centered_gibbs3: centered_gibbs3

In kwajiehao/ghInf: Inference for Gaussian hierarchical models

Description

Gibbs sampler for centered 3-level Gaussian hierarchical model according to derived full conditionals

Usage

centered_gibbs3(i, j, k, ndraws, burnin, flat_prior = TRUE, tau = 1,
  tau_a = 1, tau_b = 1, tau_c = 1, sigma_2 = 1)

Arguments

i	number of nodes at level 1
j	number of children nodes in level 2 per node at level 1
k	number of children nodes in level 3 per node at level 2
ndraws	number of samples the user wants to have
burnin	number of samples to throw away at the start as the gibbs sampler warms up
flat_prior	determines whether to use the density with flat prior
tau	variance of the root (level 0)
tau_a	variance for parameters in level 1
tau_b	variance for parameters in level 2
tau_c	variance for parameters in level 3
sigma_2	variance of the observations

Details

Assumptions: - variances are constant for parameters within the same level - all observations y_ijk are equal to 0 - the mean of the root parameter B is mu = 0 - assume a naive sampler where the variances are not updated

Value

list of means and the samples

Examples

i <- 2
j <- 3
k <- 2
ndraws <- 10000
burnin <- 1000
centered_gibbs3(i = i, j = j, k = k, ndraws = ndraws, burnin = burnin)

kwajiehao/ghInf documentation built on May 7, 2019, 10:58 a.m.