Sunday, May 23, 2010


First, the title. It comes from a discussion of the attempt to get an injunction to stop the LHC from starting.
[Law Professor Eric] Johnson quotes the British physicist Brian Cox who is reported to have said: "Anyone who thinks the LHC will destroy the world is a twat." 
That is not an encouraging sign. 
Johnson is well aware that this case may never come to court (although he points out that one like it that raises the same issues may well come about in the future).
So the real test will be how the particle physics community responds, whether with spittle-flecked ire or reasoned argument.
My fellow physicist and friend from grad school had sent this to me, and it sort of stuck as a catch-all subject line for various transgressions against science we saw out on the web.
Second, the concept. A few months earlier, he had sent me a link to the results of a study of an AIDS vaccine in the NYT, accompanied by what is best described as spittle-flecked ire. The quote from the article:
Col. Jerome H. Kim, a physician who is manager of the army’s H.I.V. vaccine program, said half the 16,402 volunteers were given six doses of two vaccines in 2006 and half were given placebos. They then got regular tests for the AIDS virus for three years. Of those who got placebos, 74 became infected, while only 51 of those who got the vaccines did.

"... how hard is it to take sqrt(8000)?" Exactly. If we take half the sample population N = 8201 in one of the groups, sqrt(N) ~ 91 is a quick measure of the size of the expected fluctuation in a measurement on that population due to random chance alone. And 91 is bigger than the difference between the number who became infected, which means categorizing this as a successful trial is dubious.

After some spittle-flecked ire from both of us on the subject, I had mentioned my long running desire to start a blog pointing out these samples of innumeracy or flaws in the application of the scientific method on the web. Not that this is a novel idea. It's more for fun. And my friend may join up at some point.

One of my personal favorites is this graph. It is the worst graph ever made. It is a good example of one of the ways you get into trouble in an investigation: you have a hypothesis you know is right and you try to look for a way to prove it. Economics has started to work that way lately ... especially in the Freakonomics vein: they look for "natural experiments" to demonstrate a hypothesis they have. Another way to put this is to have a result and then look for data that supports it. Sounds like the opposite of science to me.

Anyway, this is the sort of thing that will pop up here. I also like to make political cartoons out of graphs. So those will appear, too. Probably next post.

No comments:

Post a Comment