empirical-integrity

What problem does it solve?

Numbers in manuscript prose must come from pipeline-generated sources via code chunks or inline expressions, never hand-typed, to preserve accuracy and reproducibility.

Core Features & Use Cases

• Enforces end-to-end provenance where every quantitative or methodological claim traces through analysis/raw/ → analysis/scripts/ → analysis/results/ → manuscript (code chunks or inline expressions).
• Supports inline expressions from a project-owned stats dictionary (analysis/manuscript_stats.py with build_stats()) to render numbers in prose, ensuring consistency with the pipeline outputs.
• Provides project-specific deny rules and lifecycle guidance to prevent edits to analysis/results/manuscript_stats.json and related artefacts, preserving data integrity.
• Includes a validation test suite (test_empirical_integrity.py) to catch literals, mismatches, and rendering issues before publication and during manuscript drafting.

Quick Start

Run the project’s empirical-integrity test suite to confirm that every numeric claim in your manuscript traces to analysis/results and the stats dictionary.