Site-Pooling Diagnostics for Multisite PLS — site_pooling

Optional post-fit diagnostics for pooled multisite analyses. When observation- or subject-level site labels are available, these summaries quantify how stable a pooled solution is across sites.

Usage

site_pooling_diagnostics(
  x,
  site = NULL,
  spec = NULL,
  progress = FALSE,
  infer = c("none", "score", "full"),
  nperm = 199L,
  nboot = 499L,
  subspace_k = NULL,
  conf = 0.95
)

Arguments

x: A fitted pls_result.
site: Optional site labels. May have length equal to the number of observations or the number of subjects. If omitted, x$site is used.
spec: Optional pls_spec used to fit x. This is required for site-specific and leave-one-site-out reruns when x does not already carry enough input context.
progress: Logical; emit informational messages during reruns.
infer: One of "none", "score", or "full". "score" adds subject-block bootstrap intervals and a subject-block permutation heterogeneity test for sitewise LV score correlations. "full" additionally compares site-specific PLS subspaces against the pooled fit using cumulative RV-style concordance summaries and site-label permutation p-values.
nperm: Number of permutations used by inferential site diagnostics.
nboot: Number of subject-block bootstrap resamples used for sitewise score-correlation intervals.
subspace_k: Maximum number of leading LVs to compare in subspace concordance summaries. Defaults to the number of pooled LVs with permutation p-values $\le 0.05$, or 1 when permutation results are not available.
conf: Confidence level for sitewise bootstrap intervals.

Value

A named list containing pooled observation metadata, sitewise score summaries, sitewise score correlations, site-specific salience similarity, and leave-one-site-out rerun summaries. When infer != "none", the result also contains inferential score-heterogeneity summaries; when infer == "full", it also contains site-subspace concordance tests.