Webology, Volume 8, Number 2, December, 2011 |
| Home | Table of Contents | Titles & Subject Index | Authors Index |
Leila Dehghani
Faculty Member, Department of Medical Library and Information Science, Bushehr University of Medical Sciences, Bushehr, Iran. E-mail: leiladehghani (at) yahoo.com
Reza Basirian Jahromi
Faculty Member, Department of Medical Library and Information Science, Bushehr University of Medical Sciences, Bushehr, Iran. E-mail: rezabj (at) gmail.com
Mazyar Ganjoo
Department of Computer Science, Islamic Azad University- Bushehr Branch, Bushehr, Iran. E-mail: ganjoo (at) gmail.com
Received October 12, 2011; Accepted December 20, 2011
The h-index is one of the recent indicators in the field of scientometric which introduced by Jorge Hirsch and measures both the productivity and impact of scholars' work. It has been widely used to show the relevance of the research work of prominent scholars. But, does h-index really show this relevance? The purpose of this study is to investigate the frequency of citations to highly-cited researchers by their co-authors, as well as to quantify the weight of self-citations of highly-cited researchers' in three fields of science: clinical medicine, computer science, and economics & business. Finally, we seek to evaluate the impact of these two elements on their h-index in the Scopus database. In this research, highly-cited researchers were selected from the three fields via HighlyCited.com. The study population includes 999 highly-cited researchers in these three fields (280 authors in clinical medicine, 373 authors in computer science, and 346 authors in economics & business). The results show that the average rate of self-citations and co-author citations in clinical medicine is more than computer science, and in computer science is more than economics & business. The results show that there is a positive correlation between self-citation and total citations (P<0.01) in these three fields. The results also show that there is a positive correlation between "co-author citations" and total citations (P<0.01). Moreover, there is a positive correlation between self-citations and h-index of highly-cited researchers as well as co-author citations and h-index of highly-cited researchers in the above-mentioned three fields. It is seen that there is a significant difference in self-citations behavior and co-author citation behavior in different fields which relates to subject context. Co-author citation is an indicator which, on the one hand, not under control of own researcher and, on the other hand, represents the structure of scientific communications. Although the number of citations is not necessarily increased by raising the number of authors, but co-author citations increased by raising the number of co-authors.
Highly-cited researchers; Self-citation; Co-author; Citation analysis; H-index; Scopus
Scientometric indicators – like the number of published articles by a researcher during a period of time, the number of total citations to the articles, the average of citations to each article, the impact factor, and the rank of the journals where the articles have been published – are dependable criteria in evaluating the scientific output of researchers. Hirsch (2005) has proposed the h-index as a single number criterion to evaluate the scientific output of a researcher. According to his definition, "A scientist has index h if h of his/her Np papers have at least h citations each, and the other (Np- h) papers have fewer than h citations each".
This indictor has defined the number of scientific output of researchers as termed as scientific effectiveness of these researchers. Another characteristic and advantage of h-index includes that it is mathematically simple, it may be applied to any level of aggregation, and it is a robust indicator (Rousseau, 2007). Although this indicator is considered for calculating the period of researcher's work but it calculates total works of a researcher and neglects his/her highly-cited and lowly-cited articles. Kelly and Jennions (2006) have detected that this indicator has been designed in order to improve the other measuring indicators of science like the number of total articles, and the number of total citations; therefore, it can differentiate between affective researchers and those who published many articles.
