BvsMH: Metropolis–Hastings algorithm for Bayesian variables...

In jlin-vt/SML: Statistical Machine Learning

Description

Metropolis–Hastings algorithm for Bayesian variables selection with stochastic search. For sampling scheme, we use switch and swap proposal (Brown, Vannucci, Fearn (1998), JRSS B).

Usage

BvsMH(covariate, Y, nMc, v1, v0, w, gamma0, a, b, phi)

Arguments

covariate	n by p covariate matrix (does not include the column of ones).
Y	continuous response, n by 1.
nMc	number of MCMC iterations.
v1	p by 1 vector, the large s.d. of beta when gamma_j=1.
v0	p by 1 vector, the small s.d. of beta when gamma_j=0.
w	the probablity that gamma_j=1 for each variable.
a	prior parameters for sigma^2. (Inverse Gamma parameters).
b	prior parameters for sigma^2. (Inverse Gamma parameters).
gamma0:initial	value of tao, contains zeros and ones.

Value

sample.gamma all posterior samples of gamma

jlin-vt/SML documentation built on Dec. 5, 2019, 2:05 a.m.