> On Thu, 24 May 2001, Tim Brody wrote (about my proposed 2nd criterion for
> evaluation of an eprint archive, which was: 2) its suitability for
> yielding citation data [an 'impact-ranking' criterion?]):
>
> [tb]> One might also add the facility to export "hit" data, as an
> [tb]> alternative criterion (or any other raw statistical data?).
>
> What kind of raw statistical data might be most useful, in the future, for
> 'impact-ranking'?
Perhaps the beginning of the answer lies in what can be measured, then what
can be measured accurately, and lastly what is useful to users.
The first part is (in no particular order): hits, citations, authors,
institutions, countries, dates, and sizes, ...?
> At the arXiv archive, one section of the FAQ section (under Miscellaneous)
> addresses the question: "Why don't you release statistics about paper
> retrieval?". (See: http://xxx.lanl.gov/help/faq/statfaq).
>
> The short answer provided is: "Such 'statistics' are difficult to assess
> for a variety of reasons". The longer answer also includes the comments
> that:
> [*snip* accentuates faddishness]
> And,
> [*snip* big brother is watching]
> Thought-provoking comments?
I would say there are better reasons than the two you chose, some of which
mentioned by arXiv. For example, no system administrator would appreciate
someone downloading a paper 1000 times just to up their hits!
Also, as pointed out by arXiv, knowing how little one's research is read or
cited could put a researcher off arXiving all together.
(I provide an example of such statistics from cite-base, but leave it to the
user to decide whether they are useful or not)
All the best,
Tim Brody
Received on Wed Jan 03 2001 - 19:17:43 GMT