smoothcholCV: Performs kfolds-cross validation

In adallak/SCPackage: Learning Smooth Cholesky Factors of Large Covariance Matrices

Description

Selects tuning parameters by cross validation according to the likelihood on testing data.

Usage

smoothcholCV(k = 5, X, both.lambda = FALSE, lambda1_seq = NULL,
  lambda2_seq = NULL, max_iter = 50, init.x = NULL, band = NULL,
  n_lambda = 60, pen.type = c("HP", "fused", "l1trend"),
  ABSTOL = 0.001, stand = TRUE)

Arguments

k	Folds used in cross-validation. The default is $k = 5$
X	A n-by-p sample matrix, each row is an observation of th p-dim random vector.
both.lambda	Logical. If TRUE the cross-validation implemented for both lambdas.
lambda1_seq	A vector of non-negative tuning parameters for lambda1 to control sparsity.
lambda2_seq	A vector of non-negative tuning parameters for lambda1 to control smoothness
max_iter	Maximum number of iterations
band	Positive number of subdiagonal to be estimated. If not provided, the algorithm iterates over all subdiagonals.
n_lambda	If lambda1_seq and lambda2_seq is not provided, create a vector of lambdas with length n_lambda. Default is 60.
pen.type	Selects penalty for smoothness.
ABSTOL	The tolerence for convergence
stand	Logical, if TRUE the data will be standardized.

Value

A list object containing

• lambda1_min: Selected value of lambda1 based on cross validation.

• lambda2_min: Selected value of lambda1 based on cross validation.

• L_fit: Estimate of L corresponding to the best fit.

• lambda1_seq: lambda1 grid used in cross validation.

• lambda2_seq: lambda2 grid used in cross validation.

See Also

smoothchol

Examples

set.seed(11)
require(varband)
n = 100
p = 50
L_true = generateL(p = p, case = "c")$L
X = sample_gen(L = L_true, n = n)
L_cv = smoothcholCV(k = 5,X =  X, both.lambda = FALSE, n_lambda = 30, pen.type = "fused")

adallak/SCPackage documentation built on Feb. 7, 2022, 1:58 a.m.