The message of this book is that most of us, including, apparently, a number of senior civil servants, have almost no understanding of numbers. But we can improve our numerical literacy by asking simple questions when we are offered figures by the media. And it's not only the media that prints misleading information; so, too, do government departments.
Sophisticated mathematics are not needed to make sense of the figures we are being continually bombarded with from all kinds of sources; all that is required is a little reflection and the application of common sense. This emerges, for example, in media statements of percentage increases or decreases. A 50 per cent increase in something bad sounds alarming, but unless we know the absolute number of cases involved, it is meaningless. A 50 per cent increase in the incidence of a disease that affected only 1 in 100,000 people would not be a great cause for alarm.
Apparent correlations may be due to chance, but are often taken to indicate a causal link. The consequences of making mistaken inferences of this kind can be serious. The authors cite the cases of two doctors who were suspected because the mortality in their practices was abnormally high. This occurred in the wake of the notorious Shipman murders. But further analysis showed that the two doctors in question had unusually large numbers of nursing homes in their areas, and this had resulted in the increased numbers of deaths.
Averages are a frequent source of confusion. They can be falsely skewed if they include a few individuals or groups that lie well outside the the usual range. This does not mean that we should never use averages but we should look carefully at what they include.
Relative risk seems to be something that humans find particular difficulty in estimating. Scare stories in the media are a manifestation of this. One of the examples given here is the mobile phone alarm: in 2005 the president of the British Radiological Protection Board recommended that children should avoid using them because of the risk of brain tumours. But, in reality, the increased risk (if any) was tiny, amounting to a rise from 1 in 100,000 to 2 in 100,000. And a later, larger study did indeed show that the association was a statistical fluke.
Although we should be careful in accepting statistical pronouncements, whether official or in the media, at face value, the authors warn us that we ought not to rush to the opposite extreme and ignore all figures. 'The proper response is not to rubbish every statistical relationship but to distinguish between those that have taken some thought and those that were a knee-jerk.' We also have to watch out for the deliberate manipulation of statistics by those in authority for their own ends.
This book is an excellent corrective to the temptation to sloppy thinking that many of us have.
11 March 2009