But you are assuming with your 7.7% significance that the two issues they identify are totally independent. If they were both significant at the 10% level, and you specified beforehand that we know these are two problems that often are found in electoral fraud, the combined significance would be much lower than 10%. The article describes it as significant at the .005 level when both are considered. And even 10x this level of significance is enough for me to feel comfortable that the Irani election was fraudulent considering the polling data that exists prior to the election and the fact that the government is willing to shoot people rather than do a proper recount.

See, that's why I shouldn't type after about 10 pm.

Point # 0: I think the election was probably fraudulent, or that at least the margin was manipulated.

My post, really, was about stats. Here's the deal:

You're mistaken about my assuming anything -- I went and got the data, ran the first test the authors describe, and got a different result. I verified, read the comments and found that others had found the same thing. Once I was sure the authors had made one error, everything after that -- including the 0.005 -- simply shouldn't be trusted.

I didn't look at the second test at all, but like I say, I'm not encouraged given how simple the first test ought to have been.

Your point re:independence is also not accurate. If two events are dependent, the probability of them both occurring together is *greater* than if they were independent. If it's raining for the 50th day in a row, for example (uuuuuuuh) you don't say, "And what's even *more* unusual is that it's been cloudy every one of those days, too!" Of course it has -- clouds and rain are dependent events.

Re: prior polling, that's a valid point, something which could also be studied with a statistical test. I haven't seen one.

In re: your comfort level with these events, that's your call -- like I said, I also think that it's likely that the election was screwed up. But the fact of the matter is that the 'evidence' that these authors provide doesn't indicate that with the certainty that they imply.

From where I sit, if something's going to happen once every 10-20 times entirely by accident, then it's hardly a big deal when it does. I mean, if you flip a coin 4 times and it's heads every time, are we supposed to freak out and call the mint?