Thursday, 21 April 2011

Is the h-index a good measure of publication performance?

The h-index of publication performance was first proposed by Hirsch in 2005 as an improvement on total citations on the grounds that total citations do not measure the impact of specific papers and are easily inflated by large numbers of citations to individual papers.  In our paper in Journal of Clinical Nursing  David Thompson and I argued that the h-index, while not perfect, did indicate a sustained body of work.  This editorial has generated a commentary and a further editorial on the use of the h-index for journals as an alternative to the impact factor by Hunt & Cleary.

While the impact factor is unlikely ever to be superseded by the h-index as a measure of journal performance; for the same reasons as it is a superior measure of individual performance, it is also a superior measure of journal performance.  The impact factor - which has the virtue of simplicity, but also arbitrariness (ie the two year window for citations excluding those cited in the year of publication) - is relatively easily manipulated in much the same way as total citations (eg self-citation within the journal and maximising citation by other journals), the h-index (which is highly correlated with impact factor) is less easy to manipulate.  In the same way as more complicated and purportedly meaningful measures of citations such as the eigenfactor and the SCIimago journal and country rank (which are less easy to understand than the impact factor and offer little by way of discrimination between journals) have been developed, more complicated and concomitantly less simple meaures related ot the h-index have been developed such as the g-index.

Therefore, the h-index is a measure of how often a particular set of n papers has been cited n times.  Someone with an h-index of 10 has 10 papers that have been cited 10 times.  Some of these papers may have been cited more than 10 times but this does not influence the h-index score and, to increase the h-index, an 11th paper must be cited 11 times and all of the other papers contributing to the h-index of 10 must also have been cited at least 11 times.  Thus, the h-index is intractable compared with total citations, being less easily manipulated by a high number of citations to any single paper; therefore, it is harder to manipulate by individuals and journals than total citations or impact factor.

No comments:

Post a Comment