Library and Information Science ISSN: 2435-8495

三田図書館・情報学会
Mita Society for Library and Information Science

〒108‒8345 東京都港区三田2‒15‒45 慶應義塾大学文学部図書館・情報学専攻内
c/o Keio University, 2-15-45 Mita, Minato-ku, Tokyo 108-8345, Japan

https://mslis.jp/ E-mail:mita-slis@ml.keio.jp

https://mslis.jp/ E-mail:mita-slis@ml.keio.jp

Library and Information Science 18: 157-169 (1980)

doi:10.46895/lis.18.157

doi:10.46895/lis.18.157

発行日：1981年3月25日Published: March 25, 1981

Three methods of analyzing matrices of mutual citation frequencies among scientific periodicals are proposed in order to identify substructures, to visualize information flows among elements, and to clarify the characteristic of a specific discipline of science. Two of them are agglomerative clustering and multidimensional scaling: they are well-known statistical techniques based on a similarity or dissimilarity measure among elements. Since the measure of similarity assumes symmetric interaction, these analyses need symmetrization of the citation matrix. At the same time, this transformation ignores a distinctive feature, i.e., non-symmetricity of the citation matrix. Here the non-symmetricity occurs when a journal frequently cites another, while the latter has not many citations of the former. Therefore, another method which is called here “hierarchical grap” is presented so that non-symmetrical relationships are represented on a directed graph whose nodes are hierarchically positioned. The last method determines two kinds of arcs of the graph. Roughly speaking, an arc of one-sided influence from A to B is drawn when journal B frequently cites journal A but the converse is not true; an arc of two-sided influence between A and B is drawn when both journals have many citations of each other. In order to portray the graph, the positions of the nodes are determined so that the arcs have upward directions. An outline of the algorithm for this configuration is given with a simple numerical example.

These methods are applied to the relationship of 43 journals of environmental/civil engineering: the data are extracted from Science Citation Index (1977). The citation matrix is derived and the above methods are applied. The clustering gives nine subgroups which consist of four groups of civil engineering, three of environmental engineering, and two of interdisciplinary studies. The hierarchical graph represents the information flows, from which the studies about water systems seem to be remarkable. Further, the MDS represents the structure of this field by a two-dimensional plot: the first axis shows the relationship of civil and environmental engineering, whereas the second axis suggests nature-human relationship. Moreover, both axes show the importance of the studies about water systems in the discipline of environmental/civil engineering.

© 1980 三田図書館・情報学会© 1980 Mita Society for Library and Information Science

This page was created on 2022-07-14T10:13:51.569+09:00

This page was last modified on

このサイトは（株）国際文献社によって運用されています。