Sunday, August 23, 2009

Statistics is cool

During vacation I read this article about statistics being the job of the future (also, refer to this talk by the chief economist at Google). I've posted a lot about statistics before (here, or here, or there for example), and I seem to use it a lot in my job, especially when dealing with large sets of data on a lot of DUTs. So that article naturally jogged my brain to recall a conversation I had last month with some other engineers at work.

The field of computational physics has existed for a long while - at least as long as relatively cheap computer time. The premise of comp. phys. is that many of today's physics problems don't lend themselves to elegant mathematical solutions. You have to apply numerical techniques to simulate, prove, or disprove a theory. When I was in grad school two decades ago I knew a post-doc whose entire job (at least while I knew her) seemed to consist of running Monte Carlo simulations on potential collision results at Fermilab: computational physics.

But let's go one step further. Decades ago one of my favorite books was the award winning Startide Rising by Dr. David Brin. One premise of the novel was that, while other alien races used brute force numerical calculations to figure out engineering, the upstart humans were pursuing solutions using mathematical proofs, theorems approximating effects, etc. Good stuff, but I suspect that the aliens are right. I predict that as science progresses, more and more discoveries will be made using powerful computational tools, not "elegant mathematical formulas."

I'm not sure how much that last, rather speculative paragraph has to do with test engineering. But if scientists can theorize about something, sooner or later it has to be tested. Until then it's just theory. And I think that testing will increasingly involve statistics.

No comments: