Score a cluster-robust ensemble of rCLR, ALR, and ILR-style components — score_cluster_robust

The sample-level score is the row mean of robustly scaled rCLR, ALR, and simple ILR component scores. Cluster-robust confidence intervals are not a property of a single score vector; the benchmark script estimates them by resampling cohort/provenance clusters across held-out folds.

Usage

score_cluster_robust_ensemble(
  expr,
  meta,
  panel,
  cohort_col = "cohort",
  block_col = "provenance_block",
  pseudocount = NULL
)

Arguments

expr: Numeric matrix, samples x features.
meta: Data frame aligned to `expr`.
panel: Character vector of panel features.
cohort_col: Column in `meta` identifying cohorts.
block_col: Column in `meta` identifying provenance blocks.
pseudocount: Optional additive pseudocount before log transform.

Value

Numeric score vector. Attribute `component_scores` contains the finite component matrix.