Workshop 2 · Data

Produce one checked result from an approved data source.

Record the data source, inspection, cleaning, analysis, and result.

Data source

Confirm that the data source is approved and relevant.

• Use an approved survey, API, public-data source, or prepared fallback.
• Collect new data only when approvals and time support it.
• Do not upload restricted data to an unapproved tool.

Preserve provenance

Record the source before changing the file.

• Save where and when the data came from.
• Keep the original file unchanged.
• Capture definitions, units, access conditions, and version details.

Data inspection

Understand what one row means.

• Check row grain, fields, units, missingness, and repeated definitions.
• Look for join risks and unexpected exclusions.
• Confirm the files can actually address the question.

Plan the transformation

Make cleaning reversible and reviewable.

Propose changes first. Pause for choices that alter scientific meaning, then work from a copy with exclusions and recodes documented.

Analyze for the question

Choose one method that answers the research question.

Tie the calculation to the study logic, state assumptions, and keep causal language inside what the design can support.

Check the output

Verify that the figure or table matches the calculation and claim.

• Confirm the figure or table matches the calculation.
• Separate the result from its interpretation.
• State limitations and the next human check.

Arcade recording placeholder

Prepared example: inspect, clean, analyze, check.

Recording pending: preserve the source, inspect the data, document one cleaning decision, create one visual, and verify the calculation.