

The function simulates adaptive TMCMC chain of length nsamples using one of the three methods of adaptation - SCAM, Atchade and RAMA (analogously defined as in RWMH case).

Usage

adapt_tmcmc_metrop(target_pdf, base, nsamples, def.scale = 1,
  burn_in = NULL, a_rama = NULL, b_rama = NULL, M_rama = NULL,
  atchade_low = NULL, atchade_high = NULL, method = c("Atchade", "SCAM",
  "Rama"), verb = TRUE)

Arguments

target_pdf	The target density function from which the user wants to generate samples.
base	The starting value of the chain
nsamples	The number of samples to be drawn using the TMCMC algorithm.
def.scale	the initial scale of the proposal chosen, on which we update adaptively. Default is 1.
burn_in	The number of samples assigned as burn-in period. The default burn-in is taken to be one-third of nsamples.
a_rama	The scaling used in RAMA if norm of the chain at current iterate is less than dimension.
b_rama	The scaling used in RAMA if norm of the chain at current iterate is greater than dimension.
M_rama	The bounds on a_rama and b_rama, lower bound is -M_rama, upper bound is M_rama.
atchade_low	The lower bound of scale for Atchade scheme
atchade_high	The upper bound of scale for Atchade scheme
method	The method of adaptation used for generating the chain. May be one of 3 types- SCAM, Atchade and the RAMA methods.
verb	logical parameter, if TRUE the function prints the progress of simulation.

Value

Returns a list containing the following items

chain	The full chain produced by the adaptive TMCMC method with user-chosen method of adaptation.
post.mean	The estimated posterior mean of the adaptive TMCMC chain adjusting for burn-in.

@author Kushal K Dey

@useDynLib tmcmcR @export

kkdey/tmcmcR documentation built on May 20, 2019, 10:39 a.m.