Although h-index has some advantages over similar indicators, it has some weaknesses. "Some criticisms of the h-index have include that it is bounded by the total number of publications, it does not recognize the context of the citations received, it is limited by the completeness of the database used to do the calculation, and it does not account for the number of authors on paper" (Brown, 2009). In addition, as h-index increases by increasing the number of citations, so two effective elements – self-citation and citing co-authors – are very important. Interestingly, Hirsch (2005) concluded that "self-citation would not significantly affect the h-index of an individual author. Although self-citations can obviously increase a scientist's h-index, he contended that their effect on the h-index is much smaller than on the total citation count" (Brown, 2009). This is an assumption that has been challenged by many studies (Kelly & Jennions, 2006; Schreiber, 2007a) that have shown average decreases in the h-index of 12 percent with self-citations excluded. Also some other researchers in their articles have considered the relationship between self-citation and h-index (Couto et al., 2009; Engqvist & Frommen, 2010; Gianoli & Molina-Montenegro, 2009; Schreiber, 2007a). Of course, it is reasonable that authors citing to the previous relevant works and using their previous experiments and results. Thijs & Glänzel (2006) stated that "self-citation has been shown to vary by field and by origin of publication but average self-citation rates between 10 and 40 percent are common". Cutto et al. (2009) stated that "average self-citation rates between 10 and 30 percent are common". According to Brown, self-citation is a powerful tool within the author's power to use and may be open to abuse to increase total citation count and, through much targeted citation, to specifically increase the h-index (Brown, 2009). So it seems reasonable that h-index be considered without self-citations included for a more unbiased comparison of authors in the same field.
Researches show that the effect of self-citation on h-index among lowly-cited authors is more (Fowler & Aksnes, 2007; Gianoli & Molina-Montenegro, 2009; Iglesias & Pecharroman, 2007; Schreiber, 2007b). Although in Gianoli and Molina-Montenegro, the h-index of highly-cited researchers has been considered, but they have just focused on one field or special field (Gianoli & Molina-Montenegro, 2009). However, the effect of citing to researchers by their co-authors and their self-citations on h-index has not been investigated.
The main purpose of this study is to indicate some deficiencies in previous researches with emphasis on h-index. These deficiencies include unknowing effect of co-authors citations on h-index of highly-cited researchers and investigating of self-citations among highly-cited researchers at international level. Therefore, the major objectives of this research are to investigate the frequency of citations to highly-cited researchers by their co-authors and their self-citations and to evaluate the effect of these elements on h-index of highly-cited researchers in Scopus.
The objectives of this study are to determine:
Databases which have been used for calculating h-index are ISI Web of Science, Scopus, and Google Scholar. Although the Scopus database is weaker than ISI Web of Science in terms of time coverage and is representing citation information of articles since 1996, but Science Citation Index (SCI) and Social Science Citation Index (SSCI) are representing citation information of articles since 1974. So, h-index which has been represented for each author in ISI Web of Science is more than the Scopus database (Jacso, 2008a, 2008b, 2008c, 2008d).
The Scopus database has the possibility for calculating h-index by excluding "self-citation" as well as calculating h-index by excluding "all of co-authors"; therefore, this database was selected. Note that this selection is one of the main limitations in our research.
The data of the present research were collected in February 14-30, 2011 via Scopus. To investigate the frequency of self-citations and the number of citations of co-authors among highly-cited researchers, the list of ISI highly-cited researchers has been used which can be accessed via HighlyCited.com. This database is produced by Thomson Reuters and has information about researchers who have the most citations by the other researchers. We chose three fields – clinical medicine, computer science, and economics & business– in three main disciplines – medicine, engineering, and humanities – and finally 999 highly-cited researchers in those three fields were selected (280 authors in clinical medicine, 373 authors in computer science, and 346 authors in economics & business) as research population.
To achieve recall and precision in searching authors' names which input variously (inconsistency in inputting author names), first we referred to HighlyCited.com and required data were extracted for 999 researchers. Then, we used "author search" box to accomplish our goals. The Scopus database can maximally select 15 statuses of authors' names; due to this limitation we decided to select records with maximum number of articles.
