Measuring Quality and Impact of the Social Sciences Concepts, Opportunities and Drawbacks Pre-Conference of the 10 th International Conference on Science and Technology Indicators University of Vienna, September 17, 2008 Anthony F.J. van Raan Center for Science and Technology Studies (CWTS) Leiden University
102
Embed
Measuring Quality and Impact of the Social Sciences Concepts, Opportunities and Drawbacks Pre-Conference of the 10 th International Conference on Science.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Measuring Quality and Impact of the Social Sciences
Concepts, Opportunities and Drawbacks
Pre-Conference of the 10th International Conference on Science and Technology Indicators
University of Vienna, September 17, 2008
Anthony F.J. van Raan Center for Science and Technology Studies (CWTS)
Leiden University
This presentation will highlight recent CWTS projects:
* Benchmarking & Evaluation* HEFCE * Identification of Excellence
From these recent studies we present empirical results for social science fields particularly concerning:
* WoS coverage * Characteristics of WoS publications* Characteristics of n-WoS publications* Bibliometric results and peer judgments
First the basic principles of bibliometric analysis
Basic Concept: Quality
Scientific performance relates to achieved quality in the contribution to the increase of our knowledge (‘scientific progress’)
(1) as perceived by others: peer review (2) as measured by advanced bibliometric analysis
Basic issues for research assessment, also in the social sciences:
* Objectivity* Transparency* How to handle interdisciplinarity, definition of fields* Different ways, prestige and intensity of publication* Role of co-authors in publications* Orientation of research: local vs. global* Language* Ageing of research results* PhD training* Time dimension of awards* Socio-economic impact
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free”Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and LuísA. NunesAmaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,UniversitàdiRoma “La Sapienza,”PiazzaleAldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incominglinks to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible”to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on thedynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions thatdecay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment”[10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free”Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and LuísA. NunesAmaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,UniversitàdiRoma “La Sapienza,”PiazzaleAldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incominglinks to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible”to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on thedynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions thatdecay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment”[10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free”Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and LuísA. NunesAmaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,UniversitàdiRoma “La Sapienza,”PiazzaleAldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incominglinks to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible”to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on thedynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions thatdecay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment”[10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free”Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and LuísA. NunesAmaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,UniversitàdiRoma “La Sapienza,”PiazzaleAldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incominglinks to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible”to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on thedynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions thatdecay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment”[10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free”Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and LuísA. NunesAmaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,UniversitàdiRoma “La Sapienza,”PiazzaleAldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incominglinks to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible”to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on thedynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions thatdecay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment”[10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
Weight?
networks leading, possibly, to different dynamics, e.g., for the initiation and spread of epidemics.In the context of network growth, the impossibility of knowing the degrees of all the nodes comprising the network due to the filtering process—and, hence, the inability to make the optimal, rational, choice—is not altogether unlike the “bounded rationality” concept of Simon [17].Remarkably, it appears that, for the description of WWW growth, the preferential attachment mechanism, originally proposed by Simon [10], must be modified along the lines of another concept also introduced by him—bounded rationality [17].
We thank R. Albert, P. Ball, A.-L. Barabási, M. Buchanan, J. Camacho, and R. Guimerà for stimulating discussions and helpful suggestions. We are especially grateful to R. Kumar for sharing his data. We thank NIH/NCRR (P41 RR13622) and NSF for support.
[1] S. H. Strogatz, Nature (London) 410, 268 (2001).[2] R. Albert and A.-L. Barabási, Rev. Mod. Phys. 74, 47 (2002).[3] S. N. Dorogovtsev and J. F. F. Mendes, Adv. Phys. (to be published).[4] R. Albert, H. Jeong, and A.-L. Barabási, Nature (London) 401, 130 (1999).[5] B. A. Huberman and L. A. Adamic, Nature (London) 401, 131 (1999); R. Kumar et al., in Proceedings of the 25th International Conference on Very Large Databases (Morgan Kaufmann Publishers, San Francisco, 1999), p. 639; A. Broder et al., Comput. Netw. 33, 309 (2000); P. L. Krapivsky, S. Redner, and F. Leyvraz, Phys. Rev. Lett. 85, 4629 (2000); S. N. Dorogovtsev, J. F. F. Mendes, and A. N. Samukhin, ibid. 85, 4633 (2000); A. Vazquez, Europhys. Lett. 54, 430 (2001).[6] M. Faloutsos, P. Faloutsos, and C. Faloutsos, Comput. Commun. Rev. 29, 251 (1999); G. Caldarelli, R. Marchetti, and L. Pietronero, Europhys. Lett. 52, 386 (2000); A. Medina, I. Matta, and J. Byers, Comput. Commun. Rev. 30, 18 (2000); R. Pastor-Satorras, A. Vazquez, and A. Vespignani, arXiv:cond-mat/0105161; L. A. Adamic et al., Phys. Rev. E 64, 046135 (2001).[7] F. B. Cohen, A Short Course on Computer Viruses (Wiley, New York, 1994); R. Pastor-Satorras and A. Vespignagni, Phys. Rev. Lett. 86, 3200 (2001); Phys. Rev. E 63, 066117 (2001).[8] F. Liljeros, C. R. Edling, L. A. Nunes Amaral, H. E. Stanley, and Y. Åberg, Nature (London) 411, 907 (2001).[9] A.-L. Barabási and R. Albert, Science 286, 509 (1999).[10] Y. Ijiri and H. A. Simon, Skew Distributions and the Sizes of Business Firms (North-Holland, Amsterdam, 1977).[11] G. Bianconi and A.-L. Barabasi, Europhys. Lett. 54, 436 (2001).[12] A. F. J. Van Raan, Scientometrics 47, 347 (2000).[13] We consider a modification to the network growth rule described earlier in the paper: at each time step t, the new node establishes m new links, where m is drawn from a power law distribution with exponent gout.[14] For n(I) = const, one recovers the scale-free model of Ref. [9].[15] It is known [11] that, for an exponential or fat-tailed distribution of fitness, the structure of the network becomes much more complex; in particular, the in-degree distribution is no longer a power law. Hence, we do not consider in this manuscript other shapes of the fitness distribution.[16] L. A. N. Amaral, A. Scala, M. Barthélémy, and H. E. Stanley, Proc. Natl. Acad. Sci. U.S.A. 97, 11 149 (2000).[17] H. A. Simon, Models of Bounded Rationality: Empirically Grounded Economic Reason (MIT Press, Cambridge, 1997).
