scio.refit: Refitted Sparse Column-wise Inverse Operator

Description Usage Arguments Details Value References Examples

View source: R/scio.refit.R

Description

Refitted SCIO Estimators using Penalized Likelihood

Usage

1
scio.refit(S, Omega,thr=1e-4, pkg = c("QUIC","glasso"),...)

Arguments

S

Input covariance matrix of size p by p (symmetric).

Omega

Estimated inverse covariance matrices. Can be a matrix of size p by p from scio or a collection of matrices from sciopath.

thr

Tolerance. Small entries in magnitude (<thr) in Omega are treated as zeros when refitting the precision matrix with the same support as the scio or sciopath outputs. Default 1e-4.

pkg

R packge to be used for refitting. Default QUIC.

...

Additional options passed on to QUIC, which is the only likelihood solver called in the current release. More solvers will be included in future releases.

Details

This implements the refitting procedure discussed in Cai, Liu, and Luo (2011). The current version uses the QUIC solver for the penalized likelihood criterion. More solvers will be added.

Value

A list with one component:

w

Estimated inverse covariance matrix when a single Omega matrix is supplied, or an array of matrices when a 3 dimensional array of Omega is supplied.

References

Weidong Liu and Xi Luo (2012). Fast and Adaptive Sparse Precision Matrix Estimation in High Dimensions. arXiv:1203.3896.

Tony Cai, Weidong Liu, and Xi Luo (2011). A Constrained L1 Minimization Approach to Sparse Precision Matrix Estimation. Journal of the American Statistical Association, 106(494), 594-607.

Examples

1
2
3
4
5
6
7
set.seed(100)
x<-matrix(rnorm(50*20),ncol=4)
s<- var(x)
a<-scio(s, lambda=.01)

require(glasso) 
b <- scio.refit(s, a$w)

scio documentation built on March 3, 2020, 1:07 a.m.