bGlasso {CARlasso}R Documentation

Gibbs sampler for Bayesian Graphical LASSO and extensions

Description

Main sampling algorithm of Glasso model, note that the mean is in CAR parameterization

Usage

bGlasso(
  data,
  link = "identity",
  r_Omega = 1,
  delta_Omega = 0.01,
  n_iter = 2000,
  n_burn_in = 1000,
  thin_by = 10,
  ns = 1000,
  m = 20,
  emax = 64,
  progress = TRUE,
  verbos = TRUE
)

Arguments

data

A data.frame with all response, row as observations

link

String name of link function? Currently can be "identity" for normal response, "probit" for binary, "log" for counting, "logit" for compositional. Note that when use "logit", the last response will be used as reference.

r_Omega

Hyper-parameter for precision matrix, shape parameter of Gamma. Should be a scalar

delta_Omega

Hyper-parameter for precision matrix, rate parameter of Gamma. Shoule be a scalar

n_iter

Number of sampling iterations (i.e. after burn in) for the Gibbs sampler

n_burn_in

Number of burn in iterations for the Gibbs sampler

thin_by

Final sample was thin by this number

ns

parameter for ARS, maximum number of hulls, only used when link is "log" and "logit"

m

parameter for ARS, initial number of hulls, only used when link is "log" and "logit"

emax

parameter for ARS, tolerance for small values being 0, larger meaning we tolerate smaller values, only used when link is "log" and "logit"

progress

Bool, whether report progress from C++

verbos

Bool, whether show warnings and messages.

Value

A bglasso_out object with elements:

Examples

set.seed(42)
dt <- simu_AR1()
glassores <- bGlasso(data = dt[,1:5])
plot(glassores) 


[Package CARlasso version 0.1.2 Index]