How to Lie with Statistics by Darrell Huff.
From the introduction:
With prospects of an end to the hallowed old British measures of inches and feet and pounds, the Gallup poll people wondered how well known its metric alternative might be. They asked in the usual way, and learned that even among men and women who had been to a university 33 per cent had never heard of the metric system.
Then a Sunday newspaper conducted a poll of its own – and announced that 98 per cent of its readers knew about the metric system. This, the newspaper boasted, showed ‘how much more knowledgeable’ its readers were than people generally.
How can two polls differ so remarkably?
Gallup interviewers had chosen, and talked to, a carefully selected cross-section of the public. The newspaper had naively, and economically, relied upon coupons clipped, filled in, and mailed in by readers.
It isn’t hard to guess that most of those readers who were unaware of the metric system had little interest in it or the coupon; and they selected themselves out of the poll by not bothering to clip and participate. This self-selection produced, in statistical terms, a biased or unrepresentative sample of just the sort that has led, over the years, to an enormous number of misleading conclusions.
Machiavelli it’s not, but as the tweet from Stat Fact says, #classic.
[…] a good thing I posted about How to Lie with Statistics […]
Pingback by FISA Court Subpoena Data Released! « Another Word For It — February 3, 2014 @ 9:42 pm