Monday, December 03, 2007

Lies, damn lies and statistics.

The Royal Statistical Society organised a half day workshop on the use of statistics in reporting. A number of speakers were allocated 15 minute slots to give us a quick run down on the basics of their trade. Topics covered included: statistical terminology, data collection methods, league tables, rankings and statistical modelling.

Wading through the maths-speak (standard deviation, confidence intervals, regression to the mean, standard errors and distributions) the talks brought up a number of interesting points. The key message of the morning was the need to approach all statistics with caution, and to never accept as given what they appear to prove. The workshop was peppered with examples of numerical misrepresentations: such as the dangers of believing surveys of people's drinking habits (which only account for 50% of total alcohol sales) and the high levels of inaccuracy in ranking systems (which fail to factor in bad/good luck)

The programme was intensive, technical, but ultimately rewarding; I will certainly be more careful when passing on figures in the future.

In a follow up email I was informed that they are considering preparing some resources for their web pages to cover:
basic explanations of statistical terms, principles and methods
and 'plain English’ versions of the above suitable for use in reporting
. . . Which should prove very handy.