initOmega: Initialize Sigma and Omega using an initial clique of...

In Rmomal/nestor: Network inference from Species counTs with missing actORs

Description

Initialize Sigma and Omega by taking the principal component of cliques as initial value for the hidden variables. The PCA is done on the correaltion matrix, the variance of hidden variables are set to the empirical variances of the pca. The corresponding precision term in omega is set to 1 for identifiability reasons.

Usage

initOmega(Sigma = NULL, cliqueList, cst = 1.1)

Arguments

Sigma	variance-covariance matrix (p x p)
cliqueList	list of found cliques, given as vectorss of nodes indices
cst	small constant for positive definiteness

Value

• Sigma0: initial value of complete covariance matrix ((p+r)x(p+r) matrix)

• K0: initial value of complete precision matrix ((p+r)x(p+r) matrix)

• clique: vector containing the indices of the nodes in the clique

Examples

data=generate_missing_data(n=100,p=10,r=1,type="scale-free", plot=TRUE)
Sigma=data$Sigma
initClique=FitSparsePCA(data$Y,r=1)$cliques
initOmega(Sigma=Sigma, cliqueList=initClique)

Rmomal/nestor documentation built on Dec. 16, 2020, 7:54 p.m.