I really like this article. Its nice to see a bit of empiricism coming into the journal/impact factor/citation world.

My first and certainly simplest recommendation for improving the IF calculation is to use a median calculation and not a mean. This is all too obvious when looking at the IF of, for example, 'Briefings in Bioinformatics', which currently has an IF of 24.3. A large proportion of its citations, however, have come from one paper describing Mega 3. Although the editors freely admit this has biased the IF, a median calculation would have given us a much better value right from the start