It looks like you're using an Ad Blocker.
Please white-list or disable AboveTopSecret.com in your ad-blocking tool.
Thank you.
Some features of ATS will be disabled while you continue to use an ad-blocker.
Originally posted by Xcalibur254
reply to post by pwe11
The reason your analysis falls apart is because you are not using a random sample. The reason the projections and polls you see on the news can be accurate is because they're using data as it comes in. This tends to be fairly random. As a result you can get a more even sampling of any demographics that may be present. Your method only looks at a specific demographic. Those who live in precincts with low voter turnout. These precincts tend to correlate with more rural areas. Ron Paul does well in rural areas, but not as well in urban areas, which are areas that Romney does well. By ignoring these more urban areas you are ignoring a key demographic that votes for Romney. So of course his actual results are going to be better than his projected results and Paul's are going to be worse than his projected.
Originally posted by pwe11
I am the author of the paper being discussed here. I appreciate all of your comments in taking the time to read my work, whether you agree, disagree, like, or believe it's "the worst math ever." Please understand that I have created hundreds of graphs for counties in New Hampshire, Florida, and South Carolina. The challenge for me is getting the information in a form simple enough that most readers can comprehend and brief enough as to not lose the reader's interest. I started out with the intention of detailing all of the algorithms but realized that it's simply too much analysis for most people. I am surely impressed with anyone here that sees how "obviously flawed" my math is. Ironically enough, several statistical analysts with masters degrees feel otherwise. I can guarantee that if you go to the South Carolina Elections Commission Website and Download the data from Anderson County and load into a spreadsheet, arrange the data from smallest vote precinct to largest, and graph the running total votes vs each candidate's running vote total, your graphs will look just like mine. If the "experts" can tell us who has won the SCGOP Primary when less than 1% of the vote has been counted and using less than 0.4% exit poll subjects, the outcome can accurately be projected with 5000 votes cast in a county with 27,000 votes total from 35 precincts, especially if the precincts are representative of the whole county, as I believe they are in this case. If you think you know the demographics in Anderson better than me, let me hear from you. I lived there for 22 years.
I welcome contact from any of you who have talents that can help bring this information to light.
edit on 19-2-2012 by pwe11 because: (no reason given)
The best reason oultined so far why Romney seem to do better in Urban areas, is that the GOP fraudsters only bother to manipulate the bigger caucuses, as outlined in the OP authors post. So what is your counter argument here? That the numbers says different?
Originally posted by Xcalibur254
reply to post by pwe11
The entire methodology is flawed though. You're trying to prove vote manipulation by excluding data based on the assumption of vote manipulation but you can't do that until you reach your conclusion. Pretty much your entire model relies on using the conclusion as a premise, but you can't assign a truth value to the premise without proving it first. The logic is completely circular. Because of this you have ignored data and jumped to a conclusion not entirely supported by the data. Once again your conclusion relies on it being a premise and this has led to a clear cut case of confirmation bias. You refuse to see the more likely conclusions such as the fact that you're only looking at a specific demographic and as a result it cannot predict for other demographics. The simple fact of the matter is that your premises and conclusion are riddled with logical inconsistencies which invalidates your analysis even before you did it.
I have now finished my analysis using the methods laid out in the OP. I only looked at precincts that had less than 250 voters. I entered them in from least amount of voters to most amount of voters. Each entry was cumulative with all previous entries. This would give me a total number of voters from all the precincts, as well as the totals for each candidate from those precincts. I then ran four regressions, one for each candidate. My independent variable was total number of votes and my dependent variable was total number of votes for that candidate. This then gave me an equation for each candidate to use as a predictor as long as I knew the total number of voters. These equations are as follows:
Gingrich = Total(.443) - 9.760
Paul = Total(.214) + 12.104
Romney = Total(.171) + 2.955
Santorum = Total(.172) - 5.299
I then entered in the projected amounts and the actual amounts for the remaining counties. From these I ran a nonparametric test for each candidate comparing the projected values to the actual values. This produced the following results. The was no significant differences between Gingrich's projected and actual votes. There were significant differences between Paul's projected and actual votes. There were significant differences between Romney's projected and actual votes. There were significant differences between Santorum's projected and actual votes.
So, using the model laid out in the OP we are only able to accurately predict the votes for a single candidate in Anderson County. This would indicate that the method used in the OP is not an accurate method in projecting election results and as such his conclusions are invalid. There are a number of reasons that this model didn't work, many of which have already been mentioned. I would say chief among these are the fact that a random sample was not used and the fact a linear model is not appropriate for this data.
If I get the chance I'll come up with a model that corrects these errors and then we can compare that with the one in the OP and see which produces more accurate results.
Originally posted by sheepslayer247
After reading the OP's information and the Thread from SO, it's become pretty clear that there is election manipulation happening here!
So what can we do? We have info, but where can we take it from here? Is there someone or some group that we can contact to take this to the next level?
We have to do something, or we will get exactly what we deserve! The time for complacency is over.....let do something!