Friday, July 27, 2007

Scienciness, again

This row over at Tim Lambert's, also here, reminded me of something I've noticed around the blogosphere. There was this, too, and also this.

They're all arguments from meta-analysis of some sort, and they're all wrong. They're all wrong in the same way, too; the first, David Kane's beef with the Lancet survey of mortality in Iraq, essentially argues that excluding Fallujah from the sample was a mistake not because it tended to underestimate mortality, but because - as including the outlier increases the variance about the mean - it widened the confidence interval enough that it included zero deaths. In fact, Kane goes so far as to suggest a lower 95% bound of -130,000. Some people would stop here and review their assumptions. In this case, that would be that both the prewar mortality rate and the postwar survey samples are samples out of a complete normal distribution.

But rates of mortality can obviously never be zero; everybody dies in the end. Further, there's a pretty obvious upper bound too - otherwise there would be nobody left to survey. And there is no known way in which war reduces mortality rates whilst it's still going on. There is only one way to replace dead people; and it doesn't involve war. It's a damn sight more fun than reading David Kane, too.

When the facts change, said John Maynard Keynes, I change my ideas - what do you do, Sir?

What indeed. Consider the third link; this is another example of changing one's ideas and assuming that it has some impact on the facts. The argument is roughly that, if you assume the human population is normally distributed in time, and that you are at a random point in its existence (you have no reason to think otherwise), then the world will come to an end sooner than you think. The flaw is, of course, that all the information in the argument comes from the initial assumption about the distribution. Claude Shannon would have said that this argument actually contains no information, or at least no net gain of information; all the information in it is contained in the original assumptions, and as Shannon defined information as that which is unexpected in communication, there's no "there" there.

And frankly, Shannon information theory has been a lot more useful than any of this sciency thumb-twiddling. Moving on, Realclimate Gavin assails a paper by a couple of economists who reckon they've disproved climate change because the IPCC TARs don't conform to their definition of a "scientific forecast". I'll confine myself to pointing out that the only substantive points they make confound climate and weather, treat Piers Corbyn as a source of meaningful information*, argue from personal incomprehension of the titles of climatology papers, and assert that:
People will continue to believe that serious manmade global warming exists as they will continue to believe other things that have no scientific support (e.g., the biblical creation story, astrology, minimum wages to help poor people, and so on)
Given that essentially all the necessary pieces of climate change have either laboratory or observational evidence, this is sufficient evidence for me at least to call it partisan hackery.

So what is my point? All these arguments have this in common; they argue from assumption and assume this trumps empiricism, and strangely enough, they all get answers that suit the powerful. They argue as if there was no observable reality out there. It's harder to spot than most because it's dressed in the clothes of science - but that don't make it any better. But I suspect that the authors of this stuff really believe in it; it's quite easy to imagine that giving something a mathematical look actually gives it content. Richard Feynman, I think, would have ripped these people a second arsehole. Quite a few very bright mathematicians came to believe that numbers were the real reality, and I suspect this is what science would look like if Platonists ran it - a search for perfect form beyond the messy, sexy, noisy bazaar of reality.

It's poisonous stuff, of course; I wonder what the Stiftung Leo Strauss would say? And am I right in suspecting they may be the most important blog in the 'sphere today?

*Shannon would point out that he certainly is a source of information, because no-one sane would expect anyone to say the sort of things he does every time his mouth moves.

No comments:

kostenloser Counter