Unit 6: Journal Citation Indices

I. Orientation: What Citation Indices Measure

Citation indices are quantitative proxies for the reach and influence of scholarly output, built on the assumption that a citation is a traceable acknowledgement of use. They convert raw citation counts into comparable scores at the level of the author, the article, or the journal, and are computed over defined publication and citation windows drawn from large databases.

• Core inputs: every index needs a set of publications, a count of citations to each, and a time window: e.g. "citations received in 2023 to items published 2021–2022".
• Three data ecosystems: Web of Science (Clarivate) drives JIF and Eigenfactor; Scopus (Elsevier) drives CiteScore, SJR and SNIP; Google Scholar drives h5-index, h5-median and i-10.
• Size vs. prestige: raw metrics (h-index, Eigenfactor) grow with output volume; normalised metrics (JIF, SNIP, SJR) express average or field-adjusted intensity.
• Self-citation and gaming: counts can be inflated by author or journal self-citation, so several indices exclude or discount it.
• Key convention: a "citable item" (article, review) is distinguished from front matter (editorials, letters) that may still attract citations.

II. Author- and Journal-Level h-Family Indices

These indices summarise a whole publication portfolio in a single number that balances quantity against per-paper impact, and are applied to individual researchers and to journals.

A. h-index

A scholar or journal has index h if h of their papers have each been cited at least h times, and the remaining papers have no more than h citations each.

• Definition: rank papers by descending citations; h is the highest rank at which citation count ≥ rank.
• Worked example: citations [25, 12, 9, 6, 5, 3, 2] → the 5th paper has 5 citations (5 ≥ 5) but the 6th has only 3 (< 6), so h = 5.
• Strength: insensitive to a single highly cited "hit" and to many uncited papers.
• Limitation: can only rise with time, ignores citations above the threshold, and cannot be compared across fields with different citing rates.

B. h5-index

The h5-index is the h-index restricted to articles published in the five most recent complete calendar years, used by Google Scholar Metrics to rank journals.

• Window: for a 2024 edition it covers 2019–2023 articles and their citations to date.
• Purpose: removes the age advantage of long-established journals by capping the counting period at five years.
• Anchor: a journal with h5 = 80 published at least 80 articles in that window each cited ≥ 80 times.

C. h5-median

The h5-median is the median number of citations for exactly those articles that make up a journal's h5-index.

• Role: a companion to h5-index that shows how heavily the core (h5) articles are actually cited.
• Example: two journals both have h5 = 40; the one whose 40 core articles have a median of 95 citations is more intensely cited than one with a median of 55.
• Interpretation: always ≥ h5-index, since each core article has at least h5 citations.

D. g index

The g-index gives greater weight to highly cited papers by asking how far down the ranked list the cumulative citation total keeps pace with the square of the rank.

• Definition: the largest number g such that the top g articles together received at least g² citations.
• Formula:

find max g such that   Σ (citations of top g papers) ≥ g²

• Worked example: citations [25, 12, 9, 6, 5]; cumulative = 25, 37, 46, 52, 57. At g = 7 we would need 49 but only have (say) 57 across 5 → check g = 5: 57 ≥ 25 ✓; extend if more papers. Here g = 5 and rewards the top paper's 25 citations that h ignored.
• Advantage over h: additional citations to top papers can raise g, so a few landmark works are recognised.

E. i-10 index

The i-10 index is simply the number of publications with at least 10 citations each, reported by Google Scholar profiles.

• Definition: count of papers where citations ≥ 10.
• Merit: trivial to compute and easy to read.
• Limitation: the threshold of 10 is arbitrary and field-blind, and it ignores how far above 10 each paper reaches.

III. Altmetrics: Attention Beyond Citations

A. almetrics

Altmetrics ("alternative metrics") track the online attention a research output receives, capturing engagement that citation counts miss and doing so far sooner after publication.

