Versioning Research Data: New Recommendations for Greater Transparency

Research data are frequently corrected, supplemented, converted into different formats and republished. How is it possible to trace the data underlying a scientific analysis? A new publication in the Data Science Journal offers recommendations for the versioning of research data. The authors include ECDF Professor Heinz Pampel and ECDF board member Laura Rothfritz (both from Humboldt-Universität zu Berlin).

In the paper titled “What Do We Mean by Version? Actionable Guidelines for the Versioning of Data,“ a German-Australian research team, led by Jens Klump, Lesley Wyborn, Mingfang Wu, Dorothea Strecker, Heinz Pampel and Laura Rothfritz, examines how different data states can be clearly identified and the relationships between them can be documented transparently. Such transparent versioning helps verify research findings, makes data accessible for reuse and ensures that the contributions of the researchers involved are appropriately visible and traceable.

The team distinguishes between changes that occur during the processing of a dataset and the deliberate release of a specific data state for publication. In doing so, they also take into account various file formats and copies of a dataset hosted on different platforms. To describe these relationships systematically, the team adapts an established model from library and information science for use with research data.

The recommendations are aimed at researchers, research data management experts and data repository operators. They outline how versions can be named consistently, assigned persistent identifiers and how changes can be documented in a traceable manner. Additional recommendations address the linking of source data with derived data products, as well as the handling of retracted or no longer available data.

This work builds upon the activities of the Research Data Alliance, which is dedicated to establishing and further developing international standards, infrastructures and best practices in research data management. The recommendations presented here are based on 56 use cases from approximately 42 organizations, as well as on conversations and discussions with experts from the academic community and information infrastructure sectors. These insights were further supplemented by the results of a workshop held in Berlin in June 2024 with the support of the Berlin University Alliance. At the event, held at the ECDF, around 4 experts discussed the recommendations and contributed insights from their own professional practice.