The relative significance of the H-index

Use of the Hirsch-index (h) as measure of an author's visibility in the scientific literature has become popular as an alternative to a gross measure like total citations (c). I show that, at least in astrophysics, h correlates tightly with overall citations. The mean relation is h = 0.5(√ c + 1). Outliers are few and not too far from the mean, especially if 'normalized' ADS citations are used for c and h. Whatever the theoretical reasoning behind it, the Hirsch index in practice does not appear to measure something significantly new. Citation metrics The Hirsch index, a quantitative measure of an authors' status in the scientific literature, is defined as the paper number n in his list of publications (ranked by citations, highest first) where n equals the number of citations. (In NASA's ADS, click on 'sort by normalized citations' and scroll the results down to the last paper of which the rank is less or equal to the number of citations listed for it). It appears to have become fashionable as a more meaningful metric than, or as an alternative to, the simple metric of raw citation rates. It has been implemented also for quantifying the scientific impact of observing facilities (Grothkopf et al. 2007). The invention of new figures of merit using citation rates implies an attempt to find optimal metrics for an intuitive notion of quality ('A better than B'). Reflection on the way in which these statistics are handled in practice, however, shows that focus on the accuracy or biases of different metrics in abstracto is somewhat misguided. The uncertainty in applying such metrics is related more to the purpose they are to be used for. Unreliable work, for example, can have a high citation rate but this will be discounted accordingly if reliability is more important, depending on the purpose of an evaluation. High visibility in the popular media, on the other hand, can trump reliability, teaching skills can trump citation rates, etc. The substitution of gross citations by a new index like h as an abstract measure of 'quality' or impact tends to obscure this more basic issue. For the average astrophysical career, h scales roughly linearly with the author's career age (time since PhD). Like total citations, it can therefore not be used for comparison between authors unless compensation is made for differences in career age. The total number of all citations …