Shoe Money Tonight

Occasional ramblings by an anesthesiologist/mother (and sometimes her husband).

Tuesday, July 14, 2009

Statistical Illiteracy Again

Mark Twain, quoting Benjamin Disraeli, famously said that "There are three kinds of lies: lies, damned lies, and statistics." They were only partially right. Science doesn't lie, people do. Statistics do not lie; people's analyses do.

The chief problem occurs when people try to say that statistics are saying something they're not. Sometimes this is intentional. Other times, people are so set on the conclusion, that they impose it on the data. Worst of all is the confusion of correlation and causation.

I have written about this problem before. To make a long story short, assuming that correlation equals causation would lead people to conclude that carrying a lighter causes lung cancer, that wearing a helmet causes head injuries, and that having wrinkled skin causes drowning.

So it is with a recent, much touted study, in Scientific American. The study claimed that swearing increases one's tolerance for pain. If the article is true, however, the study said nothing of the sort.

Here is the key quote:

During the chilly exercise, they could repeat an expletive of their choice or chant a neutral word. When swearing, the 67 student volunteers reported less pain and on average endured about 40 seconds longer.

According to this, then, the subjects chose whether they were swearing or saying some neutral word. The second sentence is pretty ambiguous, but the first does seem to indicate that they were given a choice of what to do. If so, all that proves is that those using profanity had a greater tolerance for pain.

Even if the students were randomly selected, or if each did both (half swearing first, half not of course), then this still only tells us something about this particular small subgroup of college students who happened to volunteer. Going from the specific to the general is a major logical fallacy. People will make the obvious jokes. MSM outfits will draw wild conclusions. Scientifically, though, all this tells us is that there may be something here. It proves nothing about anything.

The numbers are the numbers. But people are still people.