Citing Publications
Cited Publications
All calculations are corrected for self-citations!
What do citations measure?
- Many studies showed positive correlations between citations and qualitative judgments
- In principle it is valid to interpret citations in terms of intellectual influence which is an important aspect of scientific quality
- Thus, the concepts of citation impact and scientific quality do not coincide ‘automatically’
WoS sub-universe8,000 j; 1,000,000p/y
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free” Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and Luís A. Nunes Amaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,Università di Roma “La Sapienza,” Piazzale Aldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incoming links to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible” to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on the dynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions that decay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment” [10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free” Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and Luís A. Nunes Amaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,Università di Roma “La Sapienza,” Piazzale Aldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incoming links to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible” to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on the dynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions that decay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment” [10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free” Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and Luís A. Nunes Amaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,Università di Roma “La Sapienza,” Piazzale Aldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incoming links to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible” to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on the dynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions that decay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment” [10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free” Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and Luís A. Nunes Amaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,Università di Roma “La Sapienza,” Piazzale Aldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incoming links to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible” to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on the dynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions that decay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment” [10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
Refs > nWoS
Compendex
Medline
non-WoS publ:
Books
Book chapters
Conf. proc.
Reports
ArXiv
Total publ universe
LNCS
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free” Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and Luís A. Nunes Amaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,Università di Roma “La Sapienza,” Piazzale Aldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incoming links to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible” to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on the dynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions that decay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment” [10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free” Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and Luís A. Nunes Amaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,Università di Roma “La Sapienza,” Piazzale Aldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incoming links to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible” to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on the dynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions that decay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment” [10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free” Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and Luís A. Nunes Amaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,Università di Roma “La Sapienza,” Piazzale Aldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incoming links to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible” to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on the dynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions that decay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment” [10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
VOLUME 88, Number 13 PHYSICAL REVIEW LETTERS 1 April 2002
Truncation of Power Law Behavior in “Scale-Free” Network Modelsdue to Information Filtering
Stefano Mossa,1,2 Marc Barthélémy,3 H. Eugene Stanley,1 and Luís A. Nunes Amaral11 Center for Polymer Studies and Department of Physics, Boston University, Boston, Massachusetts 02215
2 Dipartimento di Fisica, INFM UdR, and INFM Center for Statistical Mechanics and Complexity,Università di Roma “La Sapienza,” Piazzale Aldo Moro 2, I-00185, Roma, Italy
3 CEA-Service de Physique de la Matière Condensée, BP 12, 91680 Bruyères-le-Châtel, France(Received 18 October 2001; published 14 March 2002)
We formulate a general model for the growth of scale-free networks under filtering information conditions—that is, when the nodes can process information about only a subset of the existing nodes in the network. We find that the distribution of the number of incoming links to a node follows a universal scaling form, i.e., that it decays as a power law with an exponential truncation controlled not only by the system size but also by a feature not previously considered, the subset of the network “accessible” to the node. We test our model with empirical data for the World Wide Web and find agreement.
There is a great deal of current interest in understanding the structure and growth mechanisms of global networks [1–3], such as the World Wide Web (WWW) [4,5] and the Internet [6]. Network structure is critical in many contexts such as Internet attacks [2], spread of an Email virus [7], or dynamics of human epidemics [8]. In all these problems, the nodes with the largest number of links play an important role on the dynamics of the system. It is therefore important to know the global structure of the network as well as its precise distribution of the number of links.Recent empirical studies report that both the Internet and the WWW have scale-free properties; that is, the number of incoming links and the number of outgoing links at a given node have distributions that decay with power law tails [4–6]. It has been proposed [9] that the scale-free structure of the Internet and the WWW may be explained by a mechanism referred to as “preferential attachment” [10] in which new nodes link to existing nodes with a probability proportional to the number of existing links to these nodes. Here we focus on the stochastic character of the preferential attachment mechanism, which we understand in the following way: New nodes want to connect to the existing nodes with the largest number of links—i.e., with the largest degree—because of the advantages offered by being linked to a well-connected node. For a large network it is not plausible that a new node will know the degrees of all existing nodes, so a new node must make a decision on which node to connect with based on what information it has about the state of the network. The preferential attachment mechanism then comes into play as nodes with a larger degree are more likely to become known.
Source expansion
Target expansion
Scopus*CWTS is in license agreement negotiations with Scopus*CWTS currently compares Scopus- vs. WoS coverage*CWTS bibliometric algorithms can be applied
to Scopus data
Google Scholar
Network of publications (nodes)
linked by citations (edges)
Lower citation-density Higher citation-density
e.g., applied research, e.g., basic natural social sciences medical research
FCSmJCSm
CPP Expected values for normalization
Absolutely necessary but……are they appropriate?
