POCI-01-0145-FEDER-007638), to FCT/MCTES through national funds (PIDDAC), and the co- funding by the FEDER, within the PT2020 Partnership Agreement ...
CAN PHD PROGRAMS FROM DIFFERENT AREAS BE COMPARED USING BIBLIOMETRIC DATA? A. Teixeira1, T. Rocha-Santos2, G. P. Dias3, J.F. Mendes4 1
Doctoral School (EDUA), Instituto de Telecomunicações (IT), Dep. Eletrónica, Telecomunicações e Informática (DETI), University of Aveiro, 3810-193 Aveiro, Portugal 2
Centre for Environmental and Marine Studies (CESAM) & Department of Chemistry, University of Aveiro, 3810-193 Aveiro, Portugal 3
Research Unit on Governance, Competitiveness and Public Policies (GOVCOPP),School of Technology and Management of Águeda (ESTGA), University of Aveiro, 3750-127, Águeda, Portugal 4
Institute of Nanostructures, Nanomodelling and Nanofabrication (I3N), Physics department, University of Aveiro, 3810-193 Aveiro, Portugal
Abstract The bibliometric indicators for the University of Aveiro PhD programs considering the 177 PhD students that finished their PhD in the civil year of 2016 have been obtained from SCOPUS and analyzed using principal component analysis and K-means cluster. The scientific production of those students distributed among 41 PhD programs has been used in this study, of which 38,4% were male and 61,6% were female. Moreover 18.6% of the PhD students have one supervisor, 63.3% have one supervisor and one co-supervisor while 18,1% have one supervisor and two co-supervisors. The PhD programs in the areas of social sciences, arts and humanities, services and education correlate less with Number of papers, Number of paper with international Institutions and Total citations than the PhD programs in the areas such as natural and exact sciences, health sciences and engineering.
Keywords: Higher Education, PhD program, Internationalisation, bibliometrics
1
INTRODUCTION
Bibliometric databases such as SCOPUS and Web of Science have been used as quantitative methods for research evaluation [1-2]. In natural and life sciences areas, these methods are more frequently used than in humanities and social sciences areas [3]. SCOPUS covers scientific journals, books and conference proceedings in the fields of science, technology, medicine, social sciences and arts and humanities. The University of Aveiro Doctoral School has an approximate total of 1300 PhD students (2016/2017) distributed in 50 doctoral programs from the following areas: natural and exact sciences, health sciences, engineering, social sciences, services, education, arts and humanities. The PhD doctoral programs at University of Aveiro typically consist in one year of coursework with curricular unit’s specific for each program and two to three years of individual study and research work with one supervisor and up to two additional co-supervisors. Each PhD program has a director that oversees the program with guidance from the Doctoral School and the research and third cycles ViceRector. This work aims at observing if the analysis based on the chosen bibliometric indicators based on SCOPUS database for the University of Aveiro PhD programs considering PhD students that finished their PhD in 2016 is useful for helping characterizing and defining strategies for the running PhDs.
2
METHODOLOGY
The current study includes the publications of 177 students from 41 PhD programs from the 50 available at the University of Aveiro, who finished their PhD in the year 2016. These PhD programs are from seven areas: education; social sciences; engineering; arts and humanities; exact and natural Sciences; and health sciences. The bibliometric indicators have been extracted from Scopus database and included the number of publications, total citations, average citations, number of publications with co-authors from international organizations (Universities, Research Centres and Companies). All publication types extracted from Scopus are included in the analysis of the current study until
September 2017. Other information such as number of supervisors, gender, and areas has also been used. Statistical methods such as Principal component analysis and K-means clustering have been applied using the XLSTAT in order to find the correlations between chosen indicators and the specific doctoral programs. Descriptive statistics using XLSTAT has also been used.
3
RESULTS
It can be observed in Figure 1 that 38.4% of the PhD students were male and 61.6% were female. Moreover, 18.6% of the PhD students have one supervisor, 63.3% have one supervisor and one cosupervisor while 18,1% have one supervisor and two co-supervisors.
38.4%
61.6%
F
M
Figure 1. Relative frequency of female and male PhD students. From the total of 41 PhD programs under study, 24 had less than 5 students finishing their studies in the year 2016. The Principal component analysis was applied to the data obtained from SCOPUS for a) firstly, the 41 PhD programs and, b) secondly, the 13 PhD programs that had at least 5 students or more finishing their studies in the year 2016. Based on the Principal components analysis (PCA) correlation matrix obtained (for the analysis of the 41 PhD programs) it can be inferred that Total Citations variable has low correlation with the Number of papers with International Institutions variable (Table 1).
Table 1. PCA correlation matrix.
Variables Number of Papers Number of papers international Institutions Total Citations
Number of Papers
Number of papers with international Institutions
Total Citations
1
0.633
0.769
0.633
1
0.321
0.769
0.321
1
with
The first factor represents 72.26% of the initial variability of the data, and the two factors represent 95.13% of the variability, as it can be observed in Figure 2.
Figure 2. PCA biplot representing the active variables (Number of Papers, Number of Papers with international Institutions and Total Citations) and the active observations (PhD programs) for the 41 PhD programs.
