Your starting point
Prerequisites: Use the ideas and vocabulary from lesson 47; revisit its worked examples if needed.
Definition: A reproducible analysis records enough inputs, operations and assumptions for a result to be recreated or checked. It requires more than a final chart: source versions, definitions, transformations, filters, formulas and relevant software conventions all matter.
Preserve the evidence chain
Keep an unchanged source when permitted, a cleaned working copy and a transformation record. Note the source version and extraction date separately from the observation period. Record why records were removed or values replaced. A later analyst should not have to infer your cleaning decisions by comparing two unexplained files.
Separate settings from transformed data
The academy exports both project backups and settings files. A project backup includes the working dataset and saved choices. A settings file contains type, filter and chart choices but does not replay destructive cleaning. To reproduce the same result from settings, pair them with the appropriate cleaned CSV and transformation notes. Label these files so their relationship is clear.
Test the handoff
Open the exported material in a fresh session or ask another person to follow the steps. Reconcile row counts, grouped totals and plotted values. Note numerical conventions such as sample standard deviation and the quartile rule. If a result depends on an external AI response, preserve the non-sensitive prompt and the independent verification, rather than assuming the same model will reproduce identical prose later.
Two worked examples
A complete handoff
A learner exports cleaned-sales.csv, analysis-settings.json and analysis-report.txt. The report states that one confirmed repeated row was removed, revenue equals units times price, and the chart shows monthly sums. Another reader can inspect the CSV, apply the settings and check the totals against the documented rule.
An incomplete handoff
A chart image says sales improved, but includes no period, source, unit or filter. Even if its pixels are preserved perfectly, another analyst cannot reconstruct the calculation. Adding the title alone is insufficient; the input records and transformation choices must be traceable.
Where this becomes useful
A community organization revisits a report months later. A compact reproducibility bundle prevents staff turnover or a cleared browser from turning a useful analysis into an unverifiable screenshot.
Make an attempt
Complete a small workbench analysis, export the cleaned CSV, settings, report and project backup, then restore them in a fresh academy session. Compare row count and two chart values. Explain which file contains data and which contains only choices.
Misconception check
Claim: Saving a chart image preserves the analysis.
Inspect the correction
It preserves a display, not necessarily the data, definitions or operations required to verify it.
Independent application
Prepare a reproducibility bundle for a synthetic project. Include source context, cleaned data, settings, formulas, transformation notes and a scoped finding. Restore it independently and record any discrepancy or missing instruction you discover.
Completion records practice and self-review. It is not certification of statistical competence.