Journal
Field = set of journals‘established fields’
scientific medium-grained structure
+ reference-based re-definition (expansion) of fields
CWTS applies two types of field definitions:
fields
ACTA GENETICAE MEDICAE ET GEMELLOLOGIAEAMERICAN JOURNAL OF HUMAN GENETICSAMERICAN JOURNAL OF MEDICAL GENETICSANIMAL BLOOD GROUPS AND BIOCHEMICAL GENETICSANNALES DE GENETIQUEANNALES DE GENETIQUE ET DE SELECTION ANIMALEANNALS OF HUMAN GENETICSATTI ASSOCIAZIONE GENETICA ITALIANABEHAVIOR GENETICSBIOCHEMICAL GENETICSCANADIAN JOURNAL OF GENETICS AND CYTOLOGYCANCER GENETICS AND CYTOGENETICSCARYOLOGIACHROMOSOMACLINICAL GENETICSCURRENT GENETICSCYTOGENETICS AND CELL GENETICSCYTOLOGIADEVELOPMENTAL GENETICSENVIRONMENTAL MUTAGENESISEVOLUTIONGENEGENETICAGENETICA POLONICAGENETICAL RESEARCHGENETICSGENETIKAHEREDITASHEREDITYHUMAN BIOLOGYHUMAN GENETICSHUMAN HEREDITYIMMUNOGENETICSINDIAN JOURNAL OF GENETICS AND PLANT BREEDINGJAPANESE JOURNAL OF GENETICSJAPANESE JOURNAL OF HUMAN GENETICSJOURNAL DE GENETIQUE HUMAINEJOURNAL OF HEREDITYJOURNAL OF IMMUNOGENETICSJOURNAL OF MEDICAL GENETICSJOURNAL OF MENTAL DEFICIENCY RESEARCHJOURNAL OF MOLECULAR EVOLUTIONMOLECULAR & GENERAL GENETICSMUTATION RESEARCHPLASMIDSILVAE GENETICATHEORETICAL AND APPLIED GENETICSTHEORETICAL POPULATION BIOLOGYEGYPTIAN JOURNAL OF GENETICS AND CYTOLOGYREVISTA BRASILEIRA DE GENETICAANNUAL REVIEW OF GENETICSJOURNAL OF CRANIOFACIAL GENETICS AND DEVELOPMENTAL BIOLOGYJOURNAL OF INHERITED METABOLIC DISEASEPRENATAL DIAGNOSISADVANCES IN GENETICS INCORPORATING MOLECULAR GENETIC MEDICINECHEMICAL MUTAGENS-PRINCIPLES AND METHODS FOR THEIR DETECTIONDNA-A JOURNAL OF MOLECULAR & CELLULAR BIOLOGYEVOLUTIONARY BIOLOGYTERATOGENESIS CARCINOGENESIS AND MUTAGENESISTSITOLOGIYA I GENETIKAADVANCES IN HUMAN GENETICSPROGRESS IN MEDICAL GENETICSGENETICS SELECTION EVOLUTIONMOLECULAR BIOLOGY AND EVOLUTIONSOMATIC CELL AND MOLECULAR GENETICSBIOTECHNOLOGY & GENETIC ENGINEERING REVIEWSEXPERIMENTAL AND CLINICAL IMMUNOGENETICSGENE ANALYSIS TECHNIQUESJOURNAL OF MOLECULAR AND APPLIED GENETICSJOURNAL OF NEUROGENETICSTRENDS IN GENETICSDISEASE MARKERSANIMAL GENETICSGENETIC EPIDEMIOLOGYJOURNAL OF GENETICS
Main field: Social and Behavioral Sciences
SOCI AL AND BEHAVI ORAL SCI ENCES
ECONOMICS AND BUSINESS AGRICULTURAL ECONOMICS & POLICY BUSINESS BUSINESS, FINANCE ECONOMICS INDUSTRIAL RELATIONS & LABOR
EDUCATIONAL SCIENCES EDUCATION & EDUCATIONAL RESEARCH EDUCATION, SCIENTIFIC DISCIPLINES EDUCATION, SPECIAL PSYCHOLOGY, EDUCATIONAL
MANAGEMENT AND PLANNING AREA STUDIES MANAGEMENT PLANNING & DEVELOPMENT
POLITICAL SCIENCE AND INTERNATIONAL RELATIONSPUBLIC ADMINISTRATION POLITICAL SCIENCE
SOCIAL AND BEHAVIORAL SCIENCES, INTERDISCIPLINARY DEMOGRAPHY
SOCIAL ISSUES SOCIAL SCIENCES, BIOMEDICAL SOCIAL SCIENCES, INTERDISCIPLINARY
SOCIOLOGY AND ANTHROPOLOGY ANTHROPOLOGY ETHNIC STUDIES FAMILY STUDIES SOCIOLOGY WOMEN'S STUDIES
Major field, e.g. Economics & Business
journals
APPL PREV PSYCHOLAPPL PSYCHOL-INT REVBEHAV SCI LAWBRIT J GUID COUNSCAREER DEV QCOUNS PSYCHOLCYBERPSYCHOL BEHAVERGONOMICSEUR J PSYCHOL ASSESSEUR J WORK ORGAN PSYGROUP ORGAN MANAGEHUM FACTORSHUM PERFORMHUM RESOUR MANAGEINT J AVIAT PSYCHOLINT J OFFENDER THERINT J SELECT ASSESSJ APPL PSYCHOLJ APPL SPORT PSYCHOLJ BEHAV DECIS MAKING
All publication titles + abstracts (~30,000,000) have been grammatically parsed to enable bibliometric analysis by themes/concepts/ instruments and to create word-correlation based maps of science
clusterField = clusters of concept-related
publicationsnew, emerging often interdisc. fields
scientific fine-grained structure
Social Sciences Top-50 EU universities, their top-10% publications in this field
Now specific sub-field CPP/FCSm values can be calculated, for instance for research on democracy
But, obviously, the finer grained, the more ‘noisy’
Basic Performance Indicators
•P Ouput: Number of publications in internationally refereed CI-covered journals
•C Absolute Impact: Number of (self-ex) citations to these publications
•H Hirsch-index•CPP Output-normalized Impact: Average number of
cits/pub of the institute •JCSm Average number of cits/pub of the journal set
used by the institute•FCSm Average number of cits/pub of all journals of a
specific field in which the institute is active (FCSm)
