THE ROLE OF CITATION ANALYSIS IN RESEARCH EVALUATION

Philip Purnell

September 2010

HOW DO WE EVALUATE RESEARCH?

• Research grants– Number and value

• Prestigious awards – Nobel Prizes

• Patents– Demonstrating innovative research

• Faculty– Number of post-graduate researchers

• Citation analysis– Publication and citation counts

– Normalised by benchmarks

• Peer Evaluation– Expensive, time consuming and subjective

A BRIEF HISTORY OF THE CITATION INDEX

• Concept first developed by Dr Eugene Garfield – Science, 1955

• The Science Citation Index (1963)– SCI print (1960’s)– On-line with SciSearch in the 1970’s – CD-ROM in the 1980’s– Web interface (1997) Web of Science

• Content enhanced:– Social Sciences Citation Index (SSCI)– Arts & Humanities Citation Index (AHCI)

• The Citation Index – Primarily developed for purposes of information retrieval– Development of electronic media and powerful searching tools

have increased its use and popularity for purposes of Research Evaluation

WEB OF SCIENCE JOURNAL SELECTION POLICY

• Why do we select journals?

WHY NOT INDEX ALL JOURNALS?

0

20

40

60

80

100

120

0 1000 2000 3000 4000 5000 6000

# of journals

% o

f dat

abas

e

Articles Citations

40% of the journals:

• 80% of the publications

• 92% of cited papers

4% of the journals:

• 30% of the publications

• 51% of cited papers

HOW TO DECIDE WHICH JOURNALS TO INDEX

• Approx. 2000 journals evaluated annually

– 10-12% accepted

• Thomson Reuters editors– Information professionals

– Librarians

– Experts in the literature of their subject area

Web of Science

Journals under evaluation

Journal ‘quality’

THOMSON REUTERSJOURNAL SELECTION POLICY

• Publishing Standards– Peer review, Editorial conventions

• Editorial content– Addition to knowledge in specific subject field

• Diversity– International, regional influence of authors, editors, advisors

• Citation analysis– Editors and authors’ prior work

Region # Journals from Region in Web of Science

Europe 5,573 49%

North America 4,251 38%

Asia-Pacific 965 9%

Latin America 272 2%

Middle East/Africa 200 1%

Language # Journals in Web of Science

English 9114 81%

Other 2147 19%

GLOBAL RESEARCH REPRESENTATIONWEB OF SCIENCE COVERAGE

Analyses based on authoritative, consistent data from the world’s leading provider of Research Evaluation solutions

Thomson Reuters has developed a selection policy over the last 50 years designed to hand-pick the relevant journals containing the core content over the full range of scholarly disciplines

This has created a large set of journals containing comparable papers and citations

Thomson Reuters has always had one consistent editorial policy to index all journals cover-to-cover, index all authors and index all addresses. This unique consistency makes Web of Science the only suitable data source for citation analysis

SUMMARYCONSISTENCY IS THE KEY TO VALIDITY

GOVERNMENTS AND INSTITUTIONS USING TR DATA FOR EVALUATION (INCL.)

• Germany: IFQ, Max Planck Society, DKFZ, MDCUS

• Netherlands: NWO & KNAW

• France: Min. de la Recherche, OST - Paris, CNRS

• United Kingdom: King’s College London; HEFCE

• European Union: EC’s DGXII(Research Directorate)

• US: NSF: biennial Science & Engineering Indicators report (since 1974)

• Canada: NSERC, FRSQ (Quebec), Alberta Research Council

• Australian Academy of Science, gov’t lab CSIRO

• Japan: Ministry of Education, Ministry of Economy, Trade & Industry

• People’s Republic of China: Chinese Academy of Science

• Times Higher Education: World University Rankings (from 2010)

10

EVALUATING COUNTRIES

SCIENTIFIC RESEARCH IMPACT IN CENTRAL EUROPE

12Thomson Reuters InCites

OUTPUT AND PRODUCTIVITYBULGARIAN RESEARCH 1998 - 2008

13

COMPARATIVE IMPACT IN SELECTED FIELDSBETWEEN COUNTRIES

14Source: Thomson Reuters InCites

BULGARIAN RESEARCH RELATIVE PRODUCTIVITY BY FIELD

15

22% Bulgarian papers are in Chemistry

Source: Thomson Reuters InCites

<1% Bulgarian papers are in Psychiatry

EVALUATING INSTITUTIONS

EVALUATING INSTITUTIONS

Source: Thomson Reuters

North America University Science Indicators

CITATIONS PER PAPERMATHEMATICS

18Source: Thomson Reuters InCites

COMPARISON OF TOP MATHEMATICS INSTITUTES AROUND THE WORLD

19Source: Thomson Reuters InCites

WITH WHOM DOES OUR FACULTY COLLABORATE?

20Source: Thomson Reuters InCites

WHICH COLLABORATIONS ARE THE MOST VALUABLE?

21

Collaborations with these institutions have produced highly cited papers within their subject fields

Source: Thomson Reuters InCites

EVALUATING JOURNALS

CALCULATING 2009 IMPACT FACTOR - JOURNAL OF CONTAMINANT HYDROLOGY      

Citations in 2009

To items published in 2008 = 153

To items published in 2007 = 239

Sum = 392

Number of items

Published in 2008 = 97

Published in 2007 = 98

Sum = 195

392

195

= 2,01

JOURNAL IMPACT FACTORSELECTED CHEMISTRY JOURNALS

24Thomson Reuters Journal Citation Reports

25

USING THE IMPACT FACTOREVALUATING JOURNALS

• Appropriate use– To evaluate journals within a subject field

• Misuse– Comparison of journals from different fields

– Evaluation of individual articles

– Evaluation of institution or researcher

26

USING THE IMPACT FACTOR MISUSE: EVALUATING INDIVIDUAL PAPERS

30% of articles in Food Policy were not cited at all

Journal Impact Factor = 2,01

BENCHMARK YOUR PAPERS AGAINST GLOBAL AVERAGES – IS THIS A HIGHLY CITED PAPER?

27

Hematology articles from this year have been cited 18,83 times

This article is ranked in the 12,92nd percentile in its field by citations

Articles published in ‘Blood’ from 2004 have been cited 34,30 times

This paper has received 40/34,30=1,17 times the expected citations for this journal

This paper has received 40/18,83=2,12 times the expected citations for this subject category

EVALUATING INDIVIDUALS

HOW CAN WE COMPARE RESEARCHERS?

29

Author A: 60 papers Author B: 117 papers

Source: Thomson Reuters InCites

OBTAIN MULTIPLE MEASURES

RECOGNIZE THE SKEWED NATURE OF CITATION DATA

• Citation distribution is always skewed– Few highly cited papers

– Majority cited little or not at all

• Distribution type– Always distorted

– Human decision• E.g. Criticality

SUMMARY (I): TREAT AS A SCIENTIFIC STUDY

• Ask whether the results are reasonable

• Follow scientific process for evaluating data

• Apply scientific skepticism

SUMMARY (II): HOW DO WE EVALUATE RESEARCH?

• Research grants– Number and value

• Prestigious awards – Nobel Prizes

• Patents– Demonstrating innovative research

• Faculty– Number of post-graduate researchers

• Citation analysis– Publication and citation counts

– Normalised by benchmarks

• Peer Evaluation– Expensive, time consuming and subjective

THANK YOU

Philip Purnell

September 2010