Flipping the vote against Ron Paul in South Carolina?

[bbwarfield]

I still think fraud took place.. honest... i havent jumped ship on this.

Im from a background that says "look at everything.... narrow the focus.... avoid information that people will dismiss for any reason... avoid information that has a plausible but not probable other reason... avoid information gathered from people who have a clear bias unless you state that bias.... finally you have information no one can disagree with and if the other information you already were not looking at gets attacked, those attacks wont stick to your information"

Iowa rigged? thousands of ways to do it... probably was done. and your graphs would show these as well

South Carolina is the place to focus cause it is not only using electronic machines but machines that were sold after they were used in fraud in FLorida....not just the same type... the same exact machines .....

lets keep up analazying everything....everything.... but make sure we weigh South Carolina heavier cause of the different natures of the voting systems and reporting systems. Iowa shows good graphs.... but there is "possible" but not "probable" reasons for that.

 

[Liberty1789]

Oh my God. Same systematic several-standard-deviation jumps in Romney's score in NH.

 

[bbwarfield]

In the end, there will be zero question that Mitt Romney's vote receiving percentage in larger precincts of "significant" counties perfectly correlates with only one factor, and that factor is the total number of votes at each the precinct. There simply is no practical demographic combination that can arguably correlate closely enough with the number of raw precinct votes to explain this.
Definition of "significant county" by Liberty1789's standard: Within the vote flipping threshold.

I agree wholeheartedly

 

[Mark37snj]

I still think fraud took place.. honest... i havent jumped ship on this. I'm from a background that says "look at everything.... narrow the focus.... avoid information that people will dismiss for any reason... avoid information that has a plausible but not probable other reason... avoid information gathered from people who have a clear bias unless you state that bias.... finally you have information no one can disagree with and if the other information you already were not looking at gets attacked, those attacks wont stick to your information"

This is Statistical Mathmatical Warfare. You have admittedly stated that Statistics/Mathmatics is not in your area of expertise, your Background. The people who have the Statistical Mathmatical Background are saying Fraud, they are saying get it out, they are saying time to kick this up a notch. In a Statistical War I'm listening to the Statistical Generals, thats not you and you should be listening to them as well unless you have some other motive.

 

[Mark37snj]

bbwarfield this may simply come down to what you think we should be doing and what we should really be doing. The points you make are the things you do while your trying to write a story, make an argument, set a narrative. You present some data here and overlook some data there so they story flows nicely. We are not writing a story here, were crunching numbers, ALL NUMBERS. We are not leaving anything out because it does not fit well into the story we are trying to tell. We are not trying to lead people to a certain conclusion, were not trying to manipulate them. Our job is to discover the truth and get it out there regardless of whether it shows fraud in one precinct and no fraud in another. Get all the information out and let others draw their own conclusions. Let other experts draw their own conclusion, they don't need a story form us, they don't need an argument crafted for them, the data speaks for itself.

 

[chapchap70]

Oh my God. Same systematic several-standard-deviation jumps in Romney's score in NH.

Is there a strategic reason why RP was left out of this data?


Edit: I assume that the 10 percent Romney had added at the 90 to 100 percentile large precincts (compared with the smallest ones) came from Paul but not Huntsman and the others?

 

[Liberty1789]

How to detect the flip... R2 for most of us, F/t-stat for the math divas.

EDIT: Calculation started from 50% of cumulative votes cast

 

[Liberty1789]

Is there a strategic reason why RP was left out of this data?

Good question. Romney collects all flipped votes so the stats are the most spectacular there. Gingrich/Santorum are more or less untouched, so you get maximum contrast. Paul, Perry, Bachmann and Huntsman are feeders, but they share the loss of what Romney gains between them and the stats are less "screaming".

I must also confess that the table is really fastidious to assemble, so I abridged it...

 

[chapchap70]

Good question. Romney collects all flipped votes so the stats are the most spectacular there. Gingrich/Santorum are more or less untouched, so you get maximum contrast. Paul, Perry, Bachmann and Huntsman are feeders, but they share the loss of what Romney gains between them and the stats are less "screaming".

I must also confess that the table is really fastidious to assemble, so I abridged it...

I sent the image to a guy at our church who has an accounting background to see if it will peak his interest. He works for a large computer software company on Long Island. I'm not sure if it is still Computer Associates or if they got bought out and changed their name. Can't keep track of all the mergers these days.

 

[jjockers]

How to detect the flip... R2 for most of us, F/t-stat for the math divas.

Not sure what you mean here. As someone who might qualify as a math diva, I wouldn't look at the R^2 (goodness of the linear fit) or the F/T-stat first. Unless I am missing something, I'd first look at the slope (followed by the R^2). Negative slope with high R^2 => losing votes as precinct size increases, Positive slope w/ high R^2 => gaining votes as precinct size increases.

Even then, a positive slope does not necessarily imply vote flipping. It just means the candidate, for a reason yet to be determined, earned a higher percentage of votes as precinct size increased.

Also, some of the values presented are misleading. Your low R^2 values are low because of early (<=5% on x-axis) fluctuation (smallest precincts). See Santorum in the bottom graph.

