Combined Retracted Paper IDs from Retraction Watch Data and OpenAlex (December 31, 2024 Version)

What is this data? This dataset was constructed for the analysis of retracted papers by combining "Retraction Watch Data" with OpenAlex data. What is the format of this data? It adopts a simple schema with two fields: OpenAlex’s works_id and DOI. Other necessary information—such as the paper’s title, authors, and retraction reasons—can be obtained using these IDs from sources like OpenAlex, Crossref, and "Retraction Watch Data." What is the source data? "Retraction Watch Data" is the December 31, 2024 version that was acquired on January 1, 2025 via Crossref’s GitLab repository (https://gitlab.com/crossref/retraction-watch-data). Similarly, OpenAlex was obtained as a snapshot from the official S3 repository, also dated December 31, 2024. Why is this necessary? As is well known, OpenAlex incorporates "Retraction Watch Data", so one might assume that analyzing the retraction flags in the works object would suffice. However, as pointed out in the following literature (https://doi.org/10.48550/arXiv.2403.13339), there are many errors. Moreover, the retraction reasons are available only in the "Retraction Watch Data." Therefore, by independently merging both datasets, the accuracy and scope of the analysis can be enhanced. How were they combined? In merging "Retraction Watch Data" with OpenAlex, the paper’s DOI (Digital Object Identifier) is used as the primary key. Note that "Retraction Watch Data"contains two types of DOIs: RetractionDOI: The DOI of the retraction notice (i.e., the article announcing the retraction). OriginalPaperDOI: The DOI of the retracted paper (i.e., the regular paper that is being retracted). There are cases where the two DOIs are identical or where only one is available. For the merging process described above, the OriginalPaperDOI is used. The number of unique OriginalPaperDOIs is 50,831, representing the maximum number of mergeable candidates. In comparison, the simple unique count of RetractionDOIs is 51,050; excluding those identical to the OriginalPaperDOIs leaves 38,314, and further limiting to records that do not have an OriginalPaperDOI yields 2,605 unique RetractionDOIs. Even among papers with DOIs, some are not included in OpenAlex, resulting in 50,073 records being linked. Among these linked retracted papers, several were published in the 1960s. Therefore, the analysis was limited to a 10-year period from 2013 to 2022, resulting in 32,368 papers being included in the study.本数据集是什么？ 本数据集通过结合撤稿观察数据集（Retraction Watch Data）与OpenAlex学术数据库（OpenAlex）的数据构建而成，用于撤稿论文的相关分析。 本数据集采用何种格式？ 本数据集采用极简架构，仅包含两个字段：OpenAlex的works_id与数字对象标识符（Digital Object Identifier，DOI）。借助这两类标识符，可从OpenAlex、Crossref以及撤稿观察数据集（Retraction Watch Data）等数据源获取论文标题、作者、撤稿原因等其他必要信息。 数据源为何？ 撤稿观察数据集（Retraction Watch Data）采用2024年12月31日版本，于2025年1月1日通过Crossref的GitLab代码仓库（https://gitlab.com/crossref/retraction-watch-data）获取。同理，OpenAlex学术数据库（OpenAlex）的快照数据同样取自其官方S3存储仓库，版本日期亦为2024年12月31日。 为何需要构建该数据集？ 众所周知，OpenAlex学术数据库（OpenAlex）已集成撤稿观察数据集（Retraction Watch Data），因此部分研究者可能认为仅需分析works对象中的撤稿标记即可完成研究。但正如文献（https://doi.org/10.48550/arXiv.2403.13339）所指出的，此类标记存在大量错误。此外，撤稿原因仅能从撤稿观察数据集（Retraction Watch Data）中获取。因此，通过独立合并两类数据集，可提升分析的准确性与覆盖范围。 数据集是如何合并的？ 在合并撤稿观察数据集（Retraction Watch Data）与OpenAlex学术数据库（OpenAlex）的数据时，本数据集以论文的数字对象标识符（Digital Object Identifier，DOI）作为主键。 需注意，撤稿观察数据集（Retraction Watch Data）包含两类DOI： 撤稿通知DOI（RetractionDOI）：指用于发布撤稿声明的文章对应的数字对象标识符。 原论文DOI（OriginalPaperDOI）：指被撤稿的常规论文对应的数字对象标识符。 部分记录中两类DOI完全一致，或仅存在其中一类。本次合并流程采用原论文DOI（OriginalPaperDOI）作为关联依据。唯一的原论文DOI共计50831个，代表可关联的最大候选论文数量。相较之下，仅统计撤稿通知DOI的唯一值共计51050个；剔除与原论文DOI重复的条目后，剩余38314个；若进一步限定为无原论文DOI的记录，则仅剩余2605个唯一撤稿通知DOI。 即便拥有DOI的论文中，仍有部分未被收录至OpenAlex学术数据库（OpenAlex），最终成功关联的记录共计50073条。在这些已关联的撤稿论文中，部分发表于20世纪60年代。因此，本次分析将时间范围限定为2013年至2022年的10年间，最终纳入研究的撤稿论文共计32368篇。