Thursday, January 30, 2014

Silver's "The Signal and the Noise"

There are some errors in Silver's (excellent? interesting? occasionally vexing?) book from a little over a year ago, The Signal and the Noise, about prediction and Bayesian theory. The two I noticed are strange, because they are very basic. (Other people have noticed these errors.)

On page 269, the book says how 20 x 20 is 4,000, when it is actually 400. This is pretty basic. Would Silver make this kind of error? Was he rushed? Did he and the editors miss it? Did someone else ghost write parts of the book? If this error is here, what are the errors that I didn't notice?

There is also an error or two with the arrows-length illusion image on page 367. For one, the description is not clear, and it seems that Silver is describing the illusion in the wrong way (that is, he says the one that appears to be longer appears to be shorter). However I measured the lines of the arrows in my printing, and indeed the one which is supposed to appear longer via the illusion is indeed longer by a millimeter. It looks like a printing error, where the lines may have been the same length, but when the arrow heads were added, they were added compactly to the "will look shorter" line but were not to the "will look longer line". The thickness of the added arrow heads is just enough to throw the whole illustration off. So not only does Silver's description get it wrong, but then the illusion isn't an illusion, one is actually longer than the other.

This really undermines the entire book, since I'm reading it to learn about things I don't know, and if Silver can't get right the things I do know then I know I have no idea if he is getting right all of the material I don't know.

Unrelated to these errors, I did like this review: http://mathbabe.org/2012/12/20/nate-silver-confuses-cause-and-effect-ends-up-defending-corruption/