 

[bbwarfield]

http://www.electionstudies.org/nesguide/gd-index.htm#9

Mark37nj if youve read all these articles (wich i have) then you can criticize me for pointing out that yelling fraud cause a graph looks a way you dont suspect is not a good idea. Many of these articles will help you understand why I am not a fan of looking at Iowa is actual legitimate differences in the electorate and demographics.

Im fine with looking at the numbers.... I just want people to be clear when they are talking about two different states and for that matter two different counties within a state.

Or we go with your plan and apply it across the board

Lets call the whole election cause smaller states have voted and chosen and they are statistically significant enough to decide the entire election.Iowa, New Hampshire, Maine and Colorado spoke.... when added together Romney came out on top. This is what your graph would show isnt it? these four states make a statistically significant sample on wich we can decide Mitt Romney should be the nominee.....and to prove the point the large states of Nevada and Florida prooved the numbers as well.... south carolina he still came in second but thats okay cause Newt is an outlier not a statistically important candidate.......... I will spend all day making a graph that shows you this if you like? Cause there is no importance of demographics... its just the numbers like you said.... its JUST THE NUMBERS.

Turn off sarcasm............................................

Elections are decided by demographics. If you get 99% of the black vote (which obama did) then you won that demographic. So a precinct that is 92% black voters gives him roughly 90% of the vote... but how can this be cause the precinct next to it he gets 5%! why? no amount of numbers will tell you but by law we have to have MINORTIY MAJORITY districting in states were the DOJ thinks there is a significant enough chance of the minority not being represented. So the precinct line is actually DRAWN to exlude white votes from that precinct.

If you dont know how elections work..... the numbers would never tell you that and you would look at greenville counties 2008 presidential totals and screamed fraud all day.

 

[Liberty1789]

Your low R^2 values are low because of early (<=5% on x-axis) fluctuation (smallest precincts). See Santorum in the bottom graph.

Sorry, my bad, forgot to mention that the numbers are for the last 50% of the votes only

 

[Mark37snj]

http://www.electionstudies.org/nesguide/gd-index.htm#9

Mark37nj if youve read all these articles (wich i have) then you can criticize me for pointing out that yelling fraud cause a graph looks a way you dont suspect is not a good idea. Many of these articles will help you understand why I am not a fan of looking at Iowa is actual legitimate differences in the electorate and demographics.

Im fine with looking at the numbers.... I just want people to be clear when they are talking about two different states and for that matter two different counties within a state.

Or we go with your plan and apply it across the board

Lets call the whole election cause smaller states have voted and chosen and they are statistically significant enough to decide the entire election.Iowa, New Hampshire, Maine and Colorado spoke.... when added together Romney came out on top. This is what your graph would show isnt it? these four states make a statistically significant sample on wich we can decide Mitt Romney should be the nominee.....and to prove the point the large states of Nevada and Florida prooved the numbers as well.... south carolina he still came in second but thats okay cause Newt is an outlier not a statistically important candidate.......... I will spend all day making a graph that shows you this if you like? Cause there is no importance of demographics... its just the numbers like you said.... its JUST THE NUMBERS.

Turn off sarcasm............................................

Elections are decided by demographics. If you get 99% of the black vote (which obama did) then you won that demographic. So a precinct that is 92% black voters gives him roughly 90% of the vote... but how can this be cause the precinct next to it he gets 5%! why? no amount of numbers will tell you but by law we have to have MINORTIY MAJORITY districting in states were the DOJ thinks there is a significant enough chance of the minority not being represented. So the precinct line is actually DRAWN to exlude white votes from that precinct.

If you dont know how elections work..... the numbers would never tell you that and you would look at greenville counties 2008 presidential totals and screamed fraud all day.

All you did here was ignore facts, make a bunch of assumptions, give a bunch of opinions, and create a Straw Man arugument for you to slay. BORING.

 

[bbwarfield]

All you did here was ignore facts, make a bunch of assumptions, give a bunch of opinions, and create a Straw Man arugument for you to slay. BORING.

um... i think thats what you did
yes i had some poor rehtorical style.... but youve not proven me wrong on anypoint... just called them all unimportant... or non researched or whatever. when you supply evidence with quotes from the forum you including quotes of you own assumptions? thats strange to say the least

 

[Liberty1789]

I wouldn't look at the R^2. I'd first look at the slope (followed by the R^2).

I see where you are coming from and I agree with that.

Hear my central case: at mid-way, 50% of votes in, hundreds of precincts counted, I should have a pretty good estimate of the final score at 100% and dozens and dozens of counties graphed here confirm that. Let's take it as my working assumption. There is no reason for the candidates' lines to differ in volatility and t stat other that randomly.

Now a vote flipper kicks in and transfers an amount of votes from one guy to the other, proportionally to the cumulative % of votes cast. My original flatline becomes a slope, yes, as you say, but very importantly, its value is now the sum of a natural line and a mathematical straight line. The mathematical impact of that is that slope will be much "straighter". The standard deviation around the linear fit of the slope segment will collapse. Straighter segment equates a higher R2 in the slope segment than in flat liners. And it means a massive change in F factor and T stat. A true mathematical fingerprint! Because of the inclusion of of a zero-volatility linear component, F factors/t-stat jump 5 to 30 times vs the untampered line! It is what you see in Palm Beach/Darlington/Washoe above.

