varband_path: Solve main optimization problem along a path of lambda

In hugogogo/varband: Variable Banding of Large Precision Matrices

Description

Compute the varband estimates along a path of tuning parameter values.

Usage

varband_path(S, w = FALSE, lasso = FALSE, K = -1, lamlist = NULL,
  nlam = 60, flmin = 0.01)

Arguments

S	The sample covariance matrix
w	Logical. Should we use weighted version of the penalty or not? If TRUE, we use general weight. If FALSE, use unweighted penalty. Default is FALSE.
lasso	Logical. Should we use l1 penalty instead of hierarchical group lasso penalty? Note that by using l1 penalty, we lose the banded structure in the resulting estimate. And when using l1 penalty, the becomes CSCS (Convex Sparse Cholesky Selection) introduced in Khare et al. (2016). Default value for lasso is FALSE.
K	Integer between 0 and p - 1 (default), indicating the maximum bandwidth in the resulting estimate. A small value of K will result in a sparse estimate and small computing time.
lamlist	A list of non-negative tuning parameters lambda.
nlam	If lamlist is not provided, create a lamlist with length node. Default is 60.
flmin	if lamlist is not provided, create a lamlist with ratio of the smallest and largest lambda in the list. Default is 0.01.

Value

A list object containing

path:

A array of dim (p, p, nlam) of estimates of L

lamlist:

a grid values of tuning parameters

See Also

varband varband_cv

Examples

set.seed(123)
n <- 50
true <- varband_gen(p = 50, block = 5)
x <- sample_gen(L = true, n = n)
S <- crossprod(scale(x, center = TRUE, scale = FALSE))/n
path_res <- varband_path(S = S, w = FALSE, nlam = 40, flmin = 0.03)

hugogogo/varband documentation built on May 17, 2019, 9:12 p.m.