Source-linked AI summary
Dimensions: A Competitor to Scopus and the Web of Science?
Mike Thelwall
TL;DR
The paper asks whether Dimensions can provide sufficiently comprehensive and plausible citation data for research assessment compared with Scopus. It examines DOI-matched Food Science articles and a random sample of 10,000 Scopus articles, finding comparable coverage and citation counts, while identifying limits for some evaluation uses.
Problem
The paper addresses whether Dimensions provides enough coverage and plausible citation counts to support research evaluations compared with established citation databases.
Method
The study compares DOI-matched Scopus journal articles in Food Science from 2008-2017 and a random sample of 10000 Scopus articles from 2012 against Dimensions.
Results
97% of the random 2012 Scopus articles were found in Dimensions, with a median field-level Spearman correlation of 0.958 and similar geometric mean citation counts.
Takeaways & Limitations
Dimensions appears essentially interchangeable with Scopus for general citation analyses involving coverage and citation counts.
Takeaways & Limitations
The study did not investigate monographs, chapters, preprints, grants, patents, clinical-trials data, or Dimensions' analytical tools.
Abstract
from arXiv · showhide
Dimensions is a partly free scholarly database launched by Digital Science in January 2018. Dimensions includes journal articles and citation counts, making it a potential new source of impact data. This article explores the value of Dimensions from an impact assessment perspective with an examination of Food Science research 2008-2018 and a random sample of 10,000 Scopus articles from 2012. The results include high correlations between citation counts from Scopus and Dimensions (0.96 by narrow field in 2012) as well as similar average counts. Almost all Scopus articles with DOIs were found in Dimensions (97% in 2012). Thus, the scholarly database component of Dimensions seems to be a plausible alternative to Scopus and the Web of Science for general citation analyses and for citation data in support of some types of research evaluations.
1. Introduction
Dimensions is presented as a partly free citation index with lower citation counts than some alternatives but greater resistance to spam. The study assesses whether its coverage and citation counts are useful for research evaluations.
- Citation counts help researchers and managers evaluate research quality or impact when peer judgment is impractical or a second opinion is needed.
- Dimensions offers a partly free citation index whose data is less easy to spam than Microsoft Academic and Google Scholar.Its web-independent data may also reduce some processing errors.
- The study examines whether Dimensions contains enough data for useful research evaluations and whether its citation counts are plausible.
2. Research questions
The study compares Dimensions with Scopus to examine coverage, average citation counts, and whether their citation counts are interchangeable. It uses Scopus as the benchmark for comprehensiveness.
- Scopus is used instead of Web of Science because prior research found greater overall journal coverage for Scopus.
- The study asks how comprehensively Dimensions covers Scopus journal articles.
- It asks whether average journal-article citation counts in Dimensions are comparable to those in Scopus.
- It asks whether Dimensions and Scopus citation counts are interchangeable because they have a very high correlation.
3. Methods
The analysis uses DOI-matched Scopus journal articles from Food Science between 2008 and 2017, plus a random 2012 sample across fields. Coverage, geometric means, and Spearman correlations are compared between Dimensions and Scopus.
- 84691 Scopus Food Science journal articles with DOIs from 2008-2017 were matched against Dimensions by DOI.Articles without DOIs and documents such as books, conference papers, reviews, and editorials were excluded.
- Coverage was measured as the yearly percentage of Scopus articles with DOIs also found in Dimensions with a DOI.
- Geometric mean citation counts were compared with and without unmatched articles assigned a citation count of 0.Geometric means were used because citation data is highly skewed.
- Spearman correlations compared Scopus and Dimensions citation counts for each year, with field-level correlations used for the random sample.
4. Results and discussion
Dimensions showed broad coverage of Scopus Food Science articles and citation counts comparable to Scopus, with very high correlations for jointly indexed articles. Differences for individual papers mainly reflected indexing errors and journal-coverage differences.
- Coverage: Almost all older Scopus Food Science articles with DOIs were found in Dimensions, except during the partial 2018 current year.Dimensions appeared slower to index some journals than Scopus for the current year.
- Citation counts: Average geometric mean citation counts in Dimensions were about the same as Scopus.Scopus averages excluded articles not found in Dimensions; including them would make most Scopus averages 3% lower.
- Citation counts: There was a very high correlation between Scopus and Dimensions citation counts for jointly indexed articles, especially over longer periods.Lower recent-year correlations were attributed to smaller citation counts and shorter accumulation periods.
- Article-level discrepancies: Individual citation-count differences arose from indexing errors, split records, and journal-coverage differences.Examples included confused errata, unrecognised journal abbreviations, and journals indexed by only one database.
- Random-sample validation: In a random sample of 10,000 Scopus articles from 2012, 9,711 articles (97%) were found in Dimensions, with a median field-specific Spearman correlation of 0.958.Median geometric mean citation counts were 5.99 for Scopus and 6.32 for Dimensions.
- Additional indicators: RCR values correlated more strongly with Scopus citations than Altmetric scores, with median Spearman correlations of 0.838 and 0.262, respectively.RCR values were available for 7,450 articles and Altmetric scores for 4,273 articles in the random sample.
5. Conclusions
Dimensions’ journal-article coverage and citation counts appear comparable to Scopus, supporting its use for non-evaluative citation analyses and some research evaluations. However, indexing differences and susceptibility to gaming constrain its use for bibliometrics-driven evaluations.
- Dimensions’ coverage and citation counts are comparable to Scopus across the Food Science analysis and the 2012 all-field sample.
- Dimensions appears essentially interchangeable with Scopus for coverage and citation counts, although large differences can occur for individual articles.The differences may result from indexing errors or topic-related differences in journal coverage.
- Dimensions may support non-evaluative citation analyses and some formal research evaluations.
- Dimensions should not currently be used for bibliometrics-driven evaluations because preprint servers allow indirect manipulation of citation counts.The concern is strongest where evaluations create direct financial incentives to game citation metrics.