 

[Mark37snj]

um... i think thats what you did
yes i had some poor rehtorical style.... but youve not proven me wrong on anypoint... just called them all unimportant... or non researched or whatever. when you supply evidence with quotes from the forum you including quotes of you own assumptions? thats strange to say the least

All your doing is Muddying the Waters. Lets take a look at all the crap you just tried to throw into this discussion. This is from one small paragraph in your last response.

demographics....99% of the black vote.....obama......demographic.....92% black voters.....90% of the vote.....5%.....law.....MINORTIY MAJORITY....districting.....DOJ.....minority.........precinct line.....exlude white votes

ALL this was in just 4 lines, one small paragraph.

 

[bbwarfield]

Those are all true.... what was your point there?
refering to South Carolina: I know the number was 99% in South Carolina but okay.... well go with 95% nationwide

http://www.dailymail.co.uk/news/art...als-black-voters-swept-Obama-White-House.html

The following article shows 92% black population in a precinct being questioned earlier

http://www.greenvilleonline.com/art...50/Minorities-largely-sit-out-Republican-race

The Act established extensive federal oversight of elections administration, providing that states with a history of discriminatory voting practices (so-called "covered jurisdictions") could not implement any change affecting voting without first obtaining the approval of the Department of Justice, a process known as preclearance

this is from wikipedia on the voting rights act (ive read the whole law... but i figured a summary would be easier)

and majority minority districting discussed in an article on the law

http://writ.news.findlaw.com/commentary/20041007_hayden.html


If you want to prove me wrong... read all the pertinent information that shows what you said was oversimplification was true

 

[bbwarfield]

All your doing is Muddying the Waters. Lets take a look at all the crap you just tried to throw into this discussion. This is from one small paragraph in your last response.

demographics....99% of the black vote.....obama......demographic.....92% black voters.....90% of the vote.....5%.....law.....MINORTIY MAJORITY....districting.....DOJ.....minority.........precinct line.....exlude white votes

ALL this was in just 4 lines, one small paragraph.

im sorry..... i didnt realize you were just creating a strawman calling this all crap not SIGNIFICANT INFORMATION that changes what you see in numbers from Greenville South Carolina were several such districts are in place.

 

[bbwarfield]

as a side not.... the majority minority districts were taken into account and were not significant in Greenvilles over all analysis but could have served as a debunk if not noted earlier. But it needed to be looked at and it was.

 

[Liberty1789]

I don't have any significant stat training, and yes, it was tricky, but I think your perseverance paid off. Thanks for this explanation, I think I get it now.

In the original Anderson County document, there seems to have been two arguments for fraud: (a) the uniformity of the "swap" -- what The Man was calling the parabola effect, and (b) the fact that Romney's total ended up where Paul's should have been and vice versa. Not being versed in the high level statistical methods, it was easier for me to grasp the idea of projected totals and results not ending up where they should have been. The subsequent discussions about R-squares seemed to me, again as non-stat guy, to be an attempt to reinforce the projections without addressing the reliability of non-random sample selection. It now seems to me that the two arguments are independent of each other. For those that have been saying this all along, my apologies. It's been a busy week and this has been a fast moving and highly technical thread.

If I may summarize in my own words, maybe you can confirm that I get it.

We don't care about projected totals (ends), we care about the nature of the plotted line itself (means). The plot lines can be on different frequencies, but they should appear as white noise. What the R-squares and t-stats are saying is that the signal itself is non-random.

I'm sure this is technically inaccurate, but is this pretty much what's going on?

This is critical and your intuiton is on the right track. I'll put it as simply as I can.

A visually straight line on the charts I publish is rarely perfectly straight. It joins points which are not perfectly aligned. The differences between the real points and the straight line that best fits them can absolutely be described a white noise. t-stat measures directly how stable is the slope of the line joining the points, how much it "vibrates" along its general direction. F factor is even smarter in some ways: it measures if all those points which are not perfectly on the line are normally distributed away from it: more points closer than far away from the line, if you will.

On segments of lines where the vote flipper I suspect is active, something extraordinary happens: the nature of the white noise is totally transformed! T-stat is up 5/10 times, F factor sometimes 20-30 times. This all indicates that the slope is not natural. it has been "smoothed" by the vote flipping. It vibrates much less and points next to the line are much closer than before, with fewer outliers.

What happened is this: you're mixing the original, "natural" line with normal white noise to a perfect straight line with no noise. The combination has therefore massively less noise than the original. t-stats and F factors shoot up : the straighter the line, the higher they go. And they shoot up massively. The radical change in the noise is a proof of tampering! What one calls a mathematical fingerprint!

Another point: I have not seen yet a big F factor (>500) on its own: if the white noise is gone from 1 line, there is always another one where the noise is gone as well. Algorithmic vote flipping would do precisely that.