A comparison of average-based, percentile rank, and other citation impact indicators

The main aim of this paper is to defend the view that, in spite of the broad agreement in favor of the MNCS and the percentile rank indicators, there are two other citation indicators with desirable properties that the above indicators do not posses: (i) a member of the family of high-impact indicators introduced in Albarránet al. (2011), and (ii) a new indicator, based in the work of Herrero & Villar (2013), which measures the relative performance of the different research units in terms of a series of tournaments in which each research unit is confronted with all others repeatedly. We compare indicators from the point of view of their discriminatory power, measured by the range and the coefficient of variation. Using a large dataset indexed by Thomson Reuters, we consider 40 countries that have published at least 10,000 articles in all sciences in 1998-2003. There are two main findings. First, the new indicator exhibits a greater discriminatory power than percentile rank indicators. Second, the high-impact indicator exhibits the greatest discriminatory power. (Author)