[Quote] The “Bible”
Although it is a great read, Numerical Recipe[1] is no more suitable as a statistical bible than Ptolemy is for astronomy.
J. Horowitz, Page 255 of Statistical Challenges in Modern Astronomy II (Eds. G.J.Babu and E.D.Feigelson), Springer, 1997
I wonder if the new edition of Numerical Recipe could change the name Ptolemy to someone or some book from the 20th century.
- W.H.Press, S.A.Teukolsky, W.T. Vetterling, and B.P.Flannery, 2nd ed., 1992[↩]
vlk:
I think Ptolemy gets a bad rap. For someone modeling geocentric projections of planetary paths as perturbative Fourier components, he was way ahead of the times.
01-08-2008, 10:36 pmTomLoredo:
Ouch! Quite a remark by Horowitz. I sympathize with it to a great extent. But I think many NR detractors mistake its purpose. I don’t know if it’s put so clearly in the books themselves, but Press & Teukolsky, in a paper on the history of the books (in Computers in Physics IIRC), say their goal was to get scientists up to 1960‘s state-of-the art in numerical technique (i.e., not circa 1990, the time the books were orginally written). Their contention is that 90% of scientists’ numerical work is adequately handled by old technology. I think this was a reasonable goal/motivation, and that they achieved it well.
As to statistical practice in the new edition, apart from a nod to MCMC, it really isn’t much improved over the original (which was okay in most respects, as far as it went). I was particularly disappointed to see no significant changes in the section on the bootstrap, which has misled several astronomers. At the least, the reference list should have been updated.
What I would have liked to see more of (in both the old and new editions) is “diagnostic” advice regarding whether your problem is adequately addressed by 1960s technology, and then a pointer or two to modern developments. I also wonder if statistics isn’t really in the same boat as numerics, when it comes to the adequacy of 1960s technology for modern problems. But I suppose every expert in some particular area will have a gripe with NR in that area (and probably happily use it in other areas!).
The NR project is so ambitious that I don’t suppose there is anything they could have done that would have really satisfied someone with significant expertise in any one of the many areas they cover. However, as a non-expert in many of those areas, I am very grateful for the books, which helped introduce me to many tools that I now take for granted. It’s hard to imagine a better job being done on that breadth of topics without it becoming a many-volume encyclopedia!
01-09-2008, 6:38 pm