A Unique Resiliency Among Journal Indicators: Pinski-Narin Influence Weights (IWs) & Self-Citations
Eugene Garfield’s Journal Impact Factor (JIF) has become the global currency for an academic journal’s scientific standing, serving as a proxy for the relative importance of a journal within its field. This metric is derived from the total number of citations of all its eligible articles published during the previous two years divided by the total number of eligible articles. Voting by in-links (first authority value of “hubs and authorities”), where all citations count and contribute equally to the metric, undoubtedly contains blemishes in its simple quantitative embodiment of scientific prestige. Consequently, an updated methodology, the notion of journal influence weights (IWs) by Pinksi and Narin (1976) extends the impact factor to account for variations in the significance of citations, an approach of repeated improvement akin to PageRank for Web pages. IWs are a dimensionless indicator that emerges from graph-theoretic properties of a citation network.
Interestingly, influence weights contain an exceptional resistance to a common taboo in academic circles—an overabundance of self-citations which dishonestly substantiate findings and excessively inflate journal indicators. A study by Fassoulaki, Paraskeva, Papilas, and Karabinis (2000) notes that the JIF of Anaesthesiology in 1995 and 1996 is propelled 35% by self-citations. Recently, Prathap, Mukherjee, and Leydesdorff (2019) demonstrate that Pinski-Narin influence weights (IWs), by first normalizing the citation matrix and performing an eigen-computation, result in a remarkable insensitivity to self-citations. A case study of biochemistry journals finds percent change between journal index values with and without self-citations of no greater than 0.21%. Following recursion, repeated improvement by iteration of matrix multiplication enables IW to be nearly independent of surplus self-citations. A caveat, however, is that IWs should only be used with homogeneous datasets and are meaningful only when studied on a field-by-field basis, after field-normalization.
Self-citation independence is a fortuitous result of two operations: the scheme of normalization adopted, and the repeated improvement of the vector values. When comparing this property to other recursive prestige procedures, only IW has this very desirable property for quality measurement. This may not be a win for lazy, self-siting academics amplifying past works to validate current ‘findings’ and their preferred journal, but it is certainly a win for the integrity of science.
Links:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1828313/
https://www.sciencedirect.com/science/article/pii/S1751157719302950
https://www.sciencedirect.com/science/article/pii/0306457378900663