• Sources tracked: social media posts, news and blog mentions, Wikipedia and policy-document references, and reference-manager saves (e.g. Mendeley readers).
• Altmetric Attention Score: a single weighted number where a news story counts more than a tweet, visualised as the coloured "donut".
• Speed: attention accrues in days, whereas citations take 1–3 years, making altmetrics useful for early signals.
• Complement, not replacement: high attention need not mean scholarly quality; controversy also drives scores.
• Limitation: platform coverage shifts and metrics are easily inflated by coordinated sharing.

IV. Journal Citation Metrics

These are journal-level scores that rank periodicals within subject categories; they differ chiefly in database, citation window, and whether they normalise for field or weight citations by source prestige.

A. JIF (Journal Impact Factor)

JIF measures the average citations in one year to a journal's citable items from the previous two years, published by Clarivate in the Journal Citation Reports.

• Formula:

JIF(Y) = (citations in Y to items published in Y-1 and Y-2)
         ÷ (number of citable items published in Y-1 and Y-2)

• Example: 500 citations in 2023 to 250 articles from 2021–2022 → JIF = 2.0.
• Numerator/denominator asymmetry: editorials can generate citations (numerator) without counting as citable items (denominator), a known distortion.
• Limitation: two-year window disadvantages slow-citing fields (mathematics, humanities); values are not comparable across disciplines.

B. JIF percentile

The JIF percentile expresses a journal's rank within its subject category as a percentile, so scores are comparable across fields.

• Formula:

JIF percentile = (N - rank + 0.5) ÷ N × 100

where N is the number of journals in the category and rank is the journal's position by JIF.

• Reading: a percentile of 95 means the journal outranks about 95% of journals in its category, regardless of the absolute JIF value.
• Use: lets a JIF of 4.0 in one field be judged against 9.0 in another via relative standing.

C. cite score

CiteScore is Scopus's journal metric, computed over a four-year window for both citations and documents, which broadens the base compared with JIF.

• Formula:

CiteScore(Y) = (citations in Y to Y-3 to documents in Y to Y-3)
               ÷ (documents published in Y to Y-3)

• Symmetry: the same document types count in numerator and denominator, avoiding JIF's asymmetry.
• Coverage: Scopus indexes more titles than Web of Science, so CiteScore reaches more journals.
• Contrast with JIF: the longer window and larger database usually make CiteScore and JIF differ in absolute value even for the same journal.

D. SJR (SCImago Journal Rank)

SJR weights each incoming citation by the prestige of the citing journal, so a citation from a leading journal counts more than one from an obscure one.

• Principle: a PageRank-style eigenvector calculation over Scopus, transferring prestige through the citation network.
• Window: three years of cited content.
• Self-citation cap: journal self-citations are limited (about 33%) to curb inflation.
• Reading: SJR is field-normalised and size-independent, expressed as a value where higher means more prestige-weighted influence per paper.

E. SNIP (Source Normalised Impact per Paper)

SNIP corrects for field citation potential, so journals in fields that cite heavily are not automatically favoured.

• Principle: raw impact per paper is divided by the citation potential of the journal's subject field (developed by CWTS Leiden, on Scopus).
• Formula (conceptual):

SNIP = (citations per paper) ÷ (citation potential of the field)

• Normalisation: a SNIP of 1.0 marks average field impact; > 1 is above the field norm.
• Merit: enables direct cross-field comparison without predefined subject categories.

F. eigen factor

The Eigenfactor Score measures a journal's total importance to the scholarly literature over a five-year window, weighting citations by the influence of the citing source.

• Window: citations in the current year to the previous five years of content.
• Self-citation: journal self-citations are removed before calculation.
• Size-dependence: it scales with the number of articles, so large journals score higher; the Article Influence Score (Eigenfactor per article, normalised to a mean of 1) provides the per-paper version.
• Network basis: like SJR it uses an iterative, PageRank-type weighting so citations from influential journals carry more weight.