A reader, Mercedes Rosas, has pointed out in the comments below to a correction in the webpage by Prof. Jonathan Taylor of Stanford University and an e-mail from Prof. Taylor to her, on the results of the recall vote in which he says he made an error earlier and corrects hs results. Prof. Taylor’s work was used by the Carter Center to “show” that the number of “Si” coincidences in the mesa (table) votes was reasonable. This result was widely used and quoted by the international press as part for the “evidence” that there was no evidence of fraud in the recall vote.
As I have reported elsewhere, Elio Valladares got quite a different result, suggesting that the probability was not that “reasonable”, in fact Elio obtained that it was small, if not miniscule. Now Prof. Taylor has corrected his work on his web page and I would like to quote him so that there is no misinterpretation of what he says or not:
“ It seems that an expected number of ties between 345 and 350 is reasonable, as it came out from many different models. Using the Poisson assumption to estimate the standard error, it seems then that the probability of observing 402 or more ties for SI is between 1 and 3 in 1000. While this probability is small, I do not feel that it should be interpreted as overwhelming evidence of fraud.”
Yes, is not overwhelming, but we have gone from reasonable to small, but it was the reasonable that led the Carter Center to its conclusion. What would they say now?. By the way, the CNE also used this result by Prof. Taylor to say that the Si vote coincidences were irrelevant.
Prof. Taylor has acted with the integrity characteristic of scientists, I wonder if the Carter Center will post a clarification to their conclusions, but doubt it. I sure hope Prof. Taylor will now calculate the coincidences at the machine level. I believe in that case he will find that the probability is even lower, if not impossible!!! That should have been the case that the Carter Center should have had him study to begin with!
Note added: I have now received a new study by Elio Valladares in which, if I understand correctly, he simulates the vote at the level of the cuadernos (each cuaderno is a machine in centers with electronic voting) rather than at the mesa level. He then calculates then the probabilities at the mesa level using the results from the simulation of the cuadernos. His conclusion from this is that the number of observed SI vote coincidences should be 1 in 10,000, which I think even Prof. Taylor would consider we are getting into the realm of the “overwhelming evidence of fraud”
Second note added: Prof. Taylor has now removed the word overwhelming from his conclusions. I wish that rather than worry about words, they would work on the real problem at the machine level.