gconcord (version 0.41)

concord: CONvex CORrelation selection methoD

Description

Estimates a sparse inverse covariance matrix from a convex pseudo-likelihood function with lasso L1 penalty

Usage

concord(data, lambda, tol = 1e-05, maxit = 100, save.iterates = FALSE, ...)

Arguments

data

Data matrix with n observations (rows) and p variables (columns)

lambda

Penalty parameter

tol

Convergence threshold

maxit

Maximum number of iterations before termination

save.iterates

Returns iterates if TRUE

...

ignored

Details

Implements the CONCORD method by Khare, Oh and Rajaratnam (2013) http://arxiv.org/abs/1307.5381

Examples

library(mvtnorm)

## True omega
omega <- matrix(0,3,3)
omega[1,2] <- omega[2,3] <- 2.1
omega <- t(omega) + omega
diag(omega) <- 3

sigma <- solve(omega)

## Generate data
set.seed(60)
data <- rmvnorm(100, rep(0,3), sigma)

## Solve
concord(data,2)