October 22, 2007

Thomson & Predicting the Nobels

I'm updating the post from a month or so ago where I talked about the attempts by Thompson Scientific to use their citation databases to predict the Nobel prizes in various categories.

As I mentioned, last year I chided Thomson ISI about their attempt to predict that year's Nobel Prizes in the various subjects based on their citation data. Since they didn't get a single one of the prizes right, I thought that they should probably give it up. After all, citation count doesn't really measure impact all by itself. Right?

My hopes were in vain as they tried it again this year. From the press release:

Each year, data from ISI Web of KnowledgeSM, a Thomson Scientific research solution, is used to quantitatively determine the most influential researchers in the Nobel categories of chemistry, economics, physiology or medicine, and physics. Because of the total citations to their works, these high-impact researchers are named Thomson Scientific Laureates and predicted to be Nobel Prize winners, either this year or in the near future. Of the 54 Thomson Scientific Laureates named since 2002, four have gone on to win Nobel honors. (Bold is mine. -JD)

Yes, 4 out of 54 is nothing to brag about.

So, how did they do this year? The Nobel Foundation announced the laureates starting October 8th here.

Below are the ISI predictions; I'll include the winners at the end of each section.
Chemistry

Samuel J. Danishefsky
Dieter Seebach
Barry M. Trost

Winner: Gerhard Ertl


Physics

Arthur B. McDonald and Yoji Totsuka
Sumio Iijima
Martin J. Rees, F.R.S.

Winners: Albert Fert and Peter Grunberg


Physiology or Medicine

R. John Ellis, F.R.S; F. Ulrich Hartl and Arthur Horwich
Fred H. Gage
Joan Massague

Winners: Mario R. Capecchi, Sir Martin J. Evans and Oliver Smithies


Economics

Elhanan Helpman and Gene M. Grossman
Robert B. Wilson and Paul R. Milgrom
Jean Tirole

Winners: Leonid Hurwicz, Eric S. Maskin and Roger B. Myerson


Yes, once again Thomson is completely shut out and didn't get any of them correct. On the plus side, they predicted a couple of this year's laureates for last year: Physiology or Medicine and Physics. In my opinion, this is probably worth about half marks or less. After all, they didn't set about predicting who would win the prize "some day" but predicting who would win in a given year. Everyone that they chose is certainly a prominent person in their fields and deserving of recognition. It's just in using their citation messures to predict the prizes on an annual basis that I object to. My main objection, of course, being that citation is only a small part of the story when you're measuring the true impact of a person's career. Not only that, with the explosion of citation databases and citation counting features in publisher databases, the data that Thomson has is only a fraction of what is available. And becoming a smaller fraction each year.

Once again, I sincerely hope that Thomson Scientific stops using the Nobels in a misguided attempt to promote their products.

(To repeat from last time: I have nothing against the scientists and economists that Thomson nominates nor do I want to cast any negative light on the work that they have done. I only want to point out the folly of the methodology that Thomson is using.)

No comments: