Citation analysis: State of the art, good practices, and future developments

Transcription

1 Citation analysis: State of the art, good practices, and future developments Ludo Waltman Centre for Science and Technology Studies, Leiden University Bibliometrics & Research Assessment: A Symposium for Librarians & Information Professionals Bethesda, Maryland October 31, 2016

2 Centre for Science and Technology Studies (CWTS) Research center at Leiden University in the Netherlands History of more than 25 years in bibliometric and scientometric research Commercial bibliometric analyses for clients worldwide 1

3 State of the art 2

4 What do citations measure? Visibility Relevance Quality Scientific impact Citations Reputation Other factors 3

5 Citizen bibliometrics vs. professional bibliometrics Citizen bibliometrics Do-it-yourself bibliometrics by researchers, research managers, and librarians Publication and citation counts, impact factor, h-index Web of Science, Scopus, Google Scholar Mainly at the level of individual researchers Professional bibliometrics Bibliometric analyses supported by professional bibliometricians or specialized bibliometric software tools Field-normalized indicators Web of Science, Scopus Mainly at the level of research institutes 4

6 Impact factor Impact factor= # citations in 2015 to # citable articles

7 Common criticism on impact factor Too short citation window Too sensitive to highly cited publications Inflation by review articles Numerator/denominator inconsistency Vulnerability to manipulation Incomparability between fields Impact factor not suitable for assessing individual publications and their authors 6

8 Skewness of citation distributions 7

9 Skewness of citation distributions 8

10 San Francisco Declaration on Research Assessment (DORA) 9

11 10

12 11

13 h-index A scientist has index h if h of his papers have at least h citations each and the other papers have at most h citations each 12

14 Common criticism on h-index Undervaluation of highly cited publications Dependence on career length Inflation by co-authored publications Incomparability between fields 13

15 Arbitrariness of the h-index 14

16 Consistency of an indicator If two scientists achieve the same performance improvement, their ranking relative to each other should remain unchanged 15

17 Inconsistency of the h-index Source: Waltman and Van Eck (2012). JASIST, 63(2),

18 Differences in citation density between fields 17

19 Field normalization Professional bibliometrics is focused mainly on the institutional level, where field normalization is essential Standard field normalization approach: Consider a publication from 2010 in an oncology journal The publication has 45 citations On average, publications from 2010 in oncology journals have 15 citations Normalized citation score of the publication is 45 / 15 = 3 18

20 Adoption of standard field normalization approach 19

21 Limitations of standard field normalization approach Inaccuracies in journal-based field classification systems Multidisciplinary journals Fields are too broad 20

22 Fields are too broad (Clinical Neurology) Neurology Neurosurgery 21

23 Alternative field normalization approaches Publication-level rather than journal-level field classification systems (CWTS Leiden Ranking) Source normalization (SNIP) Flexible field definitions, for instance based on cocitations (Relative Citation Ratio) 22

24 4000 fields in publication-level classification system Social sciences and humanities Mathematics and computer science Biomedical and health sciences Life and earth sciences Physical sciences and engineering23

25 CWTS Leiden Ranking 2016 ( 24

26 Source normalization (SNIP) 25

27 Relative Citation Ratio 26

28 Challenges in constructing field normalized indicators Accuracy: Indicators should provide accurate field-normalized citation scores Transparency/simplicity: Indicators should be transparent and easy to explain Validity: Being cited should never be harmful Indicators should behave consistently 27

29 Good practices 28

30 Diana Hicks, Paul Wouters, Ludo Waltman, Sarah de Rijcke, and Ismael Rafols Nature, April 23,

31 Keep analytical processes open, transparent and simple 32

32 Keep analytical processes open, transparent and simple Using non-transparent indicators to support expert judgment is difficult Documenting an indicator in a scientific paper doesn t give full transparency Citizen bibliometrics may be more transparent than professional bibliometrics 33

33 Field normalization based on 4000 publication-level fields: Transparent? Social sciences and humanities Mathematics and computer science Biomedical and health sciences Life and earth sciences Physical sciences and engineering34

34 Recognize systemic effects of indicators: Impact factor Source: Hicks et al. (2015). Nature, 520,

35 Recognize systemic effects of indicators: Impact factor % of journals with more than threefold overrepresentation of self-citations to past two years 36

36 Future developments 37

37 New data sources: Altmetrics 38

38 New data source: Full text Source: Bertin et al. (2016). JASIST, 67(1),

39 New data sources: Full text Source: Rahulja et al. (in press). Natural Language Engineering. 40

40 Contextualized scientometrics: Beyond just indicators Indicators should be used to complement expert judgment Complex black-box indicators tend to replace expert judgment rather than to complement it Indicators should be presented together with contextual information; one should be able to see what is behind the numbers 41

41 Contextualized scientometrics: What is behind the numbers? 42

42 Contextualized scientometrics: What is behind the impact factor? PNAS EMBO Journal 43

43 Contextualized scientometrics: Altmetric dashboard 44

44 Contextualized scientometrics: Visual user interfaces 45

45 Thank you for your attention! More information: Contact: 46