subdiag.thresh

Non-negative penalty parameter.  Controls sparsity level.

<code>(p-1)</code>-by-<code>(p-1)</code> lower-triangular matrix (above diagonal ignored).
<code>w[l,]</code> gives the <code>l</code> weights for g_l.
Defaults to <code>w[l,m]=sqrt(2 * l)/(l - m + 1)</code> for <code>m <= l<="" code=""></=></code>


To solve the unconstrained problem, R is Sigmahat.
To solve constrained problem, R is the current partial residual (excluding A terms).


Implementation of the convex banding procedure (using a
hierarchical group lasso penalty) for covariance estimation that is
introduced in Bien, Bunea, Xiao (2015) Convex Banding of the Covariance
Matrix. Accepted for publication in JASA.

subdiag.thresh: Performs a single pass of BCD on a matrix R.

Description

Usage

Arguments