Source-linked AI summary
A further step forward in measuring journals' scientific prestige: The SJR2 indicator
Vicente P. Guerrero-Bote, Felix Moya-Anegon
TL;DR
Journal prestige indicators can produce scores that require comparison across journals. This paper proposes SJR2, a Scopus-based indicator that weights citations by citing-journal prestige and cocitation-based closeness, yielding scores that do not tend to decrease as new journals are added.
Problem
Existing journal prestige scores may only be meaningful when compared with scores from other journals.
Method
SJR2 uses Scopus data to weight citations by citing-journal prestige and journal closeness estimated through cocitation.
Results
SJR2 produces values that do not tend to decrease as new journals are incorporated and that are endowed with meaning.
Takeaways & Limitations
The proposed indicator is intended to provide more precise and useful journal metrics through prestige-weighted citation measurement.
Takeaways & Limitations
The score only makes sense in comparison with the scores of other journals.
Abstract
from arXiv · showhide
A new size-independent indicator of scientific journal prestige, the SJR2 indicator, is proposed. This indicator takes into account not only the prestige of the citing scientific journal but also its closeness to the cited journal using the cosine of the angle between the vectors of the two journals' cocitation profiles. To eliminate the size effect, the accumulated prestige is divided by the fraction of the journal's citable documents, thus eliminating the decreasing tendency of this type of indicator and giving meaning to the scores. Its method of computation is described, and the results of its implementation on the Scopus 2008 dataset is compared with those of an ad hoc Journal Impact Factor, JIF(3y), and SNIP, the comparison being made both overall and within specific scientific areas. All three, the SJR2 indicator, the SNIP indicator and the JIF distributions, were found to fit well to a logarithmic law. Although the three metrics were strongly correlated, there were major changes in rank. In addition, the SJR2 was distributed more equalized than the JIF by Subject Area and almost as equalized as the SNIP, and better than both at the lower level of Specific Subject Areas. The incorporation of the cosine increased the values of the flows of prestige between thematically close journals.
1. Introduction
The introduction motivates objective journal valuation by identifying limitations in existing citation-based and prestige indicators, including size dependence and poor comparability across fields. It presents SJR2 as a cocitation-based indicator that weights citing-journal prestige and thematic closeness without arbitrary classifications or journal-set-size dependence.
- Limitations of existing indicators: Citation-based indicators are difficult to compare across scientific areas because fields have different citation habits and citation rates, while journal classification lacks consensus.The Impact Factor has been extensively used, but its values are typically lower in Engineering, Social Sciences, and Humanities.
- Limitations of existing indicators: Existing prestige indicators often produce scores meaningful only relative to other journals, and normalization causes scores to decrease as the journal set grows, complicating temporal comparisons.These indicators may represent prestige or average prestige per document, but their fixed-sum normalization can make all scores decline over time.
- Contribution and motivation: SJR2 weights citations by the citing journal’s prestige and thematic closeness to the cited journal, using an objective cocitation method rather than arbitrary field classifications.The indicator is designed to give greater weight to citations from same or similar fields while avoiding dependence on predefined scientific-area categories.
- Contribution and motivation: SJR2 additionally avoids dependence on the size of the journal set, giving its prestige score a meaning that other prestige indicators lack.This addresses the tendency of normalized indicators to decline when the number of journals increases.
- Paper scope: The paper describes SJR2’s methodology and reports its implementation using Elsevier’s Scopus database and data from the Scimago Journal & Country Rank website.The directory contained almost 19,000 scientific journals and other publications in 2009.
2. Data
The SJR2 indicator was developed using Scopus because it represents the global structure of science and supports key bibliometric data-quality requirements. The study analyzed 1,999,777 documents published in 2008 and their 26,036,560 references to earlier indexed documents.
- Data source: Scopus was selected as the data source for developing SJR2 because it best represents the overall structure of world science at a global scale.The choice reflected criteria concerning citable versus total output, document-type assignment, and reference-to-record linkage accuracy.
- Data source: 1,999,777 documents published in 2008 were used, with 26,036,560 references retrieved to documents from previous years in the database.The data came from a Scopus raw-data copy exported in May 2011.
- Classification: The Scopus classification comprises 295 Specific Subject Areas grouped into 26 Subject Areas, plus a General Subject Area for multidisciplinary journals such as Nature and Science.These classifications provide the area and category structure used to organize the indexed documents and journals.
- Classification: Scopus organizes its subject coverage into Life Sciences, Physical Sciences, Social Sciences, and Health Sciences.The listed categories contain 3,950, 6,350, 5,900, and 6,200 titles, respectively; Health Sciences includes 100% coverage of Medline titles.
3. Method
SJR2 extends SJR by weighting prestige transfer by citation counts and thematic closeness, constraining transfers, and normalizing accumulated prestige by citable documents. It is computed in two phases: size-dependent PSJR2 followed by size-independent normalization for journal comparison.
