(Huge) delegate vote anomaly in Alabama verified

[parocks]

1) Wrong vote ALL
2) Wrong vote PART
3) Right vote (ALL and PART)
4) No vote.

It's not tricky.

Is that the BEST you can do? IF this were the result of "vote all", ALL candidates would have overvotes, not just ROMNEY. Come on Parocks, WHY does Romney have 200- 300 more candidate votes than delegate votes in a precinct while the others' numbers are very close? Paul doesn't have ONE precinct where he receives more votes than delegates in the largest precincts representing 50% of the vote in Alabama!!!! Look:

 

[The Man]

I am trying hard to detect that switch, as your charts are very intriguing, but I struggle.
Below is a scatter plot of the difference in votes between the presidential preference and the 3rd delegate race, Paul vs Santorum, district by district for the whole of Alabama. You do see very clearly that Santorum gains and Paul loses, but do they correlate?
A switch from one to the other would tend to populate the chart diagonally. A simple software bug on name allocation would be very detectable here. Is it happening? Looks more like when Paul loses presidential votes, Santorum sometimes gains, sometimes he does not. You do get some (-x, +x) dots, but many (-x, 0) and (0, +x) as well. Now the important thing is that, when you cumulate (-x,0) and (0,+x) data points, it will average onto the diagonal and you will get an impeccable appearance of vote flip on cumulative charts. As things stand, I tend to think that both processes are fairly independent at district level. Only sometimes they coincided. If fraud, the data to me feels more like presidential vote suppression for Paul and/or ballot stuffing for Santorum. Well, independent or... artfully randomized... Shall we ever know?...

Hey Liberty1789, I have previously demonstrated a simple method to greatly reduce noise from voter error (delegates). Since we know that the number of votes AND the number of delegates (position 3) should ideally be equal in every precinct, we multiply each candidate's # of delegate votes X ratio of total votes/total delegates for each precinct. NOW you begin to see the REAL votes exchange between Santorum and Paul:

Delegates (Adjusted)= reported delegates (each precinct) X (total votes all candidates each precinct/total delegates (pos 3) all candidates)

 

[drummergirl]

As things stand, I tend to think that both processes are fairly independent at district level. Only sometimes they coincided. If fraud, the data to me feels more like presidential vote suppression for Paul and/or ballot stuffing for Santorum. Well, independent or... artfully randomized... Shall we ever know?...

I'm not sure if we'll ever have the complete answer on Alabama, but it sure is intriguing. My brain has not had a real life puzzle this sophisticated (games not included) in quite awhile. What surprises me is that with all the data available and the curious nature of the results, why are there no PhD statisticians drooling over this stuff? You'd think they'd be living for an opportunity like this. Anyway...

I suspect that there's a good reason your scatter plots don't show a good correlation and the PPE vote - adjusted delegates graphs The Man did do correlate. If there were two manipulative processes running at the same time, one a siphon from Paul to Santorum and the other flipping from Gingrich and Santorum to Romney a lot of funny numbers suddenly make sense.

The trouble I'm having with my hypothesis is I can't think of a way to test it. I keep running into the too many variables and not enough equations type problem. Any ideas?

 

[Liberty1789]

Delegates (Adjusted)= reported delegates (each precinct) X (total votes all candidates each precinct/total delegates (pos 3) all candidates)

Each data point in the scatter chart is made of X and Y which are both themselves the result of a substraction. Your adjustment multiplies one of the 2 elements of the substraction by an identical factor, (total votes all candidates each precinct/total delegates (pos 3) all candidates), itself not independent of the values of X and Y. So the probability of introducing a correlation artifact is really far from trivial.

To give you an idea of the complexity of it all, I have created below a scatter chart with 4 independent random variable (on the left), plotting V[SUB]1[/SUB]-V[SUB]2[/SUB] vs V[SUB]3[/SUB]-V[SUB]4[/SUB]. Then I apply a transformation along the lines of your adjustment on the right. See what I mean?

 

[drummergirl]

yep; this so feels like a 3 body problem.

 

[LibertyIn08]

I'm not sure if we'll ever have the complete answer on Alabama, but it sure is intriguing. My brain has not had a real life puzzle this sophisticated (games not included) in quite awhile. What surprises me is that with all the data available and the curious nature of the results, why are there no PhD statisticians drooling over this stuff? You'd think they'd be living for an opportunity like this. Anyway...

