Reply #49: Response to your ideas

[Time for change]

I was going to say that I believe that my analysis does provide some supporting evidence for vote fraud in general (I'm unclear as to whether or not your post is disputing that), although the main evidence is the exit polls. I would have based that on the fact that in the absence of vote fraud I would expect the discrepancies between the Gore and Kerry margins to be reasonably evenly distributed between the different types of voting machines. Yet, statistical analysis provided some evidence that the distribution is not even (This was somewhat ambiguous: ANOVA analysis was not statistically significant, yet non-parametric tests such as the Wilcoxian rank sum test and chi square Mantel Hantzel test were borderline statistically significant ) This in combination with the voting irregularities found with the Triad counties made me feel that the difference in Gore-Kerry discrepancies between the Triad and other machines offerred some support to the vote fraud theory.

But then I took your advice and did the ANOVA analysis (for all 88 counties) while controlling for county size (i.e., total 2004 vote), and there was a very strong corellation between county size and Gore/Kerry discrepancy (F=6.3, P<.001). Strangely, the corellation between county size and Gore/Kerry discrepancy held up when looking at the 47 non-Triad counties alone, but there was no correlation when looking at the Triad counties alone (maybe because they were too small.)

I'm not sure what all this means. In part, the interpretation of all this depends on whether parametric or non-parametric tests are most appropriate in this circumstance, and I'm not sure what to say about that. Obviously county size is highly related to the Gore/Kerry discrepancy -- but I can't fathom why that would be the case. Why would county size cause the percent of the Democratic vote to change so much between 2000 and 2004?

Anyhow, here are the Gore/Kerry discrepancies for all 88 counties, with the negative values favoring Gore, and the positive values favoring Kerry:
SHELBY -0.1208!
VAN WERT -0.0988!
MERCER -0.0966!
DARKE -0.0853!
LAWRENCE -0.0702!
MIAMI -0.0694!
CLINTON -0.0669!
AUGLAIZE -0.0586!
BELMONT -0.0577!
JACKSON -0.0551!
PAULDING -0.0505!
HARRISON -0.0502!
MARION -0.0466!
HARDIN -0.0416!
CHAMPAIGN -0.0413!
WYANDOT -0.0359!
HIGHLAND -0.0349!
PREBLE -0.0347!
CLERMONT -0.032!
DEFIANCE -0.0319!
MADISON -0.0295!
CLARK -0.027!
LOGAN -0.0265!
ROSS -0.0263!
MORROW -0.025!
JEFFERSON -0.024!
BROWN -0.0239!
COLUMBIANA -0.0228!
BUTLER -0.0225!
GUERNSEY -0.0214!
GREENE -0.0186!
SANDUSKY -0.0185!
UNION -0.0177!
GALLIA -0.0169!
OTTAWA -0.0168!
WARREN -0.0156!
TUSCARAWAS -0.0149!
CRAWFORD -0.0142!
WILLIAMS -0.0141!
PUTNAM -0.0125!
RICHLAND -0.0115!
TRUMBULL -0.0106!
LICKING -0.0105!
MUSKINGUM -0.0101!
HANCOCK -0.0096!
PICKAWAY -0.0072!
MONTGOMERY -0.0038!
FAYETTE -0.003!
MAHONING -0.0024!
PIKE -0.0009!
ADAMS 0.0009!
SENECA 0.0022!
HENRY 0.005!
PERRY 0.0066!
FAIRFIELD 0.0078!
HOLMES 0.0104!
FULTON 0.0108!
ALLEN 0.0121!
LUCAS 0.01255!
PORTAGE 0.01289!
NOBLE 0.0139!
LORAINE 0.01641!
ASHTABULA 0.01732!
ERIE 0.02225!
WOOD 0.0249!
SCIOTO 0.0251!
WASHINGTON 0.0254!
MEDINA 0.0256!
LAKE 0.0263!
ASHLAND 0.0286!
SUMMIT 0.03182!
STARK 0.03357!
HURON 0.0342!
MONROE 0.03532!
KNOX 0.0354!
DELAWARE 0.0359!
COYAHOGA 0.03742!
WAYNE 0.0375!
GEAUGA 0.0376!
VINTON 0.0422!
COSHOKTON 0.0422!
MEIGS 0.0472!
CARROLL 0.0543!
HOCKING 0.0618!
HAMILTON 0.0651!
MORGAN 0.0719!
FRANKLIN 0.07322!
ATHENS 0.1284!

I haven't compared this with the counties implicated by Arnebeck. Do you think that would be useful?

One thing of interest is that Hocking County, which I believe is the infamous county where a Triad official was said to have tried to make sure that the computer and hand counts matched, is the second out of 41 Triad counties most favorable to Kerry, compared to Gore. Could that mean that this whole incident was a red herring, designed to get us to take our eye off of other counties?