Wednesday, September 28, 2005

Correlation and Causality

There are a few hard and fast rules in science that you can pretty much use at all times without fear of contradiction, one of my favourites is 'Correlation does not prove causation'. Essentially what this states (in typically obscure scienc-ese) is that just because event A happened before event B, you can't neccesarily state that A caused B. This sounds obvious huh? but you can find loads and loads of examples where people consistently ignore this basic fact. For instance, we've all heard stories similar to 'my mum had a bad back and then bought a magnetic mattress cover, 3 weeks later, her back felt better'. Feel free to subsitute the exact phrase with variations involving vitamins/horoscopes/alternative medicines.
These people are confusing cause and effect, it is altogether possible that some hitherto unobserved effect of magnetic fields is that it alleviates back pain, far more likely... its a coincidence (or more subtely, a placebo effect). Scientists go to great pains to find correlations and propose a causitive link i.e. observation: people who smoke get lung cancer at a far greater rate than those that don't. hypothesis: smoking causes lung cancer. hypothesis test: if we look at people who smoke and then give up, their chances of getting lung cancer should start to approach the average of the never-smoked as their time spent smoking gets shorter and shorter. This is a predicition based on the hypothesis, and is either true or false, i'm not sure if this has ever been done however, i'm not a smoking expert, but i'd be surprised if it, or something like it, hadn't been confirmed.
If it's true, hooray, your hypothesis gets to live another day while you dream up ever more elaborate tests, if it fails you've got two choices - find a hole in your test or, as will become evident after a number of failed tests, junk the hypothesis i-t i-s w-r-o-n-g deal with it and get onto another, more fruitful, line of enquiry.
This is how science works (or at least, is the ideal it tries to approach, after all, it's done by people who are by and large, just as pig-headed as anybody else you meet with the addition that they may have spent years or decades proposing a theory that is going the way of the dodo. That can be hard to deal with. Or they're science-ho's, people who say anything for a nice fat cheque).
So, next time you find that you have a great day at work every time you rub your belly anti-clockwise when you first get up, keep in mind the phrase 'correlation is not causation'. Of course, if you find that it happens quite consistently, you may be onto something big and perhaps a more formal experiment is in order ;-)


Post a Comment

Links to this post:

Create a Link

<< Home