For data analyzing, first we analyzed collecting data by Excel; these data include: the number of articles in Scopus, h-index, the number of citations, h-index by excluding self-citation, the number of citations without self-citations, the number of self-citations, h-index by excluding "all of co-authors", h-index by excluding the citations of co-authors, the number of citations by excluding "all of co-authors", and the number of the citations of co-authors. As the required data in the Scopus database has been represented since 1996 and some of these researchers have had their more scientific activities before 1996, these researchers deleted from the same population because their h-index was low. Finally, out of those 999 scholars, 684 researchers were selected. After using Excel for primitive calculating and descriptive statistics, SPSS 16.0 was used for inferential statistics; these tests included: T-test, ANOVA test, Correlation test, and Regression test.
The results show that the average of self-citations in clinical medicine is more than computer science, and in computer science it is more than economics & business (Figure 1).
In these three fields, the results show that there is a positive correlation between self-citations and total citations (P<0.01). Figure 2 indicates the correlation in three fields.
The percentage of "self-citations" to "total citations" is varying in these three fields; this ratio is between 1% and 17% in economics & business, between 1% and 26% in computer science, and between 1% and 14% in clinical medicine (Figure 3).
The results show that the average of "co-author citations" in clinical medicine is more than computer science, and in computer science it is more than economics & business (Figure 4).
The results show that there is a positive correlation between "co-author citations" and total citations (P<0.01) in these fields. Figure 5 indicates the correlation in three fields.
The percentage of "co-author citations" to "total citations" is varying in these three fields; this ratio is between 0.1% and 8% in both economics & business and in computer science, and between 1% and 18% in clinical medicine (Figure 6).
Using ANOVA test, it is found that there is a significant difference between "self-citations" and "co-author citations" in these three fields (P<0.01).
The results show that there is a positive correlation between self-citations and co-author citations (P<0.01). Figure 7 indicates the correlation in these three fields.
The results show that there is a positive correlation between self-citations and h-index of highly-cited researchers as well as co-author citations and h-index of highly-cited researchers in three fields (Table 1).
| Economics & Business | Computer Science | Clinical Medicine | |
|---|---|---|---|
| Effect of self-citations on h-index | 0.277 P<0.001 | 0.609 P<0.001 | 0.740 P<0.001 |
| Effect of co-author citations on h-index | 0.195 P=0.195 | 0.263 P<0.001 | 0.303 P<0.001 |
To answer this question, T-test was used and the results show that the average of h-index by excluding self-citations is more than the average of h-index by excluding co-author citations among highly-cited researchers (P<0.01). The effect of self-citations on increasing h-index among highly-cited researchers in computer science and economics & business is more than clinical medicine (Table 2).
| Economics & Business | Computer Science | Clinical Medicine | Total | |
|---|---|---|---|---|
| Average of h-index by excluding self-citations | 0.34±0.77 | 0.8±0.96 | 1.21±1.42 | 0.72±1.10 |
| Average of h-index by excluding co-author citations | 0.15±0.47 | 0.51±0.62 | 1.67±1.50 | 0.68±1.11 |
| P Value | P<0.001 | P<0.001 | P<0.001 | P<0.001 |
Scientometric indicators are used to measure scientific output, research performance, institutional performance, scientific collaboration, and knowledge growth. They are also used in university ranking, researcher ranking, and research policy-making.
Researchers require previous related researches for indicating continuity of communications between previous researches. Also, "co-author citation" indicates the scientific communications between scholar and his/her colleagues. "Self-citation" and "co-author citation" are parts of the process of scientific communications. The authors sum up the discussion with evidence from earlier research in the following ways:
Scientometric indicators based on the number of publications, authors, references and citations are increasingly used to evaluate the pattern of research performed by researchers, institutions, and countries. This study aims to investigate the relationship between co-author citations and self-citations of highly-cited researchers and their h-indexes. Findings reveal that the number of citations is not necessarily increased by raising the number of authors, but co-author citations increased by increasing the number of co-authors. The results also show that there is a positive correlation between self-citations, co-author citations, and h-index of highly-cited researchers in clinical medicine, computer science, and economics & business.
The authors would like to thank the anonymous reviewers for their valuable comments that helped to improve the paper.