BBC article on cycling accidents

I love the way people keep going on about the 'small sample size' like they know anything about statistics or probablility. The samples in this case are male and female cyclists in London in a year, the deaths are observations from those samples. I can't be arsed to do the maths but getting seven deaths from the female cyclist population and one from the larger male population is going to be pretty far from the null hypothesis (ie chance).

The Poisson distribution is about this exactly. In small samples, random distribution doesn't look random. It clusters. For example, in an infinite coin-tossing episode, there will be clusters of heads. If you look at the distribution of results close-up, you cannot tell that the chance of getting heads is actually 50-50. I don't know if I'm explaining this very well. I'm exceedingly drunk.