If you are interested, here is the first scientific analysis of Florida voting machine irregularities submitted by a team at UC Berkeley:
The link is from electoral-vote.com, a very good site even though I don’t share the webmaster’s politics. In a nutshell, they assert that Bush got 130,000 extra votes in a manner that cannot be explained statistically. Although they do not assert fraud, they strongly suggest it.
Being a stathead myself due to my baseball obsession, I found the study misleading because there is no adjustment for the undulating voting population variables (the unspoken assumption is that the very same group that voted in 2000 and 1996 voted in 2004, which is a fallacy) which transcend raw numbers. How do you normalize for a better job the republicans did of getting out their base than the democrats? I know of a baseball statistician who can prove that Dave Kingman was a better slugger than Babe Ruth. He selects narrow hitting statistics and puts them in a formula which exploits Kingman’s one strength and ignores the plurality of Ruth’s superior output. It can be done in any study of numbers: baseball, elections, economics, you name it.
Narrow Selection of Data + Carefully Crafted Formulae/ Poor Normalization for Variables= Skewed Result.
Their adjustment for turnout is fallacious. They assert change in voter turnout does not explain the alleged anomalies in their study summary but there is no explanation of this assertion in the data. Even if what they say were taken as canon gospel and one concluded that fraud occurred, it would not have affected the final result in Florida, where Bush won by over 375,000 votes (The study concluded that the anamoly was 130,000 votes in W’s favor). I wonder if the same study were adapted to the 1972 election what the findings would yield. It would seem to me that if you apply a statistical analysis to a narrow set of categories and have no element to normalize for the dozens of other fuzzy factors affecting the outcome, you can prove anything you wish. I would hasten to add that to think that a group of Berkeley sociology PhD. candidates does not have a partisan presdisposition is either naïve or intellectually dishonest.
A little knowledge is a dangerous thing. Most people who read the study or the resulting discussions couldn’t tell you the difference between standard deviation and margin of error to save their first-born’s life, so what little they know applied to their internal logic cicuitry will yield whatever their ideological view wants to conclude. For example: A recent letter decrying Bush to The Journal News here in Westchester County, NY spoke of how the election results “were within the margin of error.” The message was that Bush had no mandate borne of an electoral majority. The poor slob doesn’t understand that a margin of error only applies to statistical samples which are a tiny fraction of a large group, not the actual result after the whole group has spoken. The between the lines message is that the election result was just an unfortunate mistake. So was the Red Sox Victory over the Yankees in the 2004 playoffs. I’d love a recount there, but that’s baseball.
In the interest of balance, I googled the Berkeley study.
Naturally, the study itself came up first. These are the next two results, which more or less dismiss the finding and link to other sites which do:
Nothing about either site made me think it was particularly partisan vis a vis dailykos.com.
Keith Olberman, whom I dislike going back to his ESPN days, called the study out as well. In this case, I give Olberman more consideration because of his sports pedigree. As I said above, baseball fans know how stats can be manipulated to mislead. Although Olberman wasn’t quite hip to all the mathematical spaghetti in the study, he rightly pointed out that the data were too narrow in scope. He also pointed out that the same analysis in Ohio found “absolutely nothing” that would raise an eyebrow. Of course, Ohio was where the rubber his the road in this election. If you could prove systemic fraud there, it would affect the election. But since it didn’t fit the agenda for this study, if you can’t beat them, smear them.
Here’s my solution: George Soros, Larry Flynt or someone else with money to burn offer $15 million for concrete proof of hacking or other election stealing nefarious deeds. Once they weed out the gold diggers, they might find a Deep Throat. My opinion is that they won’t, but they will get plenty of Linda Lovelaces.