Friday, March 20, 2009

The Oops-Leon Particle

I think this is a great 3-paragraph story:

In short, in 1976 Fermilab thought it discovered a new particle of matter, but turned out to be a mistake. It was originally called the "upsilon", but after the mistake was caught, it was referred to as the "Oops-Leon", in a pun on the lead researcher, Leon Lederman. I love that wordplay.

The other thing I love is that, like all modern science, the mistake is partly due to statistics, which we must understand as being based on probability. Looking at a spike in some data, it was calculated that there was only a 1-in-50 chance for it not to have been caused by a new particle (that is, a P-value). But with further experimentation it turned out that that was a losing bet; it actually had been some random coincidence that caused the data spike.

That's the kind of thing you need to accept when using inferential statistics; all the statements are fundamentally probabilistic, and some times you're going to lose on those bets (and hence so too with all modern science). Apparently the new standard before publishing new particle discoveries is now 5 standard deviations likelihood, or more 99.9999% likelihood that your claim is correct.

And you know what? Someday that bet will also be wrong. Such is probability; so is statistics; and hence so is science.

No comments:

Post a Comment