Objectives
- build a provenance manifest and rerun a deterministic calculation
- Preserve a machine-readable provenance record.
- Quantify one uncertainty or sensitivity effect.
- Communicate a result with a figure, caption, and limitation.
Materials and setup
Use Python 3 and reproducibility_records.csv. Record file hash, row count, columns, source URL, release, access date, and provenance level.
Computed check: A 5-record manifest has retention fraction 3/5 = 0.6.
Procedure
- Inspect the raw file and write its schema before filtering.
- Run the task and save query, code version, and output name.
- Perform the perturbation: remove one required provenance field and document the loss
- Compare baseline and perturbed results with units and direction of change.
- Write a caption distinguishing data, transformation, inference, and limitation.
Analysis record
| Record | Required entry |
|---|---|
| Provenance | Record service, query or search string, release, code commit, environment, input hash, output, and access date. |
| Result | Formula or decision rule, value, units, and code reference. |
| Uncertainty | Statistical term plus calibration, selection, or model sensitivity. |
| Validation | Independent recomputation or schema check. |
Deliverables
Submit script or notebook, provenance manifest, labeled figure, 500-word methods memo, and revision note. Full credit requires traceable inputs, correct reasoning, and a specific limitation.
Data provenance
These supplied records are level-3 instructor-provided teaching artifacts, not new observations. Cite the official service and its citation policy.