Reply #158: No, the numbers are 4101 and .01

[TruthIsAll]

1)4101 is the number of integers, assuming the range given.
You can argue about the size of the range if you wish, but my numbers were based on reasonable precinct vote and registration ranges.

We are calculating the percentage of votes/registrations, not the difference between votes or the difference between registrations.


2) Here are two examples of why .01 is the correct number to compare to the difference between two adjacent precincts:


476 225 460 48.9 2.9 0
477 218 555 39.3 9.6 0
478 233 498 46.8 7.5 0
479 225 481 46.8 0.0 1 Match deviation <.01 ok

Changing the difference test to .1 produced this invalid match:
1573 214 598 35.8 29.7 0
1574 208 582 35.7 0.0 1 Match deviation <.1 ng

Here are more from the same run:

3618 255 484 52.7 6.6 0
3619 295 559 52.8 0.1 1 Match deviation <.1 ng
3620 220 537 41.0 11.8 0
3621 250 501 49.9 8.9 0
3622 211 407 51.8 1.9 0
3623 252 487 51.7 0.1 1 Match deviation <.1 ng

3079 286 542 52.8 11.6 0
3080 290 563 51.5 1.3 0
3081 296 574 51.6 0.1 1 Match <.1 ng
3082 259 506 51.2 0.4 0
3083 220 409 53.8 2.6 0












Correction: Only the heading was wrong.
Prob>=8 match, given that the avg. Sim. match =3:
Prob =1.19% =1-BINOMDIST(7,3800,3/3800,TRUE)

Sim. average match
match Match 3800 Assume average simulation matched pairs of 3800
Prob>=n 1 2 3 4 5 6
1 63.22% 86.47% 95.03% 98.17% 99.33% 99.75%
2 26.42% 59.41% 80.10% 90.85% 95.97% 98.27%
3 8.03% 32.33% 57.69% 76.21% 87.55% 93.82%
4 1.90% 14.28% 35.28% 56.66% 73.52% 84.90%
5 0.37% 5.26% 18.47% 37.12% 55.96% 71.52%
6 0.06% 1.65% 8.38% 21.48% 38.40% 55.44%
7 0.01% 0.45% 3.34% 11.06% 23.77% 39.37%
8 0.00% 0.11% 1.19% 5.10% 13.32% 25.59%
9 0.00% 0.02% 0.38% 2.13% 6.80% 15.26%
10 0.00% 0.00% 0.11% 0.81% 3.17% 8.38%
11 0.00% 0.00% 0.03% 0.28% 1.36% 4.25%
12 0.00% 0.00% 0.01% 0.09% 0.54% 2.00%
13 0.00% 0.00% 0.00% 0.03% 0.20% 0.88%

14 0.00000% 0.00000% 0.00034% 0.00754% 0.06911% 0.35998%