# Dpmdtl: Return the result of difference of two precision matrices... In Difdtl: Difference of Two Precision Matrices Estimation

## Description

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

## Usage

 ```1 2 3``` ```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

 `Dpmdtl` The result of estimation by d-trace loss with lasso penalty. `lambda` The lambda used in the lasso penalty `nlambda` The number of lambda used in the lasso penalty `opt` Number of best lambda chosen by different matrix norms.

Huili Yuan

## 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

 ``` 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15``` ```##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") ```

Difdtl documentation built on Jan. 15, 2017, 8:21 a.m.