Thursday 20 September 2007

Lies, damn lies, and bloody idiots.

In my experience, the problem with statistics is not so much that they are regularly prodded by purveyors of half-truths — after all, such shenanigans can be discovered and exposed — but that they are accepted so uncritically by people with no real intention of being dishonest. It's much more difficult to debunk a lie when it has no liars in its origins.

What the hell am I on about now? Well, here's a for instance. The Hotel and Catering Industry Training Board once asked one of their staff to perform a study into recruitment and staffing. He duly conducted a load of surveys and discovered, perhaps interestingly, that a large number of graduates were working as hotel porters. This, of course, was mainly because they were going on to do post-graduate academic work and were earning a bit of easy money over Summer or during a gap year or while studying. But the thing is, that last sentence is the result of common sense, not statistical analysis; it is the result of criticising, not accepting, a perfectly true statistic that makes too little sense. The HCITB employee in question didn't do this; he simply stuck all his data through his stats rules and presented his conclusions to the board, recommending, among other things, a vigorous graduate recruitment program to meet the industry's needs for hotel porters over the coming years. For this, he was rightly fired. True story.

A lot more damage is done to the reputation of the discipline of statistics by this common failure to factor in the thought that what you're measuring is usually too messy to be measured in the way you'd like. See also the profoundly held belief among too many statisticians that the effects of intentional acts can be measured in the same way as the effects of blind chance.

Anyway, 6 out of 10 Britons would rather die than exercise, according to The Mail:

Six out of 10 Britons would not be motivated to do more exercise even if their lives depended on it, a poll has found.

Pick a hundred Britons at random. Place them in a narrow alleyway with high walls. Drive a truck with electrified razor-sharp spikes welded to its front down the alleyway towards them at 6 mph. Apparently, only forty of your subjects will run away.

I realise they have an agenda to push and that they have our best interests at heart, but that's no excuse for the British Heart Foundation to be publishing such utter bollocks. And, even allowing for the number-blindness of your typical journalist, was there really no-one on The Mail's staff who thought to question this? Nah, it's in a poll, so it must be true.

No comments: