The revised SNIP indicator of Elsevier's Scopus

The modified SNIP indicator of Elsevier, as recently explained by Waltman et al. SNIP value was obtained by dividing an arithmetic average in the numerator by a median value in the denominator which results in a quotient without the possibility of specifying the statistical uncertainty in the indicator. Instead of dividing aggregates, one should normalize observed against expected values in terms of distributions. This discussion followed our discussion of the " crown indicator " (Opthof & Leydesdorff, 2010) and led to an increasing awareness in the community that one should not use " rates of averages " (RoA), but instead average the rates (AoR) when normalizing values against the expected ones so that valid statistical error measures can be applied (Gingras & Larivière, 2011). The recent modifications to SNIP elegantly solve the problem of normalization by using the same premises as the original indicator, but in the appendix the authors are able to prove that the revised SNIP indicator can also be considered as an arithmetic average (Equation A2, at p. 281). However, the authors also mention that the use of arithmetic averages as expected values is unfortunate in the case of scientometric distributions because these can be extremely skewed (Seglen, 1992 and 1997). Somewhat amazingly therefore, the authors did not consider the possibility of using the median (instead of the mean) or other non-parametric statistics as these