causation <> correlation

One of my pet hates is the way science is misrepresented in the media, often through misuse (intentional or otherwise) of statistics.

The sort of claim I'm thinking of might read: "A study has found that those who eat five servings of vegetables per day have a 50% lower risk of heart attack before they turn 50." Without any further qualification, this statement is utterly meaningless. There are two major problems with it: The first and most obvious is that the base risk is not established - the risk is 50% lower than what? If the base risk for someone under 50 who does not eat five servings of veges a day is 1/500, then this suggests that the five-veg-eater's risk would be 1/1000. Even if you're in possession of this information, is a reduction in risk from 0.2% to 0.1% really that significant?

The second problem is the one that bugs me more, and it is potentially as much a fault of the scientific method as it is of the reporting, is whether causation has been proved. It is human nature to look for patterns - it's one way our brains cope with the barrage of information coming at them every second of the day - and often we manage to find them where there really are none. Just because there is correlation between one factor and the result doesn't necessarily imply causation. In this case, you need to know how the study has determined the link between veges and heart attacks. Have all other reasons for the correlation been explored and discounted? Off the top of my head there is a major one that would need to be investigated: does the person that eats five veg per day have a significantly healthier lifestyle overall? Do they eat less saturated fats? Do they exercise regularly? I can see that both of those things might be likely.

But then that wouldn't make such a sensational headline would it?