•p0 Percentage of not-cited publications
CWTS Key Research Performance Indicators:
• JCSm/FCSm Relative impact of the used journal set
• CPP/JCSm Internat. journal-normalized impact• CPP/FCSm Internat. field & doc-normalized
impact
• Pt/Πt Contribution to the top-5, 10, 20,..%• P*CPP/FCSm Size & Impact Together: Brute Force
Applied research, engineering
Basic research
high FCSm
Up to factor ~20
high FCSm, but low JCSm
low FCSm
low FCSm, but high JCSm
High CPP
low CPP
Internal WoS-coverage of social science fields
results from HEFCE and Benchmark projects
Table 3.1: Internal coverage percentages of the Thomson Scientific/ ISI Citation Indexes
I nternal Coverage Percentage 80-100% 60-80% 40-60% <40% Biochem & Mol Biol Appl Phys & Chem Mathematics Other Soc Sci Biol Sci – Humans Biol Sci – Anim & Plants Economics Humanities & Arts Chemistry Psychol & Psychiat Engineering Clinical Medicine Geosciences Phys & Astron Soc Sci ~ Medicine
Internal WoS coverage of main fields of science
major field P 00-04 Avg Nr Refs Refs<1980 % Refs<1980 Refs Non-CI Refs CI % Refs CI
Internal WoS coverage (%) of submitted publications per UoA
SS&H Sports-related Subjects 698 65 SS&H Economics and Econometrics 1,688 51 SS&H Accounting and Finance 164 40 SS&H Geography 2,697 39 SS&H Business and Management Studies 3,205 36 SS&H Anthropology 299 35 SS&H Library and Information Management 366 35 SS&H Built Environment 560 34 SS&H Archaeology 327 31 SS&H Linguistics 199 30 SS&H Art and Design 194 28 SS&H Social Work 344 27 SS&H Sociology 933 24 SS&H Education 1,230 24 SS&H Social Policy and Administration 948 24 SS&H Town and Country Planning 520 23 SS&H Politics and International Studies 1,040 18
From: Moed, Visser, Buter, 2008
Table 3.3: Estimated CI-coverage (1991-2006), based on the extent to which references within CI -covered publications are also CI -covered Field 1991 1996 2001 2006 MEDI CAL & LIFE SCIENCES AGRICULTURE AND FOOD SCIENCE 66% 66% 73% 75% BASIC LIFE SCIENCES 87% 89% 93% 93% BASIC MEDICAL SCIENCES 76% 75% 80% 84% BIOLOGICAL SCIENCES 72% 74% 80% 82% BIOMEDICAL SCIENCES 86% 87% 90% 90% CLINICAL MEDICINE 82% 82% 85% 85% HEALTH SCIENCES 50% 47% 57% 62% NATURAL SCI ENCES ASTRONOMY AND ASTROPHYSICS 75% 79% 82% 86% CHEMISTRY AND CHEMICAL ENGINEERING 77% 80% 86% 88% COMPUTER SCIENCES 38% 37% 42% 43% EARTH SCIENCES AND TECHNOLOGY 60% 60% 69% 74% ENVIRONMENTAL SCIENCES AND TECHNOLOGY 46% 46% 55% 62% MATHEMATICS 58% 57% 58% 64% PHYSICS AND MATERIALS SCIENCE 75% 78% 81% 84% STATISTICAL SCIENCES 49% 46% 52% 58% ENGI NEERING SCIENCES CIVIL ENGINEERING AND CONSTRUCTION 37% 33% 34% 45% ELECTRICAL ENGINEERING AND TELECOMMUNICATION 54% 52% 52% 53% ENERGY SCIENCE AND TECHNOLOGY 54% 48% 53% 59% GENERAL AND INDUSTRIAL ENGINEERING 42% 37% 44% 54% INSTRUMENTS AND INSTRUMENTATION 67% 62% 71% 69% MECHANICAL ENGINEERING AND AEROSPACE 58% 53% 57% 64% LANGUAGE, INFORMATI ON AND COMMUNI CATI ON INFORMATION AND COMMUNICATION SCIENCES 32% 28% 29% 32% LANGUAGE AND LINGUISTICS 26% 33% 43% 40% SOCIAL AND BEHAVI ORAL SCI ENCES ECONOMICS AND BUSINESS 35% 36% 35% 43% EDUCATIONAL SCIENCES 27% 31% 30% 36% MANAGEMENT AND PLANNING 23% 24% 27% 36% POLITICAL SCIENCE AND PUBLIC ADMINISTRATION 17% 18% 20% 24% PSYCHOLOGY 59% 59% 66% 72% SOCIAL AND BEHAVIORAL SCIENCES, INTERDISCIPLINARY 33% 34% 36% 40% SOCIOLOGY AND ANTHROPOLOGY 22% 27% 29% 34% LAW, ARTS AND HUMANITIES CREATIVE ARTS, CULTURE AND MUSIC 17% 14% 16% 14% HISTORY, PHILOSOPHY AND RELIGION 24% 23% 25% 27% LAW AND CRIMINOLOGY 27% 32% 32% 31% LITERATURE 14% 12% 11% 11% MULTIDISCIPLINARY JOURNALS 78% 83% 87% 87%
Internal WoS coverage for all main fields of science
LANGUAGE, INFORMATION AND COMMUNICATION INFORMATION AND COMMUNIC SCIENCES 32 28 29 32 LANGUAGE AND LINGUISTICS 26 33 43 40 SOCIAL AND BEHAVIORAL SCIENCES ECONOMICS AND BUSINESS 35 36 35 43 EDUCATIONAL SCIENCES 27 31 30 36 MANAGEMENT AND PLANNING 23 24 27 36 POLITICAL SCIENCE AND PUBLIC ADMIN 17 18 20 24 PSYCHOLOGY 59 59 66 72 SOCIAL AND BEHAV SCIENCES, INTERDISC 33 34 36 40 SOCIOLOGY AND ANTHROPOLOGY 22 27 29 34
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
1991
1996
2001
2006
1991
1996
2001
2006
1991
1996
2001
2006
1991
1996
2001
2006
1991
1996
2001
2006
1991
1996
2001
2006
1991
1996
2001
2006
