R: MCMC options for SoftBart

Opts {SoftBart}

R Documentation

MCMC options for SoftBart

Description

Creates a list that provides the parameters for running the Markov chain.

Usage

Opts(
  num_burn = 2500,
  num_thin = 1,
  num_save = 2500,
  num_print = 100,
  update_sigma_mu = TRUE,
  update_s = TRUE,
  update_alpha = TRUE,
  update_beta = FALSE,
  update_gamma = FALSE,
  update_tau = TRUE,
  update_tau_mean = FALSE,
  update_sigma = TRUE,
  cache_trees = TRUE
)

Arguments

`num_burn`	Number of warmup iterations for the chain.
`num_thin`	Thinning interval for the chain.
`num_save`	The number of samples to collect; in total, `num_burn + num_save * num_thin` iterations are run.
`num_print`	Interval for how often to print the chain's progress.
`update_sigma_mu`	If `TRUE`, `sigma_mu` is updated, with a half-Cauchy prior on `sigma_mu` centered at the initial guess.
`update_s`	If `TRUE`, `s` is updated using the Dirichlet prior `s \sim D(\alpha / P, \ldots, \alpha / P)` where `P` is the number of covariates.
`update_alpha`	If `TRUE`, `alpha` is updated using a scaled beta prime prior.
`update_beta`	If `TRUE`, `beta` is updated using a normal prior with mean 0 and variance 4.
`update_gamma`	If `TRUE`, gamma is updated using a Uniform(0.5, 1) prior.
`update_tau`	If `TRUE`, the bandwidth `tau` is updated for each tree
`update_tau_mean`	If `TRUE`, the mean of `tau` is updated
`update_sigma`	If `TRUE`, `sigma` is updated, with a half-Cauchy prior on `sigma` centered at the initial guess.
`cache_trees`	If `TRUE`, we save the trees for each MCMC iteration when using the MakeForest interface

Value

Returns a list containing the function arguments.

[Package SoftBart version 1.0.1 Index]