What to Include in a Reproducible Analysis Handoff

Make the next person’s task explicit

A folder of spreadsheets and output screenshots does not explain how a result was produced. A useful handoff lets an authorized collaborator identify the data version, understand the definitions, rerun the intended steps, and connect each output to a manuscript claim.

Begin with a short README describing the study question, the unit of observation, the directory structure, and the order in which scripts run. Identify decisions that remain unresolved. Keep participant identifiers and access credentials out of general handoff notes.

Separate inputs, transformations, and outputs

Preserve source data in the institution’s approved location. Store cleaning and analysis instructions separately from generated tables and figures. Use versioned filenames or a suitable version-control workflow for code and documentation. A new dataset should not silently overwrite the one used for a submitted table.

List software and package versions, random seeds where relevant, and any manual steps that cannot yet be reproduced. Document excluded records, recoded values, and derived variables. If two analysts interpret a category differently, resolve the definition before comparing models.

Connect a result to its origin

For each manuscript table, record the generating script or saved analysis, the input dataset version, and any formatting steps. Label exploratory results clearly. A collaborator should not need to guess which of several similarly named output files contains the final analysis.

A clean rerun is a useful check, but it does not establish that the statistical method is appropriate. Review the scientific assumptions separately from the ability to execute the workflow.

Agree on a minimal delivery package

Use the analysis handoff worksheet to assign owners and list missing components. Pair it with the Data Collection Builder when variable definitions still need work.

For a free graphical analysis environment, jamovi and JASP provide open-source options. Confirm that the selected modules support the intended method, and review saved analyses before treating the output as final.