AGRICULTUREAND FOODSCIENCE
BASIC LIFESCIENCES
BASIC MEDICALSCIENCES
BIOLOGICALSCIENCES
BIOMEDICALSCIENCES
CLINICALMEDICINE
HEALTHSCIENCES
References non-ISI
References ISI
purple: non-WoS reflight blue: CI ref
1991-2006
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
19
91
19
96
20
01
20
06
19
91
19
96
20
01
20
06
19
91
19
96
20
01
20
06
19
91
19
96
20
01
20
06
19
91
19
96
20
01
20
06
19
91
19
96
20
01
20
06
19
91
19
96
20
01
20
06
ECONOMICS ANDBUSINESS
EDUCATIONALSCIENCES
MANAGEMENTAND PLANNING
POLITICALSCIENCE AND
PUBLICADMINISTRATION
PSYCHOLOGY SOCIAL ANDBEHAVIORALSCIENCES,
INTERDISCIPLINARY
SOCIOLOGY ANDANTHROPOLOGY
References non-ISI
References ISI
purple: non-WoS reflight blue: CI ref
1991-2006
External WoS-coverage of social science fields
results from HEFCE and Evaluation projects
TABLE 2: EXTERNAL COVERAGE FOR UPPSALA PAPERS 2002 - 2006
Research Unit Wos papersJournal Papers
Journal Coverage
Total papers (in journals,
procs, books)
Total paper
Coverage
UPPSALA 8.502 11.403 75% 16.436 52%
Arts 61 567 11% 1.089 6%Centre for Gender Research 17 37 46% 77 22%
Centre for Multiethnic Research 1 12 8% 43 2%Dep of ALM (Archives, Libraries, Museums) 3 20 15% 43 7%
Dep of Archeology and Ancient History 5 67 7% 142 4%Dep of Art History 1 12 8% 34 3%
Dep of Cultural Anthropology and Ethnology 6 94 6% 151 4%Dep of History 9 72 13% 152 6%
Dep of History of Science and Ideas 5 77 6% 120 4%Dep of Literature 3 140 2% 242 1%
Dep of Musicology 1 7 14% 18 6%Dep of Philosophy 10 29 34% 67 15%
The Uppsala Progr for Holocaust and Genocide Studies 1 4 25% 9 11%
IMPACT COMPARED TO WORLD SUBFIELD AVERAGE2000 - 2006
ECONOMICS
LSE
NORTHWESTERN
OXFORD
EUR
MANCHESTER
UvT
DUKE
CAMBRIDGEUvA
VU
UM
LEUVEN
VIRGINIA
RUG
PITTSBURGH
IMP COLL
MCGILLJOHNS HOPKINS
GENT
UTRECHT
RU
BARCELONA
TORINO
LAUSANNE
KINGS COLL
BASEL
GENEVE
MILANO
LEIDEN
NAPOLI
0,00
0,50
1,00
1,50
2,00
2,50
3,00
0 100 200 300 400 500 600
TOTAL PUBLICATIONS
Black coloured squares above (below) the horizontal reference line represent groups for which the impact (CPP) is significantly above (below) world average (FCSm)
Top-10% (of impact) of EU publications in Political Science, Economics, and Psychology
1997-2003, 4-y citation window (to calculate their impact)
From references all WoS-references removed, only non-WoS references (with freq > 2) have been analyzed
Total about 28,000
Percentage of references dating prior to 1980
0%
10%
20%
30%
40%
50%
60%
70%
Politicalscience
Economics Psychology
% <1980
% <1980 articles
% <1980 books
% <1980 handbooks
% <1980 theses
From: Nederhof, van Leeuwen, van der Wurff 2008
From these:
Books and journals dominant among post-1980
references to non-WoS items
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Political science Economics Psychology
% Books
% Journals
I tems with small frequencies among post-1980 references to
non-WoS items
% total items (> 1979)
0%
1%
2%
3%
4%
5%
6%
Political science Economics Psychology
% Handbook % Thesis % Proc % Reports % Working P
(> 1980)
Top-ranked publications according to document type
Field Books J.Articles Chapters Manuals Handbook Edit.vol Unknown
Political sc 84% 7% 0% 0% 0% 5% 5%
Economics 89% 3% 0% 0% 3% 0% 6%
Psychology 68% 3% 3% 24% 0% 0% 3%
Top-50 non-WoS >1980 references by document type
Bibliometric results and peer judgments
results from HEFCE and Benchmark & Evaluation projects
Qualitative versus quantitative assessment
peer review reputation may have strong influenceincludes 'tacit knowledge' (e.g., instrument building)includes credits: expectations, we believe that…, ahead of time…takes products other than journals papers into accountfashion and hypes perhaps less influential
bibliometric reputation much less influential analysis only 'codified knowledge'
no credits: only past performance, evidence-basedproducts other than journal papers less importantfashion and hypes perhaps more influential on the short term
TABLE 5: BIBLIOMETRIC STATISTICS FOR DEPARTMENTS AND INSTITUTES 2002 - 2006
CPP/ CPP/ JCSm/ SelfDepartment P C C+sc CPP Pnc JCSm FCSm FCSm Cit
Centre for Image Analysis 22 59 81 2,68 32% 0,84 0,71 0,85 27%Dep of Information Technology 265 589 868 2,22 54% 1,22 1,24 1,01 32%
Dep of Mathematics 167 195 291 1,17 59% 1,13 0,97 0,86 33%
