To mark European Statistics Day, here is a study conducted by NCI psychology lecturer, Dr Fearghal O’Brien, as part of his postdoctoral fellowship at the US National Institutes of Health, which demonstrates the application of statistics in helping us verify and understand our observations of the world around us, allowing us to respond.

We’ve all observed that teenagers simply don’t listen to good advice, they have to learn from their own mistakes – but there are some mistakes so serious that we hope they never make them.

Analysing data on driving behaviour and car crashes amongst young drivers, the study notes an oddly positive result of teens experiencing a bad collision, i.e.; one that is “police-reportable”, causing airbag deployment, major damage, injury or a rollover – there is an immediate change for the better in driving habits.

The study was conducted using accelerometers and cameras placed in over 250 cars across the United States. While these instruments allow for the investigation of real-world driving behaviour, the huge amounts of data that they produce present some non-trivial analytic problems. The team dealt with this big data by effectively constructing separate generalized linear mixed models for numerous time-points, e.g. before the collision, one month after the collision, two months after the collision, etc.

"Crashes are not a good thing and every driver should try their best to avoid having one,” said Dr O’Brien. “However, our data shows that young drivers learn from the feedback from a crash – there is a subsequent reduction in rapid acceleration, turns, stops and other aggressive manoeuvring for at least the following two months. More research is needed to find out how much longer these changes last and – most importantly – to use this knowledge to find safer ways to get young drivers to correct their risky driving behaviours.”