Had a study been set up, selecting some areas at random for 20mph zones (LT ‘Twenty is Plenty’ campaign) and leaving others as before, a proper controlled statistical assessment could have been made.

The great bugbear with statistics and road safety is always lack of data: road accidents are actually very rare events, so a phenomenon called ‘regression to the mean’ occurs.

This means you can have by pure freak chance a few accidents in one area, then nothing for a great length of time.

If you are just responding to these chance clusters by, say, sticking a 20-zone on to it or putting up a camera, then the ‘regression to mean’ effect can fool you into thinking that you are having an effect where you are not.

A large-scale experiment with control groups solves this: you get to compare the treatment with the control side by side with plenty of data and at the end, you know if it works or not.

Scientists always use statistics like this, so they know if an experiment has worked or not.

dan h (via website)