期刊影响因子分析 汤姆森科技信息集团中国办事处50 years ago…利用引文分析进行科研评估的理论基础: Normative Perspective on citations 美国社会学家罗伯特· 默顿(Robert K. Merton): 《科学的规范结构》文献引用的动机Citation motivation: 原文献的科学价值Citation –Impact, Influence, Importance, Performance 科学期刊在同行科学家的交流中起着根本的作用。即科学文献能够代表科学活动以及表现不同分支学科之间关系。在整个期刊系统中，可以区分出“重要”的国际性期刊学术文献被重要的期刊文献引用的次数是这一文献的影响或国际显著性的一种度量Moed et al., The use of bibliometric data for the measurement of university research performance, Research Policy, 14, 131-149,1985; Oppenheim, C. Journal of Documentation, 1995 Web of Science and its offspring Web of Science and its offspring Article-level vs. Journal-level data Journal Citation Reports 独特的多学科期刊分析评价工具自1997 年以来的数据：提供基于引文数据的期刊评价指标显示出引用和被引期刊之间的相互关系帮助用户以定量的方式了解全球的学术期刊利用JCR on Web 深入了解学术期刊Calculation of Journal Impact Factor Journal Impact Factor Average citation count in a particular year Assigned at “journal level”Assigned to journals indexed in Citation Indexes SCIE (Science Citation Index Expanded) SSCI (Social Sciences Citation Index) Included in Journal Citation Reports database Updated annually Helpful when evaluating a journal’s relative importance Not intended to evaluate individual papers/researchers Why Impact Factor? Comparison of scholarly journals Journal collection development in a library Decision-making support for article submission by a researcher Editorial policy development bya journal publisher Problems in SIMPLE comparison of citation counts Size (# of articles published in a year) Frequency (# of issues published in a year) Web of Science record example What you can tell in Journal Citation Reports Which journals publish most number of articles? Which journals are most cited in particular disciplines? Which journals have “high impact”in particular disciplines? Which journals publish “hot”topics? Journal’s self-citations Review article rates Cited-citing relationships of journals Recognize skewed nature of citation data Citation distributions are not normal (the typical bell-shaped curve) Apparent at every level of analysis Citation distributions are extremely skewed: few papers or people are cited at high frequency, most cited little Misuse of “Impact Factor”Essential Science IndicatorsSM Benchmarking data for papers Benchmarking data for papers Compare like with like THIS IS THE GOLDEN RULE OF CITATION ANALYSIS Always compare “apples with apples, not apples with oranges”Field averages are different, by 10:1 Time affects citation counts Summary Citation Indexes Web of Science Journal selection process Impact Factor Included in JCR Average citation count in a particular year for a journal Only account for recent years Helpful in evaluating a journal’s relative importance Not intended to evaluate individual papers/researchers Characteristics of citation data Different citation patterns by disciplines Skewed distribution Benchmarks for research evaluation Included in ESI Average citation rates Percentiles Match against with the actual citation counts No simplistic ways for research evaluation by citation analysis Use citation data to COMPLEMENT peer review process Thank you! 谢谢！http://www.book118.comwww.book118.com/papers/science_v122v3159p108y1955.html new approach to subject control of the literature of science 学術論文の引用関係を手がかりに情報を検索する画期的なツールとして、今からちょうど50 年前、ISI 社の創業者であるEugene Garfield によってサイテーションインデックスが提案されました。「科学のための引用索引」と題されたこの記念碑的な論文の副題は、「アイディアの関連付けを利用した新しい情報検索」というものでした。この論文の中でGarfield は、引用索引のさまざまな利点や特徴をあげていますが、その根底にある発想は、学術論文における引用が「著者自身による文献の関連付け」であるというものです。キーワードや分野分類は時代によって移り変わり、また、索引作成者（indexer ）の恣意的な判断にも左右されます。論文の著者自身が引用する文献は、著者の思考の流れやアイディアを直接的に反映しており、それをたどることによって、科学技術文献をより効果的効率的に検索し、利用できるとGarfiled は考えたのです。Journal citation data - examples Characteristics of citation data Journal citation data –other than Impact Factor 2 step maps Whether one is reviewing the papers of an individual researcher, those of a research team, papers in a single journal or group of journals, those of a specific field in a given year, or those of an entire university, lab, or industrial firm, the citation distribution of the dataset will be highly skewed. That is, a small number of papers in the population will be highly cited and the large majority will be cited little or not at all. This should not cause surprise –it is the nature of these data at every level of analysis. Even the papers of highly cited scientists and Nobel laureates exhibit this type of distribution. It is a rich area of study that has yet to be fully mined or explained, but such distributions –which are anything but so-called normal ‘bell curve’distributions –seem to be common to cases where human choice is at play. A related phenomenon is the concept of criticality. At some critical point –somewhat mysterious because has yet to be defined mathematically with a single equation –a paper achieves enough citations that other citations to it seem to accelerate. Then the paper, in citation terms, takes off on a different and higher trajectory. Another fascinating topic for the mathematically inclined and one that has begun to attract the attention of theoretical physicists. Here I reach the heart of my rules and this presentation. The methodology for a bibliometric analysis must always compare like with like, or as we say, “apples with apples, not apples with oranges.”Different fields of research exhibit quite different citation rates or averages, and the difference can be as much as 10:1. The average 10-year-old paper in molecular biology and genetics may collect 40 citations, whereas the average 10-year-old paper in a computer science journal may garner a relatively modest 4 citations. Even within the same field, one should not compare absolute citation counts of an eight-year-old paper with those of a two-year-old paper, since the former has had more years to collect citations than the latter. Likewise, there is little sense in comparing the thick publication dossier of a researcher who has been publishing for 30 years and runs a large laboratory with the handful of recently published papers from a newly minted Ph.D. in the same field. This is all really no more than common sense. Still, comparing like with like is, for me, the “golden rule”of citation analysis. * Journal Citation Reports - Journal impact factor Essential Science Indicators Citation and citation analysis Citation Analysis via Peer Assessment Garfield, E. “Citation indexes for science: A new dimension in documentation through association of ideas.”Science, 122 (3159), p.108-111, July 1955. Cited Referencesassociation of ideas of its author Citation Indexes  new approach to subject control of the literature of science 2004 Cited References 2000 1998 2001 1993 2003 Times Cited 2003 2004 Cited References Paper/Article Citing papers How to define “citation”? 引文分析的前提Web of Science Journal Citation Reports Journal Impact Factor Stats by journal Stats by address Essential Science Indicators Baselines Customize Research Services Group National Science IndicatorsInstitutional Citation ReportsTopic Citation Reportsetc…Science Citation IndexExpanded Web of Science Journal Citation Reports Journal Impact Factor Stats by journal Example: 2004 Impact Factor for New England Journal of Medicine 2002 2003 2004 366 papers published in 2003 14147 times cited in 2004 378 papers published in 2002 14549 times cited in 2004 378+366 14549+14147 ＝744 28696 ≒　38.570 Journal Citation Reports 2004 Science Edition Impact Factors only take the recent 3 years into consideration. What about the fields with a longer