The BBC has a story about quintuplets born in the UK last Saturday. On the radio news, they said the odds against natural quintuplet birth are 65 million to one. Which makes me think about our perceptions of probability
A previous set of natural quins (ie without fertility drugs) was born in the UK in 2002. They are only the 12th known set. (The latest set were born here only because their Russian mother came here for good medical care, so the UK doesnt have a statistically significant quintuplet tendency!)
Which makes me also think of the tragic cases involving Sir Roy Meadow, about whom the least said the better. He testified in court that the chances of two cot deaths occurring in the same family were 73 million to one, and partly on the strength of this a jury found Sally Clark guilty of murdering her two sons in 1999. (She was cleared by a Court of Appeal in 2003, released, and recently died of acute alcoholic poisoning after suffering “enduring personality change after catastrophic experience, protracted grief reaction and alcohol dependency syndrome.”)
So the fact that something is a 65 (or 73) million to one outside chance doesnt mean that it cant happen.
According to a very interesting website about probability, the chances of winning the UK National lottery jackpot are 1 in 13,983,816.
According to Wikipedia, 11.3 people are killed on UK roads per 1 billion vehicle kilometers.
According to my (shaky) maths, that means one person is killed for every 88,495,575 km driven. So if you drive more than 7km to buy your lottery ticket, you are more likely to be killed than you are to win the National Lottery. Yet lots of people do just this every week
I touched on the issues around probability in an earlier posting. The Sally Clark case should serve to remind us that probability, though useful to modellers and simulators, is widely misunderstood: but also that it sometimes matters very much that we get the maths and the modelling assumptions right.