Scientific validation utilities — scientific

Deterministic fixtures, independent numerical calculations, PCA subspace comparison, and the bundled scientific-validation suite.

Usage

validation_tolerances()

validation_statistics(
  genotype,
  sample_ids = rownames(genotype),
  variant_ids = colnames(genotype)
)

align_pca_signs(target, reference)

compare_validation_values(
  observed,
  expected,
  tolerance,
  label = "comparison"
)

manual_ibs_matrix(genotype)

compare_pca_subspaces(observed, expected, components = 2L)

manual_population_diversity(genotype, populations)

run_scientific_validation(integration = FALSE, threads = 1L)

Arguments

genotype: Numeric sample-by-variant genotype dosage matrix containing values 0, 1, 2, or NA.
sample_ids: Optional character vector of sample identifiers.
variant_ids: Optional character vector of variant identifiers.
target: Numeric PCA score matrix whose component signs will be aligned.
reference: Numeric reference PCA matrix with matching dimensions.
observed: Observed numeric vector or matrix. For PCA subspace comparison, the observed score or eigenvector matrix.
expected: Expected numeric vector or matrix. For PCA subspace comparison, the independent reference matrix.
tolerance: Maximum permitted absolute numerical difference.
label: Character label attached to the comparison result.
components: Positive integer number of leading PCA components.
populations: Population label for every genotype-matrix row.
integration: Logical; run the complete SNPRelate and independent reference integration checks.
threads: Positive integer number of requested threads. LD validation is capped at four threads.

Value

The individual helpers return tolerance lists, validation tables, similarity matrices, PCA-subspace metrics, or population-diversity tables. run_scientific_validation() returns an overall pass flag and detailed checks.