Redundancies in 'H' Index Variants and the Proposal of the Number of Top-Cited Papers as an Attractive Indicator

Ruscio, Seaman, D'Oriano, Stremlo, and Mahalchik (this issue) evaluate 22 bibliometric indicators, including conventional measures, like the number of publications, the "h" index, and many "h" index variants. To assess the quality of the indicators, their well-justified criteria encompass conceptual, empirical, and practical issues: ease of understanding, accuracy of calculation, effects on incentives, influence of extreme scores, and validity. Although they found that some indicators are more attractive for research evaluation purposes than others, the correlation between most of the indicators is (very) high. A generally high correlation between the indicators points out a redundant contribution each to each other. Calculating a comprehensive metaanalysis, Bornmann and colleagues (Bornmann, Mutz, Hug, & Daniel, 2011) came to a very similar result: Only a few "h" index variants make a nonredundant contribution to the "h" index; most of the variants correlate very highly with the "h" index. They presented the first metaanalysis of studies that computed correlations between the "h" index and variants of the "h" index that have been proposed and discussed in the literature. Thus, the study was not about the added value of the "h" index or "h" index variants to other bibliometric indices, e.g., total number of citations. They tested whether there is a generally high (or low) correlation between the "h" index and the variants. The metaanalysis included a total of 135 different bivariate correlation coefficients between the "h" index and an "h" index variant (such as the "g" index). In total, they analyzed 37 different "h" variants. Their results show, with an overall mean value between 0.8 and 0.9, there is a high correlation between the "h" index and the "h" index variants. According to Navon (2009), "high correlations indicate that despite the differences in how the metrics are calculated, there is too much redundancy in the information they yield." Even if the "actual" correlations might be higher than those claimed in the original papers introducing the new "h"-type indices due to systematic sampling selection effects, a mean correlation coefficient of between 0.8 and 0.9 is still high and might not justify the development of more and more "h" index variants. However, not all "h" index variants have a high correlation with the "h" index. Their results indicate that some "h" index variants have been developed that have a relatively low correlation with the "h" index, and it can be assumed that they can make a nonredundant contribution to the "h" index. These variants are mainly the MII (Sypsa & Hatzakis, 2009) and the "m" index (Bornmann, Mutz, & Daniel, 2008). For evaluative purposes, then, the "h" index could be combined with these "h" index variants to better depict research performance bibliometrically. (Contains 1 table.)