Medicine 3.556 24.034 30.552 6,76 28% 1,08 1,22 1,13 21%
Dep of Genetics and Pathology 535 4.579 5.820 8,56 24% 1,04 1,37 1,31 21%Dep of Medical Biochemistry and Microbiology 509 3.623 4.946 7,12 23% 0,88 1,14 1,30 27%
Dep of Medical Cell Biology 253 1.004 1.467 3,97 34% 0,70 0,74 1,06 32%Dep of Medical Sciences 975 7.376 8.981 7,57 28% 1,20 1,38 1,15 18%
Dep of Neuroscience 490 2.872 3.872 5,86 27% 1,03 1,03 1,00 26%Dep of Oncology, Radiology and Clinical Immunology 524 3.684 4.771 7,03 26% 1,14 1,17 1,03 23%
Dep of Public Heatlh and Caring Sciences 372 1.568 2.030 4,22 35% 1,03 0,92 0,89 23%Dep of Surgical Sciences 443 3.272 3.856 7,39 29% 1,24 1,52 1,22 15%
Dep of Women s and Children s Healthエ エ 196 579 763 2,95 38% 0,77 0,74 0,96 24%
Centre for Image Analysis 22 59 81 2,68 32% 0,84 0,71 0,85 27%Dep of Information Technology 265 589 868 2,22 54% 1,22 1,24 1,01 32%
Dep of Mathematics 167 195 291 1,17 59% 1,13 0,97 0,86 33%
Medicine 3.556 24.034 30.552 6,76 28% 1,08 1,22 1,13 21%
Dep of Genetics and Pathology 535 4.579 5.820 8,56 24% 1,04 1,37 1,31 21%Dep of Medical Biochemistry and Microbiology 509 3.623 4.946 7,12 23% 0,88 1,14 1,30 27%
Dep of Medical Cell Biology 253 1.004 1.467 3,97 34% 0,70 0,74 1,06 32%Dep of Medical Sciences 975 7.376 8.981 7,57 28% 1,20 1,38 1,15 18%
Dep of Neuroscience 490 2.872 3.872 5,86 27% 1,03 1,03 1,00 26%Dep of Oncology, Radiology and Clinical Immunology 524 3.684 4.771 7,03 26% 1,14 1,17 1,03 23%
Dep of Public Heatlh and Caring Sciences 372 1.568 2.030 4,22 35% 1,03 0,92 0,89 23%Dep of Surgical Sciences 443 3.272 3.856 7,39 29% 1,24 1,52 1,22 15%
Dep of Women?s and Children?s Health 196 579 763 2,95 38% 0,77 0,74 0,96 24%
Dep of Business Studies 21 44 52 2,10 62% 0,80 1,23 1,52 15% 4,0Dep of Domestic Sciences 25 117 156 4,68 24% 0,99 0,79 0,80 25% 2,5Dep of Economics 41 75 88 1,83 49% 0,64 0,85 1,34 15% 4,0Dep of Government 27 54 73 2,00 52% 1,44 1,13 0,78 26% 3,5Dep of Information Science 31 347 411 11,19 35% 3,61 2,40 0,66 16% 4,5Dep of Psychology 149 549 737 3,68 42% 0,96 0,99 1,03 26% 3,5Inst for Housing and Urban Research 31 67 89 2,16 35% 0,92 1,16 1,26 25% 4,0
Social Sciences and Humanities
1 5 0.45 0.29 0.40 0.45 0.65
2 58 1.12 1.92 0.18 0.64 1.24
3a 291 0.92 1.36 0.17 0.68 1.20
3b 145 0.82 1.23 0.00 0.62 1.10
4 366 1.02 1.80 0.37 0.84 1.31
5 389 1.26 1.30 0.53 1.03 1.59
5* 151 1.34 1.27 0.64 1.09 1.68
Normalised citation impact parameters per subject group and per rating Normalised Citation Impact Distribution
RATING # Depts MEAN STD P25 MEDIAN P75
Comparison WoS vs. Scopus
results from one of the HEFCE projects: see tomorrow Martijn Visser and Henk Moed: “Comparing Web of Science and Scopus on a paper-by-paper basis”
Conclusion
Advanced bibliometric analysis is a powerful tool to make research assessment more objective, transparent and effective, particularly in the natural science and medical fields, and also in many of the engineering and social science fields but both internal and external WoS/Scopus coverage are absolutely necessary parameters to assess the validity of WoS/Scopus based measurements (including the non-WoS/Scopus publications…)
As always, never use it as a stand-alone tool. But also: it is an effective instrument for measuring interdisciplinarity, knowledge flows and knowledge diffusion -even for non-WoS/Scopus publications!
Thank you for your attention
more information: www.cwts.leidenuniv.nl
Appendix
According to an influential Swiss scientist:
Bibliometric investigations are clearly not very reliable…. In particular, the "frequency of citation" does not account for the quality of the researchers, because
(1) it depends more often on the social recognition of the researcher than excellence of his/her scientific work;
(2) it favors researchers who work on fashionable topics;
(3) it favors the fields of knowledge which traditionally publish shorter articles compared to those where publications are longer;
(4) it cannot differentiate between the fashion and the substance of a paper;
(5) it can favor the authors of "surveys", who are very frequently cited, compared to the authors of focused research papers;
(6) a position article or even an erroneous article can be criticized and consequently well cited.