I suspect that there's a good reason your scatter plots don't show a good correlation and the PPE vote - adjusted delegates graphs The Man did do correlate. If there were two manipulative processes running at the same time, one a siphon from Paul to Santorum and the other flipping from Gingrich and Santorum to Romney a lot of funny numbers suddenly make sense.

The trouble I'm having with my hypothesis is I can't think of a way to test it. I keep running into the too many variables and not enough equations type problem. Any ideas?

There are plenty of graduate level statisticians on this forum. There's a reason most don't touch these nutty threads with a ten foot pole.

 

[RonRules]

There are plenty of graduate level statisticians on this forum. There's a reason most don't touch these nutty threads with a ten foot pole.

The chairman of the Statistics Department at my local university disagrees with you. He's taught thousands of graduate level statisticians.

Jus' sayin.

 

[drummergirl]

There are plenty of graduate level statisticians on this forum. There's a reason most don't touch these nutty threads with a ten foot pole.

Well, they may not want to take their time to deal with noob questions and trolls. But if there are statisticians avoiding these things for that reason, they are doing nothing to improve public opinion of statistics. Most people hold to "there are lies, damn lies, and statistics." As long as the statisticians stay in the ivory towers, that view will remain.

 

[The Man]

Each data point in the scatter chart is made of X and Y which are both themselves the result of a substraction. Your adjustment multiplies one of the 2 elements of the substraction by an identical factor, (total votes all candidates each precinct/total delegates (pos 3) all candidates), itself not independent of the values of X and Y. So the probability of introducing a correlation artifact is really far from trivial.

I respect what you are saying Liberty1789. But clearly there is heavy voter error regarding delegates which tends to mask any inverse relationship between two candidates' vote-delegate differentials. What you term "identical factor" may hold TECHNICAL merit AND it is true that the "factor" relies partially on the number of that particular candidate's reported delegate votes. But we're in unchartered mathematical waters that require "outside-the-box" thinking. So I decided to create the same relationship "votes minus adjusted delegates" for all possible relationships in this Alabama GOP Race. Are you telling me this is just a coincidence? NO other relationshipe shows ANY kind of inverse correlation except Paul vs Santorum. As a reference, the red dotted line represents the perfect ideal inverse relationship (Candidate A's loss = Candidate B's gain).

RED LINE = PERFECT INVERSE RELATIONSHIP

 

[LibertyIn08]

The chairman of the Statistics Department at my local university disagrees with you. He's taught thousands of graduate level statisticians.

Jus' sayin.

If he is that confident of your process, he would be able to generate at least one solid scholarly article (likely more) and would probably be in the running for multiple awards. Professors don't turn down those kinds of opportunities. Since he hasn't come forward and given his actual name, I guess I'll have to take your word on the above anecdote.

I look forward to reading his published article proving fraud in the 2012 election.

As for the ivory tower comment posited by drummergirl, since many basic principles of observational sciences and statistics are being ignored here and in the other thread, the qualified persons probably see no reason to engage you.

That being said, I'll leave the critiques for the criticism thread and make sure this is my last post on the matter on either of the flipping threads. The only reason I came to this thread is the nonsensical posting of "vote flipping 'proof'" in unrelated threads; given the preponderance of qualified individuals on this forum who don't find your "proof" reliable, I suggest you leave the discussion to these threads and stop crossposting it elsewhere claiming it proven.

 

[drummergirl]

given the preponderance of qualified individuals on this forum who don't find your "proof" reliable, I suggest you leave the discussion to these threads and stop crossposting it elsewhere claiming it proven.

Who?

I have yet to hear from anyone even claiming to be a professional statistician. I could be wrong, because there are thousands of posts, but I don't recall any posts from a credentialed statistician on this topic (I think it would've registered as important).

Your statement leaves me feeling a bit like Indiana Jones at the end of Raiders of the Lost Ark when told, "we have top men working on it."

Who?

I know if you read my summary, it's clear that I'm not a statistician. But I do know basic statistics well enough to have used that skill set many times for various reasons. My graduate work involved quite a bit of statistics (and yes, it's published in a professional, peer reviewed journal in my field) I don't think anyone is claiming that a series of forum posts is the same as a peer reviewed journal article. I don't think professors would say that a discussion over a few beers is the same as a symposium either. However, I know several professors that actually got some of their best research ideas from the discussions with beer.

