Elio Valladares, who is at the University of Virginia, has completed a simulation that is very interesting because it looks at the problem of coincidences at the ¨mesa¨ table level, rather than at the machine level.
Recall that while the CD was talking about the anomalies in the number of coincidences at the center level, the Carter Center and the CNE were quick to dismiss that there was no such anomaly and the results were reasonable. Recall also that each Center may have a number of ¨mesas¨ tables and that each table may have one or more machines. Thus the two sides seemed to be talking about two different things, coincidences in the tables, of which there were 402 for the SI´s and 311 from the NO´s in our review of the machine results, or coincidences in the machines of which there were 805 in the Si´s at the center level or 647 for the No´s. However, the coincidences for the Si´s at the Center level translated to 1879 machines.
In its final report, the Carter Center said that it had consulted a Prof. from Stanford University who we understand was Jonathan Taylor from the Dept. of Statistics of that University. According to the final report by the Carter Center, these results based on the table coincidences are ¨probable¨. However, no details has been ever been given of how exactly this conclusion was reached. In the mean time, studies showed that the coincidences at the machine level were not that probable.
What Valladares has done is to simulate the probability of a coincidence using the total number of votes at each table, from the results for the referendum. He then uses the real numbers to simulate 10,000 elections and calculates the number of times these coincidences occur. The results for this calculation are shown in the plot below:
This is the total number of coincidences seeing at the table level, the distribution peaks around 345 with a fairly narrow distribution. According to the study, Valladares concludes that the probability of having 393 coincidences for the SI is 0.0028, which according to him is the number of Si coincidences per table reported by the CNE. Our calculations are that the number is 401 which is even less likely to occur.
Even more interesting is the fact that the number is of ¨NO¨ coincidences in the same calculation is not found to be so unlikely, with a probability of 0.17 of finding 311 cases in which the No´s coincide.
It would be interesting to know of either Prof. Taylor or the Carter Center have anything to comment on this, as Valladares´ results contradict their conclusion and tend to support quantitatively the thesis that there was some form of fraud on Aug. 15th.