SCImago rankings

This is an interesting (and free) alternative to ISI’s Journal Citation Reports: http://www.scimagojr.com/

Unlike ISI’s impact factor, the scoring system used by SCImago is based on weighted citations: a citation to a paper counts higher if it is made by authors who themselves have a high score. This leads to an interesting iterative algorithm for estimating scores. It is similar to the affinity propagation algorithm (this idea keeps popping up in different places).

There is an important parameter in this scheme, namely how strongly the citation count depends on the citing journal’s score (how strong is the weighting). With small or no weights, everyone’s citations count equally: you get a “democratic” system, but tend to favor “mainstream” opinions and hype; publications can then go viral, and the system is easy to game. With very strong weights, you avoid these problems, but this can lead to conservative, entrenched systems where only the most cited journals count, and alternatives to this “establishment” are suppressed. A middle ground is probably preferable, but where exactly this lies seems subjective. But it sure is interesting! :-)

It would be great to see scheme this applied to score individual papers rather than journals. Probably it has, but I haven’t seen it. Anyone?