coupled_half_t_mcmc: coupled_half_t_mcmc

In niloyb/CoupledHalfT: What the package does (short line)

coupled_half_t_mcmc

Description

Coupled Blocked Gibbs MCMC for Bayesian shrinkage with half-t priors

Usage

coupled_half_t_mcmc(
  mc_chain_size,
  X,
  X_transpose,
  y,
  a0 = 1,
  b0 = 1,
  std_MH = 0.8,
  rinit = NULL,
  approximate_algo_delta = 0,
  epsilon_eta = 0.5,
  epsilon_xi = Inf,
  epsilon_sigma2 = Inf,
  nrepeats_eta = 1,
  verbose = FALSE,
  preallocate = 100,
  max_iterations = Inf,
  totalduration = Inf,
  xi_fixed = FALSE,
  sigma2_fixed = FALSE,
  lag = 1,
  t_dist_df,
  two_scale = TRUE,
  xi_interval = c(0, Inf)
)

Arguments

mc_chain_size	minimum length of Markov chain
X	n by p matrix
X_transpose	Pre-calculated transpose of X
y	length n vector
a0	positive scalar
b0	positive scalar
rinit	Initial distribution
approximate_algo_delta	approximate MCMC error (non-negative scalar)
epsilon_eta	eta common random numbers/ maximal coupling coupling threshold
epsilon_xi	xi common random numbers/ maximal coupling coupling threshold
epsilon_sigma2	sigma2 common random numbers/ maximal coupling coupling threshold
nrepeats_eta	number of slice sampling steps
verbose	boolean for printing/ not printing run time
preallocate	pre-allocated memory
max_iterations	number of maximum iterations
totalduration	maximum duration
xi_fixed	boolean for fixing / not fixing xi
sigma2_fixed	boolean for fixing / not fixing sigma2
lag	Lag
t_dist_df	degree of freedom v>=1 for Half-t(v)
xi_interval	support of prior distribution of xi

Value

(beta, eta, sigma2, xi) sampled from the half-t prior blocked Gibbs kernel

niloyb/CoupledHalfT documentation built on Aug. 31, 2022, 2:35 a.m.