Select v0 via K-fold cross-validation

Evaluates each v0 value in the ladder using K-fold cross-validation with Gaussian negative log-likelihood as the scoring criterion. Selects the v0 that minimizes the total CV loss.

Usage

SSJGL_select_v0_cv(
  Y,
  v0s,
  folds = 5,
  seed = 1,
  penalty,
  lambda0,
  lambda1,
  lambda2,
  v1 = 1,
  doubly = FALSE,
  rho = 1,
  a = 1,
  b = 1,
  maxitr.em = 200,
  tol.em = 1e-04,
  maxitr.jgl = 200,
  tol.jgl = 1e-05,
  truncate = 1e-05,
  normalize = FALSE,
  c = 0.1,
  impute = TRUE,
  verbose = TRUE
)

Arguments

Y

List of K data matrices.

v0s

Numeric vector of v0 values to evaluate.

folds

Integer number of CV folds. Default 5.

seed

Integer random seed. Default 1.

penalty, lambda0, lambda1, lambda2, v1, doubly, rho, a, b, maxitr.em, tol.em, maxitr.jgl, tol.jgl, truncate, normalize, c, impute

Arguments passed to ssjgl.

verbose

Logical. If TRUE, prints fold progress. Default TRUE.

Value

A list with elements:

v0_best

The selected v0 value.

i_best

Index of the best v0 in v0s.

cv_score

Named numeric vector of total CV scores per v0.

folds

Number of folds used.

seed

Random seed used.

Examples

if (FALSE) { # \dontrun{
sim <- simulate_ssjgl_data(K = 2, p = 15, n = 100, seed = 42)
cv_res <- SSJGL_select_v0_cv(
  Y = sim$data_list,
  v0s = c(0.1, 0.05, 0.01, 0.005, 0.001),
  folds = 3,
  penalty = "fused",
  lambda0 = 1, lambda1 = 0.5, lambda2 = 0.5,
  maxitr.em = 50, maxitr.jgl = 50,
  normalize = TRUE, impute = FALSE
)
cv_res$v0_best
cv_res$cv_score
} # }