Numbers Prove Nothing
Numbers can prove lots of things. But not in the case of supposed election fraud in Iran.
Bernd Beber and Alexandra Scacco’s Washington Post story of statistical anomalies seems at first read to prove the Iranian election was a fraud. I would have been happy to have been convinced. Sadly (because having that proof would have made complaints of a lack of concern for democracy against the government of Iran much stronger) the numbers prove nothing of the sort.
John Graham-Cumming explains (from a statistician’s point of view) why the Washington Post’s analysis is faulty. I found his article hard going in places (I did stats at university when I studied psychology, but only one semester).
Hannah Devlin’s article on Times Online is a bit easier to follow.
This doesn’t mean the election was OK. I still think it probably was not. It just means statistics based on oddities in the count are not going to give us a definitive answer.