Wednesday, August 15, 2007

Lies, Damn Lies... and Statistics

Three statisticians and three epidemiologists are traveling by train to a conference. The statisticians ask the epidemiologists whether they have bought tickets. They have.

“Fools!”, say the statisticians. “We've only bought one between us!”

When the ticket inspector appears, the statisticians hide together in the toilet. The inspector knocks and they pass the ticket under the door. He clips the ticket and slides it back under the door to the statisticians.

The epidemiologists are very impressed, and resolve to adopt this technique themselves. On the return they purchase one ticket between them, and share the journey with the statisticians, who again ask whether they've all bought tickets. “No”, they reply, “we've bought one to share.”

“Fools!”, say the statisticians, “we've not bought any.”

“But what will you do when the inspector comes?”

“You'll see.”

This time when the inspector appears, the epidemiologists hide together in the toilet. The statisticians walk up to door and knock on it. The epidemiologists slide their ticket under the door, and the statisticians take it and use it as before- leaving the epidemiologists to be caught by the inspector.

The moral of this story is that you should never use a statistical technique unless you are completely familiar with it.

Retold by Frank Shann in The Lancet 1996; 348: 1392
Anyway, speaking about statistics, it's funny if you realize what most medical researches are about. What most journal articles will give you a result like this:
We are 95% confident that people on average will be about 1.5 - 2.1 times more likely to have accidents if they have obstructive sleep apnoea.
And that's all we are proud of, a double layers of probability. I'm not saying that statistics are bull shit; it's just that with bad curriculum in this subject, most healthcare providers are unfamiliar with it. That makes it easier for those in the know to manipulate it, possibly to the detriment of the patient's welfare.

In fact, statistics is the backbone to evidence-based medicine. If doctors know that doing certain things will very likely benefit your health, they must have learnt it from good studies, and certainly, good statistics.


bluez_aspic said...

Reminds me of Stephen Jay Gould's "The median isn't the message":

I was never fond of statistics - but it made me well-inducted in the art of nefarious statistical manipulation, and to distrust surveys.

And here's a classic on poll manipulation :)

Neptune Chye said...

There're three great lies in the world,

Lies, damned lies, and statistics.

youngyew said...

bluez_aspic: Haha yeah, I think part of why statistics is so often manipulated is because so few people understand it. That gives us even more reason to familiarize ourselves with it, isn't it? :)

neptune: Haha yeah, I think someone said this line. The title of the post (which comes from the quoted story) is a reference to this famous line, by the way.