Plant Pft Statistics

Module: spectralbridge.plant_pft_statistics.

Conditional empirical PFT contrasts; fitted quantile curves are never replicates.

These are associational randomization diagnostics of the supplied sampled plants, conditional on exchangeability within recorded blocks, not causal/transfer tests.

API glossary · Workflow / scientific context

Signatures and documentation below are collected from the current source. Advanced objects are not newly declared stable by inclusion here.

Audience: advanced.

PFTStatisticsConfig dataclass

PFTStatisticsConfig(permutations: int = 9999, bootstraps: int = 999, min_plants: int = 10, seed: int = 20261014, stages: tuple[str, ...] = ('raw', 'corr', 'olioli'))

Explicit, reproducible inferential choices (not acceptance thresholds).

validate

validate()

Audience: advanced.

holm

holm(pvalues)

Holm family-wise adjustment; valid under arbitrary dependence.

Audience: advanced.

permute_within

permute_within(labels, blocks, rng)

Shuffle whole plant labels jointly across bands, preserving block counts.

Audience: advanced.

contrast

contrast(values, labels, blocks, *, permutations, bootstraps, seed)

Block-adjusted binary coefficient with fixed null scaling and max-T.

Mixed blocks only; beta weights block contrasts by nA*nB/(nA+nB). Null scaling uses block-centered variation independent of PFT labels. Whole-curve max |T| and each-band max-T probabilities share one permutation distribution. Bootstrap intervals are pointwise/descriptive, not simultaneous.

Audience: advanced.

run_pft_statistics

run_pft_statistics(comparisons, output, *, config=None, crosswalk=None)

Read verified compact products in place and write resumable statistics.