Мазмұнға өту

Қайталанымдылық · Бастапқы

Template 6: Reproducibility README

Place this with the data and analysis materials that support the paper. Replace all prompts with actual information and test the instructions from a clean starting point.

3 мин оқу · 2026 ж. 26 қаз.

Бұл бет әлі аударылмаған, сондықтан ағылшын нұсқасын оқып отырсыз.

Негізгі тұжырымдар

  • Adapt the prompts to the actual study.
  • Complete every statement from verified records.
  • Follow the target journal and institutional requirements.
Мазмұны

Place this with the data and analysis materials that support the paper. Replace all prompts with actual information and test the instructions from a clean starting point.

Place this with the data and analysis materials that support the paper. Replace all prompts with actual information and test the instructions from a clean starting point.

Project and release

Title and related manuscript: [Title and actual identifier, if assigned]

Release version and date: [Version; date]

Contributors and contact: [Names, roles and appropriate contact]

Purpose: [Which results this package reproduces]

Licence and restrictions: [Data, code and material licences; confidential elements]

Contents

Raw inputs: [Files, origin and access conditions]

Processed data: [Files and transformations]

Dictionary: [Location]

Analysis materials: [Scripts, notebooks or reproducible procedure]

Outputs: [Tables and figures with corresponding manuscript numbers]

Decision log: [Exclusions, amendments and sensitivity analyses]

Environment

Operating system and hardware where relevant: [Details]

Software and package versions: [Exact versions used]

Installation or environment instructions: [Actual steps]

Randomness: [Seed handling and any known nondeterministic operations]

Expected runtime and resources: [Measured or clearly estimated requirements]

Reproduction sequence

  1. Obtain [permitted inputs] from [verified source].
  2. Confirm [version, checksum or identifying information].
  3. Prepare [environment] using [instructions].
  4. Run [documented entry point or procedure].
  5. Compare [outputs] with [expected values or files].

Expected result: [A small verification result that demonstrates successful execution]

Known limitations: [Unavailable restricted data, hardware dependence or manual steps]

Restricted access route: [Who can request access, criteria and responsible decision maker]

Changes from earlier versions: [Description and impact]

Do not place credentials or participant identifiers in this document. If data cannot be shared openly, provide a truthful access statement and enough nonconfidential documentation to understand the analysis.

Tutorial: reproduce one reported result from the beginning

Read Article 6 and Article 29. Choose one figure and trace it backwards to its input files. Then write the forward sequence a new reader would follow. State the expected working directory, input filenames, software versions, configuration and output filenames. A list of packages alone does not explain how the analysis runs.

Hypothetical workflow: The reader begins with a preserved measurement file and a dictionary. A cleaning script checks identifiers and units, a second script creates the prespecified summaries, and a final script generates the table and figure. The README states which outputs are regenerated and which files contain manual annotations. If a step requires restricted data, explain the access process and provide a suitable nonidentifying example when possible.

Record uncertainty honestly: A random seed can help reproduce a stochastic computation, but it does not document every software or hardware dependency. If exact numerical equality is not expected across environments, state a justified comparison tolerance and explain what readers should inspect.

Completion check: Run the documented workflow in a clean working folder. Verify at least one manuscript value and one figure against the generated outputs. Record what was actually tested, and do not describe an untested installation path as verified.