Difdtl (version 2.0)

Dpmdtl: Return the result of difference of two precision matrices estimation by d-trace loss with lasso penalty.

Description

Calculate the result of difference of two precision matrices estimation by d-trace loss with lasso penalty, given two sample classes.

Usage

Dpmdtl(X1, X0, lambda = NULL, nlambda = 10, lambda.min.ratio = NULL, 
rho = NULL, shrink = NULL, prec = 0.001, correlation = FALSE, 
tuning = c("none", "aic", "bic", "nbic"))

Arguments

X1

A nXp matrix.

X0

A nXp matrix.

lambda

The tuning parameter of lasso penalty.

nlambda

The number of tuning parameter of lasso penalty for selection.

lambda.min.ratio

rho

The parameter in augmented Lagrange method. The rho here equals the 2*rho in the reference paper.

shrink

prec

correlation

tuning

The method used in the lambda selection.

Value

References

Huili Yuan, Ruibin Xi and Minghua Deng(2015). Differential Network Analysis via the Lasso Penalized D-Trace Loss. http://arxiv.org/abs/1511.09188

Examples

##generate samples
library(MASS)
set.seed(1);
Sigma1 = genp(50,0.2,0.5)
set.seed(1);
Sigma2 = Sigma1+genp1(50,100,0.5)
tdelta = Sigma2-Sigma1
SigmaX<-solve(Sigma1)
SigmaY<-solve(Sigma2)
n<-200
p<-50
X1<-mvrnorm(n,rep(0,p),SigmaX)
Y1<-mvrnorm(n,rep(0,p),SigmaY)
##use of Dpmdtl
dpmdtl<- Dpmdtl(X1,Y1,nlambda=10,tuning="bic")