Long-term availability of data associated with articles in PLOS ONE

Abstract

The adoption of journal policies requiring authors to include a Data Availability Statement has helped to increase the availability of research data associated with research articles. However, having a Data Availability Statement is not a guarantee that readers will be able to locate the data; even if provided with an identifier like a uniform resource locator (URL) or a digital object identifier (DOI), the data may become unavailable due to link rot and content drift. To explore the long-term availability of resources including data, code, and other digital research objects associated with papers, this study extracted 8,503 URLs and DOIs from a corpus of nearly 50,000 Data Availability Statements from papers published in PLOS ONE between 2014 and 2016. These URLs and DOIs were used to attempt to retrieve the data through both automated and manual means. Overall, 80% of the resources could be retrieved automatically, compared to much lower retrieval rates of 10–40% found in previous papers that relied on contacting authors to locate data. Because a URL or DOI might be valid but still not point to the resource, a subset of 350 URLs and 350 DOIs were manually tested, with 78% and 98% of resources, respectively, successfully retrieved. Having a DOI and being shared in a repository were both positively associated with availability. Although resources associated with older papers were slightly less likely to be available, this difference was not statistically significant, suggesting that URLs and DOIs may be an effective means for accessing data over time. These findings point to the value of including URLs and DOIs in Data Availability Statements to ensure access to data on a long-term basis.