Sunday, October 02, 2011

Unconditional Justice

My Bayesian friends and colleagues are not going to be happy about a UK court ruling that Bayes' theorem cannot be used to analyze statistical evidence in trials.

I have five comments:

(1) Is one still allowed to use Newton's laws for evidence in a course case? You know, just in case gravity might be relevant. about a minute ago

(2) Maybe this is what EPSRC had in mind with respect to building UK capacity in statistics?

(3) Seriously, what the fuck?

(4) Facepalm!

(5) I guess UK justice is unconditional after all.

(Tip of the cap to Mariano Beguerisse Díaz.)

Update (10/03/11): Cosma Shalizi has posted some nuanced commentary on his blog (much more nuanced than my commentary, as I was quite obviously in full attack mode). As I stated in my response to his comment in this space (when my head was a bit cooler than when I wrote the original post), I do agree with his point in general, so we will see if reasonable uses of Bayes' theorem remain permissible in UK courts. I am most definitely cynical enough to doubt it, so I am still not happy about the situation, but we will see what happens in practice. Yesterday, I admittedly went into my usual Kill-Billish red-alarm mode when I saw the article in The Guardian. However, Cosma and I might have to have some words about confounding me with Brits. :) I am only Brit-employed, and you won't find me going out in the midday sun* anytime soon.

* Except for the last few days, in which we actually had midday sun.

2 comments:

Cosma said...

I like harshing on lawyers as much as the next scientist, but actually reading the news story suggests something much more reasonable. The judge on the appeals court was not, it seems, throwing out Bayes's theorem, but rather refusing to be cowed by Bayes's theorem when the base rates and the likelihoods appearing in it are wild-ass guesses. To quote the news story you link to: "And so he decided that Bayes' theorem shouldn't again be used unless the underlying statistics are "firm"." This is, of course, the completely correct attitude; otherwise, the Bayesian posterior is simply without any evidential value whatsoever, and the difference between an expert stating "I'm, like, really sure" and "My posterior probability is 0.99" is entirely spurious precision.

Mason said...

I do agree with that completely. But I also think there is a happy medium between using it intelligently rather than blindly and not being allowed to use it at all. I think one should be allowed to use it reasonably and give the argument behind things. I think that the notion of "firm" will amount to throwing it out entirely rather than not being stupid with it, but hopefully I'm wrong and the notion of firm will be what you describe.

In short, I suppose that my cynicism got the better of me a bit when I wrote this.