CHAPTER 21 - MULTIVARIATE DATA-ANALYSIS METHODS IN BIBLIOMETRIC STUDIES OF SCIENCE AND TECHNOLOGY

Abstract

In the past decades quantitative studies in social and behavioral sciences have benefitted both in range and depth from the application of statistical methods, more in particular the multivariate analysis (MVA) methods. These methods are called for when assessing quantitative relations between two or more units of analysis. MVA methods have also become increasingly important analytic aids in large-scale bibliometric (i.e. quantitative, scientific literature-based) studies of science and technology. This paper centres on basic possibilities of this analytic toolkit for this type of studies. An introductory non-mathematical overview is presented for important subclasses of MVA methods suitable for the analysis of multivariate bibliometric data. Section 1 starts with a brief general account of the relevance of bibliometric studies followed by argumentation to justify a structured overview of applicable MVA methods in such studies for those who are unfamiliar with MVA. In Section 2 we continue with a sketch of the role of data-analysis methods in general within this type of studies. Section 3 begins by presenting some data, obtained from scanning articles in bibliometrics-oriented journals, to give an indication on the extent of MVA-usage in bibliometric research articles during the last decade. This introduction is followed by a discussion of some basic properties of multivariate data in general and an outline of a conceptual framework of bibliographic information encompassing all general types of multivariate bibliometric data-arrays. Section 4 is used for classification of general classes of MVA methods based on the previously discussed inherent properties of MVA and user-imposed constraints on multivariate data. This section is primairily meant as a preliminary to Section 5 in which we present a classification of MVA methods which figure prominently in bibliometric studies. This is followed by synopses of subclasses of MVA methods and the general types of data they can be applied to. In addition, this structured overview is larded with more recent bibliometric applications of this methods. Subsequently, Section 6 contains an illustrative application of two of the discussed MVA methods, as an addition to the basic data as given in Section 3, thus yielding a more detailed insight in the magnitute and trends with respect to the usage of separate MVA methods in bibliometric studies. Finally, in Section 7 we draw some general conclusions and advance our views on the prospects of MVA in bibliometric S&T-studies.