So, if you know any hardcore statisticians avoiding this thread, I toss down the gauntlet. Look at the data and analyze it or be known as an intellectual lily-livered, yellow bellied chicken.

 

[The Man]

When the precinct totals are averaged lot size= 20 you get this inverse correlation between Paul and Santorum. Liberty1789, if this doesn't at least earn a second look from you, then it's time to move on to the next analysis. Honestly, this graph shows absolutely that Santorum receives Paul's votes. As I have maintained, you can't look at individual precincts unless you filter noise. Then, the obvious emerges. We could go through individually and do some obvious filtering each precinct one-by-one and I will guess that this graph would look like a 45% straight line. For example, the obvious ballot stuffing in Jefferson County causes my method of filtering to err; Romney's ridiculous gift of 100's of votes causes my method of adjustment to subtract votes from the other candidates. The graph below is in precincts less than 690 votes, which corresponds to when Romney starts receiving votes out of thin air.

 

[The Man]

Same Graph for ALL precincts

 

[RonRules]

The clustering on this chart is interesting to me. Pure randomness can have clusters, but more often than not, it's more like white noise on an old TV. How can we isolate the cause of this clustering?

 

[The Man]

The clustering on this chart is interesting to me. Pure randomness can have clusters, but more often than not, it's more like white noise on an old TV. How can we isolate the cause of this clustering?

The point is that plotting candidate A vs candidate B "votes minus delegates' should produce a line with a -1 slope (y=-x) in the ideal 1:1 vote swap. Liberty1789 brilliantly introduced this as a new method to deduce that a candidate's votes transfer to another. His initial analysis of the Paul-Santorum relationship showed no such relationship. I have filtered the noise and averaged the results. You can see that no other relationships in this race produce anywhere near these results. My method of filtering is less than perfect- but good enough to see this parasitic relationship very clearly.

 

[The Man]

If you don't like my delegate adjustment method, here is Paul vs Santorum votes minus delegates (reported 3rd position) each point averaged (sample=20). The red dotted is the line y=-x, the ideal perfect parasitic vote swapping relationship.

 

[drummergirl]

If you don't like my delegate adjustment method, here is Paul vs Santorum votes minus delegates (reported 3rd position) each point averaged (sample=20). The red dotted is the line y=-x, the ideal perfect parasitic vote swapping relationship.

Even unadjusted, that's a pretty strong correlation.

 

[The Man]

Even unadjusted, that's a pretty strong correlation.

I guess we tend to lose level perspective of this Alabama delegate situation. Literally there is nowhere in the 1864 precincts where Paul does anything but lose votes versus delegate votes. In the only precincts where Paul has more votes than delegates (44 of them using position 3 delegate), he picks up a whopping 116 votes total. In thes same 44 precincts, Santorum and Romney gain more than 600 votes a piece versus delegates. Romney gains more than 200 votes in at least 20 precincts while Santorum gains 100+ votes in 46 precincts. Paul loses more than 100 votes in 60 precincts and his largest gain is 13 votes in a precinct, where he still gains less than 2 of the other 3 candidates. Nowhere in this thread is there any credible explanation of this. I'm not sure how it's possible to look at these latest graphs and not acknowledge the correlation. BUT you have to accept the fact that delegate data is only valuable when looking at groups of precincts.

 

[RonRules]

We're at the analysis paralysis stage now. We've got the evidence. It's up to the State to determine the actual cause.

Has anybody called the Secretary of State yet?

I was on the phone with Milwaukee today. I kinda adopted Wisconsin and don't really want to cross lines with Alabama.

 

[Liberty1789]

When the precinct totals are averaged lot size= 20 you get this inverse correlation between Paul and Santorum. Liberty1789, if this doesn't at least earn a second look from you

How do you group the 20 precincts together? Are they sorted?

The reason I am asking is that when you look at Paul's share of votes vs Santorum's per county, which should cancel out quite a bit of noise, there is no sign of correlation whatsoever. None. The 5% "skimmer" is not apparent here either, as plenty of counties score above 5% and Paul's best score is almost all the way up to 8%.

Some more head scratching at my end, I'm afraid.