Fit SSJGL at best v0 with bootstrap confidence intervals

Fits the model on the full data at a single v0 value, then computes bootstrap confidence intervals for partial correlations using the percentile method.

Usage

SSJGL_final_with_pcor_CI(
  Y,
  v0_best,
  B = 200,
  ci_level = 0.95,
  seed = 1,
  penalty = "fused",
  lambda0,
  lambda1,
  lambda2,
  v1 = 1,
  doubly = FALSE,
  rho = 1,
  a = 1,
  b = 1,
  maxitr.em = 500,
  tol.em = 1e-04,
  maxitr.jgl = 500,
  tol.jgl = 1e-05,
  truncate = 1e-05,
  normalize = FALSE,
  c = 0.1,
  impute = TRUE,
  verbose = TRUE
)

Arguments

Y

List of K data matrices.

v0_best

Scalar v0 value to use.

B

Integer number of bootstrap samples. Default 200.

ci_level

Numeric confidence level (e.g., 0.95). Default 0.95.

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 bootstrap progress. Default TRUE.

Value

A list with elements:

v0_best

The v0 value used.

fit

The ssjgl fit on the full data.

theta_hat

List of K estimated precision matrices.

pcor_hat

List of K partial correlation matrices.

CI_lower

List of K lower CI bound matrices (pcor).

CI_upper

List of K upper CI bound matrices (pcor).

boot_pcor

List of K arrays (p x p x B) of bootstrap pcors.

B

Number of bootstrap samples.

ci_level

Confidence level used.

seed

Random seed used.

Examples

if (FALSE) { # \dontrun{
sim <- simulate_ssjgl_data(K = 2, p = 15, n = 100, seed = 42)
boot_res <- SSJGL_final_with_pcor_CI(
  Y = sim$data_list,
  v0_best = 0.01,
  B = 20,
  penalty = "fused",
  lambda0 = 1, lambda1 = 0.5, lambda2 = 0.5,
  normalize = TRUE, impute = FALSE
)
# Edges where 95% CI excludes zero
sig <- (boot_res$CI_lower[[1]] > 0) | (boot_res$CI_upper[[1]] < 0)
diag(sig) <- FALSE
sum(sig[upper.tri(sig)])
} # }