.
Attempts have been made to classify all of the proposed indices. Schreiber et al. (2012) suggested a classification system based on two dimensions of scientific performance: quality and quantity. This classification system considers that some indices have a stronger tendency than the rest to measure the quantity of research output, while other indices (g‐, A‐, and R‐indices) have a stronger tendency to characterize the quality of research output. It is suggested that two complementary indices, for example, h‐ and A‐indices, should be used in the scientometric analysis of an individual researcher's productivity and impact (Schreiber et al. 2011).
9.5 A General Criticism on the Use of Metrics
Although the use of single metrics (based on bibliometric measurements) for the comparison of researchers has steadily gained popularity in recent years, there is an ongoing debate regarding the appropriateness of this practice. The question is whether single measures of research performance are sufficient to quantify such complex activities. A report by the Joint Committee on Quantitative Assessment of Research argues strongly against the use of citation metrics alone as a tool in the field of mathematics. Rather, the committee encourages the use of more complex methods for judging the impact of researchers; for example, using evaluation criteria that combines citation metrics with other relevant determinants, including membership on editorial boards, awards, invitations, or peer‐review activities (Adler et al. 2008). Along these lines, Egghe (2007) stated earlier, “The reality is that as time passes, it's not going to be possible to measure an author's performance using just one tool. A range of indices is needed that together will produce a highly accurate evaluation of an author's impact.” Shortly thereafter, Bollen et al. (2009) offered empirical verification of Egghe's intuitive hypothesis. Based on the results of a principal component analysis (PCA) on a total number of 39 existing indicators of scientific impact, an argument was made that ideal scientific impact should be multidimensional and cannot be effectively measured by a single numeric indicator.
9.6 Citation Data Sources
There is an ongoing debate on the issue of multiple citation data sources (Jacso 2008). Citation data sources, or databases, are web‐based data sources that can be accessed freely or through a subscription cost. These databases provide the meta‐data of scientific publications and their citation information. The debate mainly concerns the fact that the various available data sources are likely to produce different citation data for the same publication. In fact, a comparison has been made to test the robustness of citation outputs from Thomson Reuters Web of Science (WoS) – formerly known as Thomson Corporation's Institute for Scientific Information (ISI) Web of Knowledge – and Google Scholar. The latter uses Publish or Perish, a software program that retrieves citations from Google Scholar and analyses them to present various metrics, a common application in bibliometric literature.10 The WoS results tend to underestimate the citations because WoS covers solely journals included in the ISI list. Google Scholar, on the other hand, tends to overestimate the citations because in addition to covering more journals, it also retrieves citations to working papers, books, and more (Falagas et al. 2007; Meho and Yang 2007; Jacso 2008; Franceschet 2010). Although several groups have supported the idea of using Google Scholar to implement citation‐based statistics, many scientific authorities claim that the Google Scholar data are often inaccurate (Adler et al. 2008). Meho and Rogers (2008) further examined the differences between Scopus and WoS, and reported that no significant differences exist between the two databases if only journal citations are compared. Nevertheless, we have to stress that the specific database used for collecting bibliometric citation data can, in fact, influence scientometric analysis. Many authors have cautioned against the use of citation data without further evaluating the database for validity and verification (Dodson 2009).
9.7 Discussion
An important