Canonical quantitative baseline metrics — canonical_baseline

Construct, register, compare, evaluate, summarize, and write deterministic evidence for versioned quantitative scientific baselines.

Usage

new_canonical_baseline_metric(id, dataset_id, analysis, expected,
  comparator = "exact", tolerance = 0, version = "1.0", rationale,
  provenance = list())

validate_canonical_baseline_metric(metric)

new_canonical_baseline_registry(metrics = list())

register_canonical_baseline_metric(registry, metric, replace = FALSE)

validate_canonical_baseline_registry(registry)

compare_canonical_baseline_metric(metric, observed)

evaluate_canonical_baselines(registry, observed, dataset_id = NULL)

canonical_baseline_table(result)

write_canonical_baseline_evidence(result, output_dir)

canonical_baseline_validation(baseline_registry, observe)

canonical_baseline_coverage(registry)

Arguments

id: Stable metric identifier.
dataset_id: Canonical dataset identifier, or optional dataset filter during evaluation.
analysis: Analysis producing the metric.
expected: Expected scalar or vector value.
comparator: One of exact, absolute, relative, set, or distribution.
tolerance: Non-negative numerical acceptance tolerance.
version: Baseline version.
rationale: Scientific justification for the comparator and tolerance.
provenance: Optional named provenance metadata.
metric: A canonical baseline metric.
metrics: Optional list of baseline metrics.
registry: A canonical baseline registry.
replace: Whether an existing metric may be replaced.
observed: Observed value, or a named list keyed by metric identifier.
result: A canonical baseline result.
output_dir: Evidence output directory.
baseline_registry: Canonical baseline registry captured by a suite callback.
observe: Function accepting a descriptor and directory and returning named observations.

Value

Constructors return validated metric or registry objects. Comparisons and summaries return deterministic data frames. Evaluation returns a PopgenVCFCanonicalBaselineResult. The suite adapter returns a validation function. The evidence writer returns normalized paths.