- 3. Method: SJR2 distributes citing-journal prestige according to three-year citations and the cosine between journals’ cocitation profiles, increasing transfer between thematically closer journals.The cosine is computed from the cocitation profiles of the citing and cited journals.
- 3. Method: SJR2 caps prestige transfer at 50% of the source journal’s prestige and 10% per citation to limit link-farm-like effects.The restriction addresses journals with very few recent references or highly specialized journals.
- 3. Method: Normalizing PSJR2 by the proportion of citable documents produces values that do not tend to decrease as new journals are added and gives the scores meaning.Citable documents include articles, reviews, short surveys, and conference papers in the three-year window.
- 3. Method: The indicator has two phases: computing Prestige SJR2 (PSJR2), a size-dependent overall-prestige measure, then normalizing it into the size-independent SJR2 indicator.The normalized indicator is intended for comparing journals.
- 3. Method: Each journal starts with prestige 1/N, and iterative updates combine database inclusion, document volume, and the number, importance, and closeness of received citations.The citation network represents journals as nodes and citation relationships as directed links.
C Cos D PSJR
SJR2 combines prestige propagation with cosine similarity between journals’ cocitation profiles, giving thematically closer journals greater weight. It normalizes prestige by each journal’s share of citable documents, producing size-independent scores whose document-weighted mean is unity.
- C Cos D PSJR: The cosine excludes cocitations between the two journals and uses citations made in the target year to a three-year citation window.Excluding pairwise cocitations avoids distortions because a journal’s self-cocitations are usually much more frequent than cocitations with other journals.
- C Cos D PSJR: SJR2 weights prestige flows by the cosine between journals’ cocitation profiles, capturing thematic similarity more finely than direct cocitation alone.Similar cocitation profiles indicate related subject matter, so citations from thematically close journals receive greater weight and may normalize differences across Subject Areas.
- C Cos D PSJR: Dividing each journal’s PSJR2 by its share of citable documents converts size-dependent accumulated prestige into a journal-comparable prestige-per-document score.A value of 1.0 represents mean prestige per document; 0.8 and 1.3 represent 20% less and 30% more than the mean, respectively.
- C Cos D PSJR: The citable-document normalization guarantees that the annual SJR2 values have a document-weighted mean of unity.This contrasts with the SJR’s tendency to decline as prestige is distributed across a growing number of journals.
4. Statistical Characterization
Using stable 2008 Scopus data, the study characterizes SJR2 against JIF(3y) and SNIP through rank distributions, correlations, subject-area normalization, and citation-flow patterns. SJR2 is less concentrated than JIF(3y), approaches SNIP’s equalization, and increases prestige flows between thematically related journals.
- Overall statistical characterization: SJR2, JIF(3y), and SNIP all follow approximately logarithmic rank distributions, but SJR2 declines most gradually, whereas JIF(3y) declines most sharply.The results indicate that prestige is less concentrated than citation, with fewer prestigious journals than highly cited ones; the three metrics are strongly correlated.
- Subject-area normalization: In the JIF(3y) comparison, one Economics, Econometrics and Finance journal with impact 6.29 receives SJR2 16.87, illustrating SJR2’s normalization across citation habits.Biochemistry, Genetics and Molecular Biology journals lie above Economics, Econometrics and Finance journals in JIF(3y), while SJR2 reduces this separation.
- Subject-area normalization: SJR2 produces more equalized subject-area distributions than JIF(3y), approaching SNIP overall and remaining closer to SNIP for Specific Subject Areas.Excluding the special General area, SNIP is most equalized, followed closely by SJR2 for Subject Areas; SJR2 remains closer for Specific Subject Areas.
- Citation-flow analysis: Citation rates vary substantially across fields: same-Specific-Subject-Area citations range from 17% in Nursing to 63% in Dentistry, while SJR2 increases same-area flow percentages.The increase is greater when the cosine is included, indicating that thematic closeness raises the prestige flow between journals.
5. Conclusions
SJR2 addresses declining prestige scores by normalizing prestige against citable documents and incorporates thematic closeness through cocitation-profile cosine weighting. It equalizes scores across subject areas and represents a step toward measuring journals’ real prestige.
- SJR2 overcomes the tendency for prestige scores to decrease as more journals and documents enter the calculation.
- SJR2 divides a journal’s prestige portion by its citable-document portion, making the mean score unity and keeping the weighted average equal to unity regardless of dataset size.A score above unity means prestige exceeds the journal’s portion of citable documents, and a score below unity means the reverse.
- Cosine weighting gives greater influence to citations from thematically close journals, especially within the same Specific Subject Area, while reducing influence from other areas.This thematic weighting is based on the cosine between citing and cited journals’ cocitation profiles.
- SJR2 greatly equalizes prestige distributions across Subject Areas and especially Specific Subject Areas, making scores from different areas more comparable without arbitrary journal classifications or weights.
- Despite high overall and area-level correlations with SNIP and JIF, SJR2 is presented as a step toward a better representation of scientific journals’ real prestige.