Description Usage Arguments Details Value Author(s) References Examples
MCMC algorithm for Gaussian Non-decomposable graph determination using PAS algorithm.
1 | GWishart_PAS_DMH(b_prior, D_prior, n, S, C, beta, burnin, nmc)
|
b_prior |
d.f. |
D_prior |
location |
n |
|
S |
|
C |
initial partial covariance matrix |
beta |
|
burnin |
number of MCMC burnin |
nmc |
number of MCMC samples |
Sample C from Gwishart distribution with density: p(C) \propto |C|^{(bG-2)/2} exp(-1/2 tr(C DG)) where (1) bG : d.f. (2) DG: location C: initial partial covariance matrix;
C |
Samples from G-Wishart distribution.
The result is 3D array with |
Sig |
Inverse of C
The result is 3D array with |
adj |
Sampled adjacency matrix.
The result is 3D array with |
Hao Wang ; Sophia Zhengzi Li
Wang and Li (2011) "Efficient Gaussian Graphical Model Determination without Approximating Normalizing Constant of the G-Wishart distribution " http://www.stat.sc.edu/~wang345/RESEARCH/GWishart/GWishart.html
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 | p = 6; tedge = p*(p-1)/2; beta = 0.5;
indmx = matrix(1:(p^2),p,p);
b_prior = 3;
D_prior = diag(p);
n = 3*p;
b_post = b_prior+n;
A = toeplitz(c(1,0.5,matrix(0,1,p-2)));
A[1,p] = 0.4;
A[p,1] = 0.4;
S = solve(A)*n;
D_post = D_prior + S;
adjTrue = 1*(abs(A)>0.001);
burnin = 30; nmc = 100; C = diag(p);
resPAS_DMH <- GWishart_PAS_DMH(b_prior,D_prior,n,S,C,beta,burnin,nmc);
C_save <- resPAS_DMH[[1]]
Sig_save<- resPAS_DMH[[2]]
adj_save<- resPAS_DMH[[3]]
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.