(UPDATE: apparently Thomson ISI have already thought of this – you can calculate an h-index for any search – journal, author, keywords, etc… – you do on Web of Science)
I’ve been mulling this one over for a while. It may mean nothing, it may mean something, I’ll let you decide.
You may have heard of the h-index, it is the highest number of papers a scientist has published that have each received at least that number of citations, i.e., if 20 of your publications each have at least 20 citations, your h-index is 20 (to get up to an h-index of 21, not only does your 21st paper need to receive 21 citations, but all of those other 20 papers – each with at least 20 citations – need to reach 21 cites as well). So, as you can see, increasing your h-index just one point is not necessarily an easy thing to do, and the bigger the number, the harder it is to increase (at least that’s how I see it).
All in all, it seems to be a measure of consistency – rather than just considering the number of papers someone has published – which may include many mediocre ones that are rarely cited – the h-index takes into account how much impact a body of work has made in the community, based on citations.
It got me thinking. How about doing this for journals? Yes, we have the impact factor and associated numerical wizardry, but does that measure consistency (it may well do, I’m just asking…). As you probably know, the impact factor for a journal for any given year, let’s say 2004, is calculated by dividing the total number of citations received in 2004 for all papers published in that journal for the preceding 2 years (i.e., 2002 and 2003) by the total number of papers from that period. So, if The Journal of Marvellous Research published a total of 100 papers in the period 2002-2003, and in 2004 those papers are cited a total of 1500 times, its impact factor is 15.
Here’s my problem with that – what if most of those 1500 citations come from two or three review articles? Or five or six really good primary research papers? In other words, perhaps The Journal of Marvellous Research has published a few gems, but the rest is not quite up to scratch? So, how about an h-index for journals – the J-factor?
Let’s look at the five-year period from 2000-2004 and calculate a J-factor, i.e., the highest number of articles the journal has published during that time which each have at least J citations.
Here are the results:
JACS 133 (papers published 13,606)
Angewantde 114 (papers published 5,423)
ChemComm 75 (papers published 6,655)
Here are the impact factors for 2005 by comparison:
So, JACS has a lower impact factor than Angewandte, but a higher J-factor… does this mean anything? I’m not sure. Obviously JACS published far more papers, but that is not necessarily a measure of quality – after all, ChemComm published more papers than Angewandte in that five-year span, but has a significantly lower J-factor than Angewandte.
Just to be cheeky, here are the 2000-2004 J-factors for two other journals:
Nature 301 (papers published 13,679)
Science 287 (papers published 13,433)
…and compare this to the 2005 impact factors:
So, I’m not really sure what this all means, if anything. The J-factor is a constantly shifting metric, the 2000-2004 J-factors for all of these journals may be different tomorrow, likely different next week, and certainly different next year. Will they change in proportion to one another? That’s something else to watch.
And remember, you can prove anything with statistics…
Stuart Cantrill (Associate Editor, Nature Nanotechnology)