Run Split-Half Validation

Performs split-half cross-validation where the sample is randomly split, PLS is run on each half, and the correlation between the resulting weights (u) and loadings (v) is computed.

Usage

pls_splithalf_test(
  stacked_datamat,
  stacked_behavdata = NULL,
  stacked_designdata = NULL,
  num_groups,
  num_subj_lst,
  num_cond,
  method,
  num_split,
  num_outer_perm = 0L,
  clim = 95,
  bscan = NULL,
  meancentering_type = 0L,
  cormode = 0L,
  is_struct = FALSE,
  outer_reorder = NULL,
  inner_subject_perms = NULL,
  progress = TRUE
)

Arguments

stacked_datamat

Stacked data matrix

stacked_behavdata

Behavior data matrix (methods 3-6)

stacked_designdata

Design contrast matrix (methods 2, 5, 6)

num_groups

Number of groups

num_subj_lst

Subjects per group

num_cond

Number of conditions

method

PLS method (1-6)

num_split

Number of split-half iterations

num_outer_perm

Number of outer permutations for significance (includes the unpermuted reference sample as the first permutation, MATLAB missnk_rri_perm_order style).

clim

Confidence level used for optional CI reporting (percent).

bscan

Conditions for behavior block

meancentering_type

Mean-centering type

cormode

Correlation mode

is_struct

Structure PLS flag (do not permute conditions within-subject)

outer_reorder

Optional matrix of outer permutation orders (nrow(stacked_datamat) x num_outer_perm). If provided, overrides the internally generated outer permutation orders (MATLAB permsamp style).

inner_subject_perms

Optional nested list specifying the within-group subject permutations for each outer permutation and split. If provided, it must have shape [[op]][[p]][[g]] where each entry is a permutation of 1:num_subj_lst[g]. This enables deterministic split-half runs for testing.

progress

Show progress

Value

pls_splithalf_result object