I have seen several different metrics that are applied for ranking journals: Impact Factor, SJR, and H-index. Which of these is the most useful and robust for comparisons with respect to the quality and prestige of journals?
- Impact factor is the traditional measure of mean citations per year shortly after publication. It can refer to either the official Thomson-Reuters computation (which has lots of problems), or other competing similar computations.
- SJR attempts to improve on impact factor by using an algorithm like PageRank to more heavily weight citations from "better" publication venues.
- H-index for journals is just the same as for people, essentially looking for the volume of consistently good work. The only large place I know of that applies this to journals is Google Scholar metrics.
All of these are subject to manipulation and gaming, all are highly field-dependent, and none are very good actual ratings of quality. Still, they can be used to get a rough sense of the significance of a venue that you are unfamiliar with.
One difference between the traditional Impact Factor (JIF) (the official Thomson-Reuters score) and the SJR score (SCImago Journal Rank) is that the former is based on the ISI Web of Science database and the latter on the Scopus database. Depending on the field you are in this can translate to differences in coverage, which can affect the relative rank of journals. The SJR score is also a weighted score, which takes journal prestige into account, and this is not the case for the Impact factor.
The H-index is strongly influenced by the number of papers published by a journal, since this translates into the number of "attempts" of a journal to obtain highly cited papers. This is the main reason why journal rankings based on the H-index usually deviates quite alot from ranking based on avarage-based metrics (such as the Impact factor and SJR). When using the H-index to rank journals a small top-journal within a field (based on article impact) can be surpassed by large journals. As an example, journals from the publisher "Annual reviews" generally come out much lower on H-index rankings compared to IF or SJR-rankings. Personally, I find the H-index a poor proxy for journal prestige, but it might be somewhat useful to determine overall influence on a field (as in number of relatively highly-cited papers). Overall, I think that the H-index is much better for evaluating individual scientists than to compare journals. Why Google Scholar chose it as the only score to present when comparing journals is beyond me.
However, there also exists many other indicies of journal performance, influence and prestige. Hocking (2013) provides a comparison between 11 different journal metrics, using ecology journals as a case study. This study show that while all indicies are positively correlated, they form 3 relatively distinct clusters. The traditional impact factor is closely related to SJR as well as to Article influence (AI) and Source Normalized Impact per Paper (SNIP), while the H-index and the Eigenfactor are part of other clusters. This means that journal rankings based on e.g. JIF, SJR, SNIP will be very similar, while rankings based on the H-index or Eigenfactor will emphasize other aspects of journal performance.
Among widely-available indicators, SJR and the Article Influence Score (AIS) are best for characterizing journals. An important property of those two indicators is that they weigh citations with the relevance of the citing journal, whereas in computing the impact factor (IF) or the h-index all citations are considered equally, regardless of the relevance of the citing publication. While all these metrics are typically highly correlated, taking into consideration the relevance of the citation sources can matter especially when trying to assess journals that are not among the top ones. For example, in the case of a journal that would artificially inflate the IF through self-citations or by participating in a citation ring within a publisher, or in the case of a journal hosting publications of a relatively isolated field, the SJR and AIS would remain relatively low if citations would come from low-quality sources. Another advantage is that the typical values of SJR and AIS vary less among different fields than the values of the IF and h-index. There are established methods for field-normalizing the IF (see, e.g., the SNIP indicator), however is more difficult to field-normalize the h-index.