Self-citations again

[Quentin L. Burrell]

My recent note on this topic provoked some interesting responses and I'd like to comment briefly on some of thepoints raised.

(i) Eric Ackermann. While I have symapathy with Eric's scholar working in a new or unfashionable area, I would contend that discounting self-citations does not necessarily imply that they are in any way tainted. My view is that if one is doing a citation analysis, it is not to measure the inherent worth of a paper but simply to measure in some way the impact which the paper makes. But it must be the impact on the wider scientific/scholarly community which is paramount, in which case discounting of self-citations must take place. This may seem harsh on the solitary scholar, but a voice in the wilderness is just that until someone hears!

(ii) Albert Tabah. Much of what I have said above is also relevant here. So far as the Tony van Raans and Gene Garfields of this world are concerned, I am sure that they would have little to fear using a discounting scheme since the true value of their work is marked by the evident extent to which it is referred to and used by other workers. Again it is the impact on the wider community which is important, and which provides the evidence!

(iii) Ronald Rousseau. I would have thought that the lower bound is given by the scoring system which completely omits self-citations, a totally discounted system. In the original set-up we imagined a total of N citations, a proportion p of which are self citations so the totally discounted score is (1-p)N, while the suggested "discounted citation score" had (1-p^2)N. The full score of course is just N.
Note that since 1 > 1-p^2 > 1-p, the DCS does give a sort of mid-way discounting.

Finally, I would hesitate to support any method of applying subjective judgments to an author's intent in making self-citations but, because of the external impact consideration, would argue in favour of some sort of automatic discounting. The DCS has a nice intuitively reasonable construction and a simple resulting formula to apply. I do not have access to the relevant citation databases so if anyone would like to try it out, I would be most interested to hear from them either directly or through the list.

Thanks for comments.

Quentin Burrell



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.asis.org/pipermail/sigmetrics/attachments/20001213/bb2adca2/attachment.html>