From the PCA biplot it can be observed different grouping of PhD programs. At least four different groups can be perceived. The yellow group constituted by Energy systems and climate change and biochemistry doctoral programs seems to have more significance in terms of Total Citations. The green group is constituted by Chemistry; Nanosciences and nanotechnology; Informatics; Health sciences and technology that have significance both in the Number of Papers and Total Citations variables. The red group is constituted by Biomedicine; Marine sciences; Gerontology and geriatrics; Materials science and engineering; Physics; Physical engineering; Biology; Biology and ecology of global changes; Electrical engineering; Mechanical engineering; Environmental sciences and engineering; Telecommunications seems to have more influence on the Number of papers with international Institutions. Finally, the magenta group constituted by Marine and environmental sciences; Geosciences; Civil engineering; Refining, petrochemical and chemical engineering; Plants biology, Multimedia in education; Industrial engineering and management; Tourism; Mathematics, Information and communication in digital platforms; Higher education studies; Education, Psychology; Marketing and strategy; Cultural studies, Chemical Engineering; Mathematics and application; Music; Design; Accounting; Geotechnology; Literacy studies are likely to have less influence in the three active variables of this study. K-means clustering was applied to the same data and shown 5 clusters (data not shown). These clusters are slightly different from the ones obtained by PCA analysis, but one cluster is very similar to the magenta group resulting from PCA analysis. This similar group includes all the PhD programs in the social sciences and education areas, besides some from other areas. In fact, these were the PhD programs with less publications in SCOPUS. This can be a consequence of the use of bibliometrics on the basis of SCOPUS database due to a higher portion of journals which are not included in the database and also to a larger share of book contributions and monographs [3]. Following, as referred, the PCA has been applied to the data obtained from SCOPUS for the 13 PhD programs that have less than five students finishing their PhD in the year 2016 and Figure 3 shows the PCA biplot.
Figure 3. PCA biplot representing the active variables (Number of Papers, Number of Papers with international Institutions and Total Citations) and the active observations (PhD programs) for the 13 PhD programs. As it can be observed in the PCA biplot (Figure 3) the two factors represent around 98% of the variability of the data. It can also be observed that the Number of papers with international Institutions variable continues to less correlate with Total of Citations variable. Despite small changes in the arrangement of the PhD programs, Civil engineering; Tourism; Music; Design; Education, Psychology continue to be the PhD programs that are likely to have less influence in the three active variables of this study.
4
CONCLUSIONS
It can be concluded that, for the sample data used, the Number of papers with international Institutions has low correlation with Total citations; while the number of papers has a high correlation with the total citations, which means that the number of citations is not highly dependent of the presence of partnerships with International Institutions in the scientific papers. Moreover, the PhD programs which show the highest level of internationalization in their publications were Biomedicine; Marine sciences; Gerontology and geriatrics; Materials science and engineering; Physics; Physical engineering; Biology; Biology and ecology of global changes; Electrical engineering; Mechanical engineering; Environmental sciences and engineering; Telecommunications, which were PhD programs from areas of exact and natural sciences, health sciences and engineering. PhD programs such as Chemistry; Nanosciences and nanotechnology; Informatics; Health sciences and technology; Energy systems and climate change; and Biochemistry are the ones who present higher publication track record of scientific papers and have more citations. The PhD programs in the areas of social sciences, services, education, arts and humanities are the less correlated with number of papers, number of papers with international Institutions and total citations while the PhD programs in the areas such as exact and natural sciences, health sciences; and engineering are more correlated with those variables. In face of the results one may conclude that bibliometric data based on the SCOPUS database seems suitable for research evaluation on the areas of exact and natural sciences, health sciences; and engineering and probably not so suitable for the areas of education, services, arts and humanities and social sciences. However the method of the PCA biplot may be of relevance for analysing the existing production and from there establish policies to improve these matrices.
ACKNOWLEDGEMENTS This work was also supported by national funds through FCT/MEC (PIDDAC) under project IF/00407/2013/CP1162/CT0023. Thanks are due for the financial support to CESAM (UID/AMB/50017 - POCI-01-0145-FEDER-007638), to FCT/MCTES through national funds (PIDDAC), and the cofunding by the FEDER, within the PT2020 Partnership Agreement and Compete 2020.
REFERENCES [1]
L. Bornmann, B. F. Bowman, J. Bauer, W. Marx, H. Schier, M. Palzenberger, Standards for using bibliometrics in the evaluation of research institutes in Next generation metrics (B. Cronin & C. Sugimoto eds.), pp. 201-223, Cambridge/ USA: MIT Press, 2014.
[2]
L. Bornmann, W. Marx, How to evaluate individual researchers working in the natural and life sciences meaningfully? A proposal of methods based on percentiles of citations, Scientometrics, vol. 98, no. 1, pp. 487-509. 2014.
[3]
L. Bornmann, A. Thor, W. Marx, H. Schier, The application of bibliometrics to research evaluation in the humanities and social sciences: an exploratory study using normalized Google Scholar data for the publications of a research institute, Journal of the Association for Information Science and Technology, vol. 67, no.11, pp. 2778–2789. 2016.