Write your name on papers by your PhD students Ignore your publisher’s copyright: put your paper online Work in a popular area so that many others can cite you Write survey papers, not research papers Never change your established research area Avoid innovative and new (but risky) projects Chose catchy titles for your papers Emphasize quantity instead of quality Do not lose valuable time, avoid events like this one Concentrate on paper production, not good teaching Heavily cite you own (and your friend’s) papers Never publish more than a single ‘least publishable unit’ Cannibalize your old papers: refurbish and republish them
According to an influential Swiss scientist:
How to increase your ‘bibliometric values’
Main anecdotal objections against citation analysis
- Mendel Syndrome- Wittgenstein Syndrome- Lowry Effect- Einstein effect- Old boys clique- Disgusting anyway
A scientist has index h if h of his/her N papers have at least h citations each and the other (N-h) papers have no more than h citations each
Hirsch (h-) index AFJ van Raan =
18
Correlation of h-index (h) with number of citations (C)for all chemistry groups in the Netherlands
y = 0.394x0.4543
R2 = 0.8793
1
10
100
1 10 100 1000 10000
C
h
Correlation of h-index (h) with number of publications (P)for all chemistry groups in the Netherlands
y = 0.7293x0.5186
R2 = 0.4859
1
10
100
1 10 100 1000
P
h
Correlation of h-index (h) with CPP/FCSmfor all chemistry groups in the Netherlands
y = 6.9566x0.5331
R2 = 0.2161
1
10
100
0.10 1.00 10.00
CPP/FCSm
h
Large European University
(A&H)
APC
BIOL-AP
BIOL-HU
CHEM
CLM ECON
ENG
GEO
MATH
MOLB
(MULTI)
PHYS
PSY
SOC-MED
(SOC)
0
25
50
75
100
0255075100PUBLICATION RANK PTCL
CIT
AT
ION
I P
AC
T R
AN
K P
CT
L
Among top 25 % in publication output and citation impact
Top 25%
Bottom 25%
Impactranking
Publ.ranking Top 25%Bottom 25%
‘ Top’ research university
(SOC)
SOC-MEDPSY
PHYS (MULTI) MOLB
MATH
GEO
ENG ECON
CLM
CHEM
BIOL-HUBIOL-AP
APC
(A&H)
0
25
50
75
100
0255075100PUBLICATION RANK PTCL
CIT
AT
ION
I PA
CT
RA
NK
PC
TL
University has a top position
in each discipline
Bottom 25% Publ.ranking Top 25%
Top 25%
Impactranking
Bottom 25%
Citation-counting scheme based on‘roof-tile’ method:
Citation years 1995 1996 1997 1998 1999 2000 2001 20021995
Let 4181 4264Edi 1313 905Other 1421 909Total 7872 14037 (c)
ISI IF
Citations in 2002__________ Citeable documents in 2000 and 2001
14037 (c) 957 (a)
IF=14.7
CWTS IF Citations to Art/Not/Rev in 2002 Art/Not/Rev in 2000 and 2001
7959 (b) 957 (a)
Citations to Art/Let/Not/Rev in 2002 Art/Let/Not/Rev in 2000 and 2001
7959+4264 957+4181
IF=8.3
IF=2.4
Manipulability of citation indicators proposed in this study
To which extent are our citation-based indicators sensitive to manipulation?
Can one increase actual citation impact by:
(1) Increasing author self citation?
Author self-citations are not included: increasing author self-citation has no effect
(2) Publishing in high impact journals?
A case study of 2,000 UK senior authors with >10 p/y revealed that journal impact explains ~20% of the variance in the citation impact rates. Journal impact is therefore not a dominant determinant of actual citation impact at the level of individual senior authors.
(3) Collaborate more intensively?
Some studies report positive correlation between number of authors and citation impact, but they ignore differences in authoring practices among research fields. Author self-citations are not included in this study. It all depends upon who collaborates with whom. There is also the issue of causality: ‘good’ research may attract high-impact collaborators.
(4) Publishing with US authors because they overcite their own papers?
Studies found no conclusive evidence that US scientists in science fields excessively cite papers originating from their own country.
(5) Publishing less, only the very best papers?
One would expect a higher citation impact per paper. Longer term effects of such a publication strategy are uncertain. PhD students need papers in their CV’s. It may become difficult for a group to attract good PhD students if its policy is to let them publish only a few papers. Another factor is that publications also enhance the visibility of a group’s research activities. If a group starts publishing substantially less papers, this may lead to a lower visibility and hence to a lower citation impact, even per paper.
(6) Making citation arrangements?
A high impact group receives its citations from dozens of different institutions. The distribution of citations amongst citing institutions is very skewed. The contribution of the tail of the distribution to the citation impact is relatively large. Making arrangements with a few institutions will not lead to a substantial increase in citation impact.
