Build an interactive workbench that helps researchers and analysts assess whether a published quantitative claim can be reproduced from supplied data and methods.
Users can upload datasets, research notes, calculation descriptions, tables, and relevant source documents.
The application should:
Extract quantitative claims and map them to the data and calculations intended to support them.
Reproduce supported descriptive statistics, ratios, and comparisons.
Identify missing variables, undocumented transformations, inconsistent sample sizes, and unclear exclusion criteria.
Compare reported results with independently recalculated values.
Allow users to record analysis assumptions, filtering steps, and software versions.
Provide sensitivity analysis for configurable analytical choices.
Maintain a reproducible execution log for supported calculations.
Generate a discrepancy report showing which claims were reproduced, partially reproduced, contradicted, or could not be tested.
Include a data preview, calculation notebook, claim-to-evidence map, and exportable report.
Do not imply that reproducing a numerical result validates the entire study or establishes causation. Preserve the distinction between data errors, methodological choices, and interpretation. Never invent missing observations or undisclosed methods, and make every transformation inspectable.
0 Comments