>I hate silver and grey cars because they are impossible to see at dawn and dusk. However I was very surprised to read this:
>
>http://www.bmj.com/cgi/content/full/327/7429/1455

The story sounds convincing, but in the end still inconclusive. First, there's a few hundred cars which did crash, then about the same amount of those which did not. It may come down to about a hundred of those of a particular color, which is IMO dangerously close to a small sample. What was it, 30 cases for a single variable, right? And we have dozens of variables here.

Besides, why take a sample? There should be official data somewhere, covering all the cars in the country. Take them and start data mining.