Application of Thomson-ISI Impact Factors for research performance evaluation is irresponsible
* Much too short ‘Citation window’* No Field-specific Normalization * No distinction between document types * Calculation errors/inconsistencies nominator/denominator* Underlying citation distribution is very skew:
IF-value heavily determined by a few very highly cited papers
* What is quality?* Numbers are order of magnitude lower > examples (e.g., profiles)* H-index example social sciences* National publications (also the case for engineering!)* Coverage* Figures about the role of books vs. journal papers (Uppsala data?)* Language* Citation window* non-WoS analysis, target known* non-WoS analysis, target unknown (CHE-2 results)* Societal relevance of social sciences > try sustainability maps
for social science themes!* Database problems (EC-ASSIST list)* Social science data from benchmark studies* Norwegian Association of Higher Education Institutions classification of sources* European Reference Index for the humanities journal classifications ERIH-ESF-HERA* Library Catalog Analysis: number of library copies per book title, e.g., Worldcat (Linmans); exploratory study of 43 catalogs in economics (ask HM…)* Use slides of HMs and TNs CHERPA presentations!* bibliometrics is more than an instrument of research performance analysis, it van also reveal patterns of knowledge development and diffusion
* Is the lifetime of a book longer than that of a journal article?* Flemish study on social sciences and humanities!* Nature of citations may be different* Hierarchy of books through reputation of publishers?* Results of our HEFCE analysis of the RAE 2001* Figures on p.37-40 in HEFCE Scoping report* Leiden Benchmarking: social sciences and humanities: order of magnitude,
Output and impact compared world field average1999 - 2006
Erasmus University and 20 benchmark institutions
NORTHWESTERN UNIV (USA)
DUKE UNIV (USA) UNIV N CAROLINA - CHAPEL HILL (USA)
DARTMOUTH COLL (USA) UNIV PENN (USA)
UNIV MARYLAND - COLLEGE PARK (USA)
MIT (USA) STANFORD UNIV (USA)
HARVARD UNIV (USA)
UNIV CHICAGO (USA)
STOCKHOLM SCH ECON (SWEDEN)
COPENHAGEN BUSINESS SCH (DENMARK)
LONDON BUSINESS SCH (GREAT BRITAIN)
UNIV OXFORD (GREAT BRITAIN)
INSEAD (FRANCE)
UNIV MAASTRICHT (NETHERLANDS)
UNIV GRONINGEN (NETHERLANDS)
UNIV TILBURG (NETHERLANDS)
UNIV AMSTERDAM (NETHERLANDS)
ERASMUS UNIV ROTTERDAM
(NETHERLANDS)
0.50
1.00
1.50
2.00
2.50
3.00
3.50
4.00
4.50
5.00
5.50
0 100 200 300 400 500 600 700
TOTAL PUBLICATIONS
Black coloured squares above (below) the horizontal reference line represent groups for which the impact (CPP) is significantly above (below) world average (FCSm)
CP
P/F
CS
m
Top-ranked publications according to
country of origin
Country of first author of Top 50 publ > 1980
Field US UK EU continent Other Unknown
Political sc 50% 20% 18% 5% 7%
Economics 71% 9% 9% 6% 6%
Psychology 50% 12% 21% 12% 6%
In the set of ‘best’ publications submitted to the 2001 RAE it was found that journal articles constitute 73% of submitted papers from all Subject Groups.
For science-related Units of Assessment we find 92%. The profile for Mathematics is quite similar to that for Science.
In Social Sciences and Humanities books are important publication sources. The shares of authored books and book chapters are 15 and 24%, respectively.
Most highly cited non-WoS items by document type
Book
Journal
Article
Hand-
book Manual
Soft-
ware
Proc-
eedings Thesis Other
Political sc 54 19 10 3 2 14 4 5
Economics 75 52 42 2 2 4 5 8
Psychology 150 112 30 430 32 11 11 2
The comparison of WoS and Scopus coverage of the 2001 RAE ‘best’ publications shows that Scopus coverage is especially better in the Subject Groups Subjects allied to Health (e.g., clinical dentistry, nursing, pharmacy), and to a lesser extent also in Engineering & Computer Science and Health Sciences.
In Clinical Medicine, Biological Sciences and Physical Sciences, however, Scopus coverage is slightly lower than WoS coverage.
from: Moed and Visser 2008, Appraisal of Citation Data Sources, HEFCE-report
Percentage of WoS papers found in Scopus
Publ Year WoS Database Segment
Nr WoS-covered J ournals
Nr WoS Articles and
Reviews
% WoS Articles and
Reviews found in Scopus
1996 Science 5,320 673,271 93
1996 Soc Sc & Hum 2,610 88,583 57
1996 Total 7,930 761,854 89
2005 Science 6,146 867,748 97
2005 Soc Sc & Hum 2,719 94,629 72
2005 Total 8,865 962,377 95
0
20
40
60
80
100
0 10 20 30 40 50 60 70 80 90 100
% WoS Papers found in Scopus
% W
oS
Jo
urn
als
1996
2005
Distribution of %WoS papers found in Scopus, science fields
WoS vs Scopus coverage of 2001 RAE ‘best’ publications Total
submitt publ
% in WoS
% in Scopus
Δ
Discipline
Science 95,056 84.1 84.4 0.3 Mathematics 6,634 81.8 80.1 -1.7 Social Sciences and Humanities 91,324 24.9 25.9 1.0
Total submitt
publ
% in WoS
% in Scopus
Δ
SSHU Town & Country Planning 1,478 38.1 57.4 +19.4 SSHU Social Work 1,642 22.9 36.7 +13.8 SSHU Accounting and Finance 779 21.7 34.9 +13.2 SSHU Built Environment 2,471 24.5 35.9 +11.4 SSHU Social Policy & Administr 3,912 25.8 34.1 +8.3 SSHU Politics and Internat St 4,382 26.4 34.6 +8.1 SSHU Sociology 3,519 29.2 37.0 +7.8 SSHU Business & Managem St 9,746 37.9 45.5 +7.6 SSHU Geography 4,890 61.6 68.6 +7.0 SSHU Education 8,662 16.0 22.5 +6.5 SSHU Econom & Econometrics 2,879 67.5 72.0 +4.5 SSHU Sports-related Subjects 1,301 60.5 63.4 +2.9 SSHU Library & Inform Manag 1,259 31.7 34.4 +2.7 SSHU Anthropology 1,180 27.5 28.0 +0.4
INTERNAL COVERAGE OF THE CITATION INDEX BY MAIN FIELD