DOI and versioning

Why NEMAR has its own DOI namespace

NEMAR mints dataset identifiers in its own 10.82901/NEMAR namespace through EZID and publishes the associated DataCite metadata. The namespace gives the NEMAR resource a stable way to connect three things that otherwise tend to drift apart: the dataset as a continuing scholarly object, a specific released version, and the landing page that explains how to use it.

This is not just a shorter download link. A minimal citation record might provide a title, an author, and a URL. A populated NEMAR record can carry much more context when the source metadata supports it:

• authors, ORCID identifiers, and affiliations;
• an abstract, acknowledgement text, and MeSH-validated keywords;
• funding information and related publications or identifiers;
• issued and, when supplied, collection dates;
• language, geographic information, file formats, sizes, and license;
• the dataset version and the relationship between the concept and its versions.

The result is a richer citation and a better machine-readable description of what the dataset is. “Can carry” matters: the record is only as complete as the metadata supplied and successfully validated during curation.

Concept DOI and version DOI

The concept DOI identifies the dataset across its useful lifetime. A version DOI identifies one released state. For example, a dataset might have a concept DOI like:

10.82901/NEMAR.nm000103

and a version DOI associated with a release such as:

10.82901/NEMAR.nm000103.v1.0.0

The exact identifier shown on a dataset landing page is authoritative. The important relationship is conceptual: cite the concept when discussing the dataset broadly; cite the version when an analysis depends on the exact files, metadata, or manifest used.

Why versioning is part of FAIR practice

FAIR data must be findable and reusable, but reuse also requires knowing which state was used. A DOI should not turn a dataset into an unchangeable museum object, and an update should not silently rewrite the bytes behind an old citation.

NEMAR therefore treats a release as a fixed citable state. A later correction or addition goes through a reviewable pull request, receives a new dataset version, and can receive a new version DOI. The concept-level identity connects the history; the version-level identity protects the meaning of an earlier analysis.

For a concrete comparison with an imported OpenNeuro dataset, see Why use NEMAR?. The comparison uses the HBN EEG on005506 / ds005506 records and distinguishes the two repositories’ version numbers.

The DOI itself is permanent, while its registrar metadata can be corrected or enriched when a curation error is found. That distinction lets NEMAR improve descriptions and relationships in public without pretending that a previous release never existed.

A practical citation rule

When you publish an analysis, record:

1. the NEMAR dataset ID;
2. the version DOI or explicit version tag;
3. the analysis code and relevant parameters;
4. any derivative outputs and their own provenance.

The dataset versioning guide explains the contributor workflow. The Data API exposes version lists, metadata, and manifests for tools that need to resolve the same record programmatically.