Percentile ranks and the integrated impact indicator (I3)

Loet Leydesdorff,Lutz Bornmann
DOI: https://doi.org/10.1002/asi.22641
2012-01-01
Journal of the American Society for Information Science and Technology
Abstract: We tested Rousseau's (in press) recent proposal to define percentile classes in the case of the Integrated Impact Indicator (I3) so that the largest number in a set always belongs to the highest (100th) percentile rank class. In the case a set of nine uncited papers and one with citation, however, the uncited papers would all be placed in the 90th percentile rank. A lowly-cited document set would thus be advantaged when compared with a highly-cited one. Notwithstanding our reservations, we extended the program for computing I3 in Web-of-Science data (at http://www.leydesdorff.net/software/i3) with this option; the quantiles without a correction are now the default. As Rousseau mentions, excellence indicators (e.g., the top-10%) can be considered as special cases of I3: only two percentile rank classes are distinguished for the evaluation. Both excellence and impact indicators can be tested statistically using the z-test for independent proportions.
What problem does this paper attempt to address?