DU doesn't understand polling

[Master of Ossus]

I've been watching DemocraticUnderground.com as they try and explain how Bush cheated to steal the election. It is horrifying, yet hilarious at the same time.

Essentially, a poster named "The Truth is All" has been HUGELY misleading people there, if not outright lying, and virtually EVERYONE BELIEVES HIM.

+http://www.democraticunderground.com/di ... 7461&page=

His argument holds that in states without a paper trail tied to their voting machines, Bush earned a huge boost during the actual vote tallying from the exit polls, and that this constitutes evidence that Bush rigged the election.

There are SO many problems with his "analysis" that a detailed rebuttal would be almost impossible, and would take more than my limited time can handle. However, lest anyone be fooled into believing morons who perpetuate his arguments (and many of the posters there seem to have uncritically accepted his work), here are a few of the basic flaws.

1. He cannot, or refuses to, explain how he determines which states had a paper trail and which did not. He, in fact, admitted after posting that "every single state" without a paper trail registered a boost for Bush that he did not know if Ohio used electronic voting or not. When questioned on how he made this determination, he has not posted a reply.
2. Many of the states he includes in the "no trail" category in fact DID have paper trails in most of their counties. Apparently he failed to realize that only two states have completely switched over to electronic voting.
3. His interpretation of the Margin of Error is laughable. From his posts, he seems to believe that it is based on how accurately they determine the sample's opinion. This is laughable, since the number ASSUMES that the sample faithfully, honestly, and accurately reported their vote. Instead, the Margin of Error is based on the SAMPLE SIZE, and is intended to measure how accurately the sample is likely to be to the entire voting population.
4. Exit polling is not particularly accurate, even though he seems to think that it is. You can easily convince yourself of this fact simply by looking at the 2004 data, which showed Kerry up by more than 15% in New HAMPSHIRE. Why isn't exit polling good? It costs so much money to send people to poll voters after they cast their votes that most samples have enormous biases, since pollsters usually pick only a few precincts per state and stay there only for a small fraction of the time that the polls are open. This means that the samples they use often differ dramatically from the rest of the state, and introduces systematic error into the polls. In particular, geographic areas are under-represented.
5. The pollsters base their determination of which precincts to poll based on what they expect voters to do (ie. where they expect people to vote), but in this election so many unexpected people showed up to the polls that their predictions were often wrong.
6. Although he dismisses the possibility, people can and do lie to exit pollsters.
7. He assumes that if something is outside of the Margin of Error then the vote tally must be wrong. There are actually a couple of problems associated with this. First of all, the Margin of Error is a statistical measure which attempts to determine how LIKELY a sample is to capture the true population mean. Although there is only a 5% chance that a truly non-biased sample will be outside the Margin of Error, because there are fifty states this is actually expected to occur between 0 and 5 times, and it could potentially happen everywhere! Second of all, why in the world must the exit pollsters be right and the actual vote be wrong? Think about this for a minute. It's possible that all of those states used fraud to let Bush win voters. It's possible, but what's more likely? That a sample of a few thousand voters wasn't reflective of the population, or that there was a deliberate effort to skew the election results?

The Truth is All obviously has no clue what he's doing with statistics, analysis, or polling. His refusal to answer to challenges on his methodology is particularly poignant. Any dumbass can put some numbers into Excel and ask it to make a pretty bar graph for them, but statisticians actually have to figure out what the hell the numbers are going to tell them, and then what the numbers say. That's why they get paid to do what they do. Apparently, The Truth is All doesn't understand this and thinks that his graphs support his inane contentions.[/list]

[aerius]

Democratic Underground is taking the Kerry loss about as well as Comical Ali took the Iraqi loss. There's striking similarities between the 2.

[Master of Ossus]

Democratic Underground is taking the Kerry loss about as well as Comical Ali took the Iraqi loss. There's striking similarities between the 2.

I also liked how some of them felt that Michael Moore had betrayed them by speaking of the election as if the Democratic Party had been beaten fair and square.