Odds of Bush gaining by 4 percent in all exit polling states 1 in 50,000

(page takes awhile to load)
http://www.bluelemur.com/index.php?p=405
By John Byrne | RAW STORY Editor

A statistical analysis of exit polling conducted for RAW STORY by a former MIT mathematics professor has found the odds of Bush making an average gain of 4.15 percent among all 16 states included in the media’s 4 p.m. exit polling is 1 in 50,000, or .002 percent.

The analysis, conducted by former Associate Professor of Mathematics David Anick, also ruled out any significance of a variance between electronic voting and paper ballot states, which RAW STORY reported last week.

In fact, the non-electronic voting states of New York and New Hampshire had higher gains for President Bush than states in the exit polls using some electronic voting: Arkansas, Colorado, Florida, Iowa, Ohio, New Jersey, New Mexico, Nevada and West Virginia.

In part, the discrepancy in the site’s earlier reporting came from the fact that New Hampshire and New York were not included in the reported 6 p.m. polling. Both states had the greatest “Bush gains,” by 8.7 and 8.6 percent respectively, and both use non-electronic balloting. The full breakdown of states by electronic and paper balloting can be found here.
<more>

Should be in the million range

I am not patting myself on the back.

The calculation is TRIVIAL.

All you need is :

1- Exit Polling data percentage split.
2- Assumed Margin of Error
3- Excel Normal Distribution function.

That's it. I setup the calculation in minutes.

I am not a Math professor.

But I do have three degrees in Math.

Exit Polling Reported Vote

State Kerry Bush Diff. Kerry Bush Diff. Bush Gain
AR 45 54 -9 45 54 -9.8 0.8
CO 49 50 -1 47 52 -5.6 4.6
FL 51 49 2 47 52 -5.0 7.0
IA 50 49 1 49 50 -0.9 1.9
MI 52 46 6 51 48 3.4 2.6
MN 52 46 6 51 48 3.5 2.5
MO 47 52 -5 46 53 -7.3 2.3
NH 54 44 10 50 49 1.4 8.6
NJ 54 44 10 53 46 6.2 3.8
NM 50 48 2 49 50 -1.1 3.1
NV 49 48 1 48 50 -2.6 3.6
NY 62 36 26 58 40 17.3 8.7
OH 51 49 2 49 51 -2.5 4.5
PA 53 46 7 51 49 2.2 4.8
WI 51 48 3 50 49 0.4 2.6
WV 45 54 -9 43 56 -13.0 4.0

That analysis depends on the validity of the poll results. A poll which does not take a true random sample, in this case the 4pm results, is inherently flawed as it excludes all voters after 4pm.

It cannot be reasonably argued that the polls at 4pm represented a true sample of the electorate. Large groups vote after work--evidenced ohio poll lines late in the day--and they likely represent working men in such a blue collar state.

So the statistical math is easy IF used for results that are truely random.

Welcome to DU!! :toast:

check his website today

that when I drive to the intersection of Garfield and Neely, I only have a 25 % chance of turning to the east.

However, a statistical analysis shows that when I actually reach that intersection, I really do turn to the east upwards of 98 % of the time.

The statistical chances of that happening are one in a billion.

But the fact is I live to the east of the intersection.

You are confusing chance and sampling. If you rolled a 4 sided die to choose the direction you turned (or u-turned) then you would have a 25% chance of turning east no matter your original direction of travel.

However, you are making a concious choice, just like voting. For someone to create a statistical model of your turning they could do one of two things: track every turn you have made and ever will make at that intersection or randomly choose a few of them, sample them.

Now it has been a while so I may make some errors here:

Now the first task in sampling is to choose the number of samples. The greater the sample number the less the uncertainty (margin of error). The next task is to take the samples, they have to be truely random. Failure to do so will introduce errors far outside the margin of error and possibly skew the results toward an untrue conclusion.

If taken only in the morning they likely introduce the error of you going to work and turning south toward work much more frequently. If taken only during the evening they introduce the error of you turning east much more frequently. If taken on weekends they introduce the error of you going to the bookstore to the west. I am sure you get the idea.

So a true random sample with a large sample size will create a curve likely a gaussian curve with two peaks representing going to work and coming from work that very closely approximates the entire set of your turn choices.

They've been heralding this "country goes red" ever since the election. What are the odds of this and how does this compare with ANY other election. Sounds to bias in one direction to me.

Similar to the electoral college maps, the county map distorts a view of the population. Every state has a few counties (usually around a major metropolitan area) which have the majority of the population. The remaining counties are sparsely populated.

So the Red county map is even more misleading than the electoral college map. Look at California, only a few blue counties easily carried that state for Kerry. But that state is awash in red, due to rural counties that went to Bush.

:hi:

Would like for you to do a statistical analysis in cooperation with TruthIsAll, if possible, using random exit poll times, if you can find them, for probability of bush gaining on Kerry in states without a paper trail, then compare same to states with a paper trail. Any deviances noted would be appreciated.

Of course there seems to be no deviance between exit polls in states with a paper trail, from what I have read.

I think the 5% is highly relevant, it is precisely what bush needed to "win" and also declare a mandate.

It's late for me. I hope to see more of your posts. I don't know what you're plumbing, but it strikes me as a cryptic handle.

:thumbsup:

No time for that, I have a paper to publish and a thesis to write. Just contributing my 2 cents.

the only paper I'll see will be backing up out of the toilet I have to snake.

I would have written a novel by now if it weren't for time spent on DU. :)

Stop in any time and contribute. I like your writing style.

read paper ballots were not more likely to have exit poll variance than other states. Because New York, which uses lever balloting, had such a large variance, the optical scan variance is within the threshold of being statistically explained by chance."

Basically - the analysis by this professor found that it does not matter what system of counting was used, the discrepancies were the same. So - for there to be voter fraud, it has to have happened in every single voting system in those states.

Undoubtedly the sample used in NY (not a swing state) was probably a good deal smaller than the samples collected in swing states so a larger variance is *expected*.

Bottom line: all the exit poll data & a comprehensive study is needed.