July 4, 2013

May university rankings help uncover problematic or fraudulent research?

Paul Wouters
Can one person manipulate the position of a whole university in a university ranking such as the Leiden Ranking? The answer is, unfortunately, sometimes yes – provided the processes of quality control in journals do not function properly. A Turkish colleague recently alerted us to the position of Ege University in the most recent Leiden Ranking in the field of mathematics and computer science. This university, not previously known as one of the prestigious Turkish research universities, ranks second with an astonishing value of the PP(top 10%) indicator of almost 21%. In other words, 21% of the mathematics and computer science publications of Ege University belong to the top 10% most frequently cited in their field. This means that Ege University is supposed to have produced twice the amount of highly cited papers as expected. Only Stanford University has performed better.
In mathematics and computer science, Ege university has produced 210 publications (Stanford wrote almost ten times as much). Because this is a relatively small number of publications, the reliability of the ranking position is fairly low, which is indicated by a broad stability interval (an indication of the uncertainty in the measurement). Of the 210 Ege University publications, no less than 65 have been created by one person, a certain Ahmet Yildirim. This is an extremely high productivity in only 4 years in this specialty. Moreover, the Yildirim publications are indeed responsible for the high ranking of Ege University: without them, Ege University would rank around position 300 in this field. This position is therefore probably a much better reflection of its performance in this field. Yildirim’s publications have attracted 421 citations, excluding the self-citations. Mathematics is not a very citation dense field, so this level of citations is able to strongly influence both the PP(top10%) and the MNCS indicators.
An investigation into Yildirim’s publications has not yet started, as far as we know. But suspicions of fraud and plagiarism are rising, both in Turkey and abroad. One of his publications, in the journal Mathematical Physics, has recently been retracted by the journal because of evident plagiarism (pieces of an article by a Chinese author were copied and presented as original). Interestingly, the author has not agreed with this retraction. A fair number of Yildirim’s publications have been published in journals with a less than excellent track record in quality control. The Elsevier journal Computer & Mathematics with Applications (11 articles by Yildirim) has recently retracted an article by a different author because it turned out to have “no scientific content”. Actuallly, it was an almost empty publication. According to Retraction Watch, the journal’s editor Ervin Rodin has been replaced at the end of last year. He was also relieved from his editorial position at the journal Applied Mathematics Letters – An International Journal of Rapid Publication, another Elsevier imprint. Rodin was also editor of Mathematical and Computer Modelling, in which Yildirim published 5 articles. The latter journal currently does not accept any submissions “due to an editorial reconstruction”.
How did Yildirim’s publications attract so many citations? His 65 publications are cited by 285 publications, giving in total 421 citations. This group of publications has a strong internal citation traffic. They have attracted almost 1200 citations, of which a bit more than half is generated within this group. In other words: this set of publications seems to represent a closely knit group of authors, but they are not completely isolated from other authors. If we look at the universities citing Ege University, none of them have a high rank in the Leiden Ranking with the exception of Penn State University (which ranks at 112) that has cited Yildirim once. If we zoom in on mathematics and computer science, virtually all of the citing universities do not rank highly either, with the exception of Penn State (1 publication) and Gazi University (also 1 publication). The rank position of the last university, by the way, is not so reliable either, as indicated by the stability interval that is almost as wide as in the case of Ege University.
The bibliometric evidence allows for two different conclusions. One is that Yildirim is a member of a community which works closely together on an important mathematical problem. The alternative interpretation is that this group is a distributed citation cartel which not only exchanges citations but also produces very similar publications in journals that are functioning mainly as citation generating devices. A cursory look at a sample of the publications and the way the problems are formulated seems to support the second interpretation more than the first.
But from this point, the experts in mathematics should take over. Bibliometrics is currently not able to properly distinguish sense from nonsense in scientific publications. Expertise in the field is required for this task. We have informed the rector of Ege University that the ranking of his university is doubtful and requested more information from him about the position of the author. We have not yet received a reply. If Ege University wishes to be taken seriously, it should start a thorough investigation of the publications by Yildirim and his co-authors.
If you see other strange rankings in our Leiden Ranking or in any other ranking, please do notify us. It may help us create better tools to uncover fraudulent behaviour in academic scholarship.

.
.

Popular Posts