LOSO cross-validation for rank selection — dkge_cv_rank_loso • dkge

Evaluates candidate ranks by recomputing LOSO bases and measuring explained variance on the held-out subject in the \(K^{1/2}\) metric.

Usage

dkge_cv_rank_loso(
  B_list,
  X_list,
  K,
  ranks,
  Omega_list = NULL,
  ridge = 0,
  w_method = "mfa_sigma1",
  w_tau = 0.3
)

Arguments

B_list: List of qxP subject beta matrices.
X_list: List of Txq subject design matrices.
K: qxq design kernel.
ranks: Integer vector of ranks to evaluate.
Omega_list: Optional list of spatial weights.
ridge: Optional ridge parameter passed to [dkge_fit()].
w_method: Subject-level weighting scheme passed to [dkge_fit()].
w_tau: Shrinkage parameter toward equal weights passed to [dkge_fit()].

Value

List containing the one-SE selection (`pick`), the best rank, and the aggregated score table.

Examples

# \donttest{
toy <- dkge_sim_toy(
  factors = list(cond = list(L = 3)),
  active_terms = "cond", S = 4, P = 15, snr = 5
)
cv <- dkge_cv_rank_loso(toy$B_list, toy$X_list, toy$K, ranks = 1:2)
cv$pick
#> [1] 2
# }