http://arxiv.org/abs/1002.0878

According to this work, the average h-index is a simple function of the total number of citation to an individual (h=sqrt(c)/2, c being the number of citation). In other words, the “typical” h-index does not contain any additional information beside the total number of citations. This has nothing to do with your comments in this article, but it’s just a curious fact that I thought you might find interesting.

So true…. It seems those days you need that Nature paper to get on the tenure track of the Big universities.

This is ironic because this skewed selection reminds me of the arguments of Nassim Taleb on luck and success. If the rewards are highly skewed, then on average it is extremely rewarding to bet everything on a few home-run.

I have seen around me students who bet almost everything on a Nature/Science paper because they were told they needed it for a decent career. Results: some of them will succeed (by luck, talent or perseverence) at getting this Nature (and not much else). This will help to pass over deserving student who got “only” a couple of PRL.

Because Academia is very much a (nonlinear) tournament, a small advantage at the beginning turns into a big reward at the end.

One can certainly argue whether or not two PRB are “about the same” as one PRL, as they would be per “integrated IF” (*). But one can’t very well argue that one PRL is “more” than 5 PRL plus 10 PRB, as it would be per “average IF”.

(*) Personally, I’d say it depends on the papers and on how well they are appreciated by the community. But I am aware that general opinion would go with the one PRL in most cases.

Assume that a paper has an objective quality measure q with weakly correlates to its (future) number of citations c. Moreover, assume that q can be determined in a somewhat costly procedure, and that review for journals essentially amounts to determining the quality of a paper, comparing it to a benchmark value b and accepting/rejecting based on that. Furthermore, assume that editors and authors behave sufficiently market-like to avoid grossly “overqualified” submissions.

Averaged over all papers appearing in a journal, the variation between q and c would (partially) cancel out, leaving the IF as a reliable estimate for b, which in turn serves as a lower bound for q. In particular, the IF could be a better estimate for q than c.

Now, I’m not claiming this model describes reality at all, but maybe some people do think along such lines?

So, having this obligatory part out of the way, I’d like to mention a reason for the use of IFs in such annual scorecards: They are available right now, while citations take longer to accumulate. The main issue is thus the focus on the extremely short timescale.