Statistics are dually known as useful and misleading. Another relevant saying is that a little knowledge is a dangerous thing.
CAR magazine used to have a short section covering the number of complaints received from readers separated according to car brand. The problem with that sort of analysis is that it ignores the relative number of cars from each brand on the road. The “exposure” of toyotas to problems is much higher than maseratis since there are rather more toyotas on our roads. If CAR magazine received an equal number of complaints from drivers of maseratis and toyotas, it would suggest anything but an equal likelihood of having problems from each of those brands. I obviously wasn’t sufficiently convincing when I offered to help them devise a less biased measurement criterion.
This is an example of a common problem with random events – it is important to consider what could have happened as well as what did happen when understanding the results.
In a related example, iAfrica has an interesting article on car hijackings in South Africa. They include a list of the top ten hikacked cars. The list is interesting, but difficult to interpret without knowing how many of each vehicle were hijacking AND how many were on the road, able to be hijacked. If the Maserati Gran Turismo were on the list, I would be wary of driving one! Continue reading