R/log_GWishart_pdf.R

In NonDecompGraph: NonDecomposable Graph

# log density function of G-Wishart(b,D,adj):
#
#  p(K) = I(b,D)^{-1} |K|^{(d-2)/2} exp(-trace(K D)/2) using Monte Carlo sample of size N;
#
# Reference: Atay-Kayis and Massam 2005 Biometrika 
#
# INPUT:   b,D,adj,IsDec
# OUTPUT:  log( p(K) ) = log[ I(b,D)^{-1} |K|^{(d-2)/2} exp(-trace(K D)/2) ]; using Monte Carlo sample of size N;

log_GWishart_pdf <- function(K,b,D,adj,N,IsDec)
{
	log_pdf_unnormalized <- ( b - 2 ) / 2 * log( det( K ) ) - sum( diag( D %*% K ) ) / 2;
	if( IsDec ){ # If graph is decomposable 
		log_pdf <- log_pdf_unnormalized + log_hiwishart_InvA_const( makedecompgraph(adj), b, D );    
	} else {  
		log_pdf <- log_pdf_unnormalized - log_GWishart_ud_const_mc( b, D, adj, N );
	}
	return( log_pdf )    
}