mvgraphnorm (version 1.81)

rmvggm: Generate samples from a multivariate gaussian graphical distribution

Description

Generate samples from a multivariate gaussian graphical distribution using two algorithms (1) Castelo & Roverato, 2006, 2009 (2) Kim & van de Wiel, 2008.

Usage

rmvggm(n.samples=100, 
  net.str=NULL, method=c("ipf", "htf", "kim" ), 
  cor=0.0, mean.vector=NULL, m.var=100)

Arguments

n.samples

is a numeric variable which represents total numbers of samples to be generated.

net.str

is a undirected graph of igraph-class object or can be a binary-matrix representation of a graph and should be symmetric.

method

Three methods are used. "ipf", "kim" and "htf". Default method is "ipf".

cor

is a numeric variable or an input symmetric correlation matrix, which represents the average correlation of edge-components of the graph given as an input. This is required for "IPF and "HTF" method. If the input matrix is not positive definite matrix we use Higham (2002) algorithm to convert the matrix into the nearest positive definite matrix.

mean.vector

is a vector which represents the mean of variables of gaussian graphical model. This option is used with "castelo" method.

m.var

number of additional auxilary variables to be used to generate a covariance matrix using "kim" method, deafult value of 'm.var' is 50. This option is not used with 'castelo' method. The minimum value of 'm.var' option should be 0.

Value

returns a list object of "rmvggm" class. The first component of the list object is a matrix of the sampled data. The rows are samples and columns represent the total number of valriables. Second component is a covariance matrix which is used to generate samples from the multivariate normal distribution.

Details

random samples are geneerated from a gaussian graphical model using covariance matrix generated from two algorithms: (1) Castelo & Roverato, 2006, 2009 (2) Kim & van de Wiel, 2008.

References

(1) Kim, K. I. & van de Wiel, M. (2008). Effects of dependence in high-dimensional multiple testing problems. BMC Bioinformatics, 9 (1), 114.

(2) Castelo, R. & Roverato, A. (2006). A robust procedure for Gaussian Graphical Model search from microarray data with p larger than n. J. Mach. Learn. Res., 7, 2621<U+2013>2650.

(3) Higham, Nick (2002) Computing the nearest correlation matrix - a problem from finance; IMA Journal of Numerical Analysis 22, 329<U+2013>343.

See Also

'pseudoinverse', 'qpG2Sigma', 'rmvnorm'

Examples

# NOT RUN {
 	library(mvgraphnorm)
 	g <- barabasi.game(100, directed=FALSE)
 	kk <- rmvggm(net.str=g)
	kk <- rmvggm(net.str=g, n.samples=1000, 
    cor=0, method="ipf")
	summary(kk)
	kk1 <- rmvggm(net.str=g, n.samples=1000, 
    cor=0, method="kim")
	summary(kk1)
###kk$dat is the data where columns 
###represent variables and rows represent samples. 
###Total 1000 samples are generated for each variable
###kk$sigma is a covariance matrix used to generate samples from gaussian graphical model.
###kk$net.str is a binary matrix of graph object given as an input.

# }