Skip to content

Latest commit

 

History

History
108 lines (80 loc) · 6.89 KB

index.md

File metadata and controls

108 lines (80 loc) · 6.89 KB

RMC.BestFit Technical Reference

Documentation home | API traceability | Documentation contract

A useful reading sequence is the model and observation likelihood, the estimator, the derived quantity of interest, and finally its evidence and limitations. Keep three objects distinct: a fitted parameter vector, a posterior distribution over parameters, and a predictive distribution over future observations. A frequency-curve credible interval describes uncertainty about a population quantile; it is not an interval expected to contain 95% of future floods. The notation and parameterization appendices provide a common reference when moving between model families.

Scope

This reference explains the scientific formulations, likelihoods, priors, estimation algorithms, uncertainty propagation, numerical behavior, verification evidence, and C# API for RMC.BestFit 2.0. It is written for statistical reviewers, hydrologic practitioners, and programmatic users. The document-control page identifies the reviewed BestFit source and RMC.Numerics 2.2.0 dependency checkpoint.

Implementation behavior is the authority for API and numerical claims. Primary literature and official standards establish theoretical context. The chapters distinguish verified results from the assumptions and evidence boundaries that govern their interpretation.

Foundations

Distribution models

All fifteen univariate family chapters have completed their source, parameterization, likelihood, API-example, and evidence audit. The Kappa Four zero-shape branch is analytically verified by independent differentiation and CDF-quantile inversion.

Estimation and diagnostics

The estimator and diagnostic chapters distinguish implemented conventions from standard theory. MLE and MAP profiling, GMM specification and covariance, DIC, WAIC, PSIS-LOO, rank-normalized R-hat, bulk and tail ESS, ARWMH adaptation, and NUTS acceptance routing have independent analytical or external-package evidence. Current limitations are stated directly in the affected chapters.

Analysis workflows

Bivariate, coincident-frequency, rating-curve, time-series, and spatial chapters describe current implemented behavior, supported parameter regions, and evidence boundaries.

Reviewer appendices

Peer-review release artifacts

The canonical book manifest defines the order of the Markdown reference. Build the reproducible book-style PDF from the repository root:

powershell -ExecutionPolicy Bypass -File scripts\build-technical-reference-book.ps1

The build first checks that the consolidated bibliography is current, then produces output/pdf/rmc-bestfit-technical-reference.pdf. Intermediate HTML, the browser-produced PDF, and rendered QA images belong under tmp/pdfs/.

The report-specific source checkpoint is recorded in report-metadata.json. SHA256SUMS identifies the retained reviewed PDF bytes. Markdown can evolve after that edition; a source edit does not update the retained PDF or establish a fresh page review. Rebuilding the PDF requires equation, navigation and page-level visual checks before replacing the reviewed artifact.

Publication states

Marker Meaning
technical-reference-status: complete Source-audited, link/citation checked, and all marked C# snippets compile from exact test-fixture regions
technical-reference-status: in-progress Useful working material that is excluded from the public book manifest
No marker Legacy page awaiting contract-based rewrite

The fast TechnicalReferenceDocumentationTests enforce local-link integrity, deleted-namespace rejection, exact compiled snippets, citation anchors on completed pages, and coverage of exported scientific API types in the traceability matrix.

Documentation home | API traceability | Documentation contract