Calculating an impact factor from a larger database satisfies one of the smallest problems with the impact factor. I think it's already been done for the Scopus database for some journals. It is still a bit absurd to draw too much inference based solely from the mean of a highly non-normal distribution (usually looks like exponential decay with a few papers getting many citations and most others having very few). There are many better citation-based metrics of influence. I have a preprint available on the subject comparing 11 metrics for 110 ecology journals: https://peerj.com/preprints/43/
I'm still revising the paper and looking for an appropriate outlet, but as the article suggests, I recommend using the Article Influence score in place of the impact factor (still reporting in the JCR). The Eigenfactor can be used to make other inferences, and the SJR and SNIP (calculated from Scopus) can be used to compare among different fields of study, controlling somewhat for differences in citation and publishing practices. Cheers, Dan
