Skip to contents

The cross-sectional tracer supports one biological layer. Binary metadata use signed rank-biserial association, continuous metadata use Spearman association, and ordered factors use Kendall tau-b against their declared level order. Unordered multilevel factors retain descriptive Kruskal-Wallis epsilon-squared evidence but cannot nominate a component. Binary factor levels declare reference then comparison; character or logical levels use deterministic lexical order. Identifier-like fields and caller-declared non-analytical fields are excluded from target association with reasons.

Usage

associate_metadata(
  std,
  specification = NULL,
  non_analytical_fields = character(),
  dataset_id = NULL,
  n_resamples = 0L,
  seed = 1L,
  exchangeability = c("independent", "not_identifiable")
)

Arguments

std

a Stage-1-complete StateTransitionData

specification

optional draft AnalysisSpecification owning target and nuisance intent for this run

non_analytical_fields

metadata fields to exclude explicitly

dataset_id

optional stable dataset identifier used in confirmed analysis IDs; defaults to metadata(std)$dataset_id when present, then a deterministic input-digest identifier

n_resamples

number of design-preserving association bootstraps; zero retains point/descriptive evidence only

seed

deterministic bootstrap seed

exchangeability

whether independent biological-unit exchangeability is defensible for permutation; "not_identifiable" preserves point evidence but prohibits a search-aware p-value

Value

a validated MetadataAssociationAtlas, or an AssociationAbstention when declared target type and observed metadata are incompatible