Skip to content
PlagiatScanner.de
Pillar · Research data

Turn research data into a traceable thesis record

A defensible data workflow connects the result in your thesis to an identified source file, documented transformations, analysis instructions and the environment that produced the output. Begin with responsibilities, consent and protection rather than waiting for submission week. If material can be shared, add clear metadata, a suitable repository, an appropriate licence and a version-specific citation. These twenty guides help you choose the next practical task without assuming that every dataset can or should be open.

Twenty focused research guidesPlan · Identify · Transform · Reproduce

Start from the evidence chain your claim will need

Research data are not limited to the spreadsheet eventually submitted. A useful record can include original files, a codebook, cleaning decisions, analysis code, software details, parameters and access conditions. Some components may be public; others may require controlled access or may not be shareable at all. Transparency therefore means explaining both what is available and why a boundary exists.

The right entry point depends on the project. Interview research may prioritise consent language and disclosure risk. A simulation may need an environment record and random seed from the first run. Whichever route you take, avoid treating documentation as a narrative reconstructed from memory. Create small, dated records while the decisions are being made.

1. Govern the material before deciding to share it

Set responsibilities, storage locations, access and retention early. Consent for reuse is not interchangeable with consent to participate, and de-identification needs more than removing names. Once the shareable material is known, a licence and repository can be assessed against rights, sensitivity, disciplinary practice and the intended audience.

Data management plan for studentsRecord ownership, storage, access, protection and close-out decisions.Consent for data reuseExplain future access and purposes without vague promises.Anonymise sensitive research dataReview direct identifiers, combinations and contextual disclosure.Choose a research-data licenceConfirm authority to license before selecting reuse terms.Choose a data repositoryCompare disciplinary fit, access controls, metadata and stewardship.

2. Give each research object a durable identity

A sound citation points to a particular dataset version rather than a project homepage. Creators, title, publisher or repository, year, version and persistent identifier work together. Metadata explain what the object contains and how it was produced. Software and code repositories require their own version-aware references because a moving branch cannot preserve the analytical state.

How to cite research dataAssemble a retrievable and version-specific reference.Minimum metadata for research dataDescribe content, method, structure, access and limitations.Obtain and use a dataset DOIConnect the identifier to a maintained landing page and version.How to cite research softwareName contributors, software title, release and archive location.How to cite a GitHub repositoryPrefer an archived release or specific commit over a live branch.

3. Record what happened between collection and result

Keep source material immutable and produce a distinct cleaned or transformed version. Each rule should explain why it was applied and which records changed. File names are useful labels, but they are not a complete version history. Spreadsheet work and statistical output also need inputs, formula or syntax evidence, manual actions and report destinations.

Separate raw and clean dataProtect the source while linking every derived analysis file.Build a data-cleaning logRecord the rule, reason, affected cases and validation result.Document versions and analysis pathsMap dependencies and mark the state approved for reporting.Document reproducible Excel analysisExpose source ranges, formulas, manual steps and exports.Audit statistical software outputLink data, syntax, generated output and the claim in the thesis.

4. Make the analysis recoverable and understandable

A future reader needs the environment and parameters as well as the script. Record the software version, packages and any controlled randomness. Decide where analysis code belongs based on readability, size and access. Backups protect ongoing work, while a reproducibility package prepares the approved data, code, instructions and metadata for examination or reuse.

Document software versionsCapture the application, release, environment and relevant extensions.Document a random seedState the random process, seed and limits of repeatability.Analysis code: appendix or repository?Balance access, version control, scale and reading context.Prevent data loss during a thesisUse independent copies and test that restoration actually works.Build a reproducibility packageBundle approved inputs, code, metadata and a clear run guide.

Workflow triage: choose the next record to create

A practical first move for four common situations
Current situationCreate firstThen verify
Collection is about to beginManagement, consent and storage recordAccess roles and foreseeable reuse
An output cannot be reconstructedRaw-to-clean map and transformation logSoftware, parameters and output destination
A dataset is ready for depositMetadata and rights reviewRepository, version, identifier and access tier
Submission is approachingReproducibility package and restore testRemove unneeded sensitive material

This triage is not a substitute for disciplinary or institutional requirements. It helps expose the next missing link so that the work can be improved before the evidence trail is lost.

Standards and institutional foundations

  1. ALLEA: The European Code of Conduct for Research Integrity – institutional principles for reliable research practice.
  2. GO FAIR: FAIR Principles – guidance on findable, accessible, interoperable and reusable data.
  3. DataCite Metadata Schema – the official metadata framework for registered research objects.