|
Edited on Tue Dec-28-04 07:27 PM by TruthIsAll
This is a correction to my previous post. I thought it deserved its own thread.
I incorrectly used the 2% MOE as input to the normal distribution function. I should have used the standard deviation = MOE/1.96 = .01
= NORMDIST(0.493,0.57, 0.01, TRUE)
It makes quite a difference. The odds are much more unlikely than 1 in 17,000. Read on.
Here is the orginal post: Keith Olbermann mentioned something in his earlier blog yesterday about Coshocton County, which was the ONLY county to do a full 100% recount.
KO said that the county picked up an additional 1000 votes, going from about 16,000 to 17,000. Of the 1000 new votes, Keith said they were "evenly split" between Bush and Kerry.
The county originally went 57%-42% for Bush. The additional 1000 votes were split about 50%-50%. To be exact, the Bush margin DECLINED by 15 votes in the 1000 added by the recount.
Bush's margin should have INCREASED by 150, based on the 57-42% split.
But instead, it DECLINED by 15. So we can assume Kerry got 508, Bush 493, ignoring 3rd party votes.
***************** VERY STRANGE, VERY STRANGE ************
Assuming that Bush had a 57-42% lead, here are the odds that Kerry would get 508 of the additional 1000, assuming a 2.0% MOE (very reasonable assumption, since these are actual votes):
Here is the calculation:
Assuming a 3% MOE, the odds are: ********* 1 out of 4 MILLION ! *******************
Assuming a 2% MOE the odds are: ********* 1 out of 44 TRILLION ! *******************
Here are the odds for various assumed MOE's:
MOE Prob Odds: 1 out of 5.0% 0.00127 787 4.5% 0.00040 2,509 4.0% 0.00008 12,396 3.5% 0.00001 123,531 3.0% 2.44E-07 4,085,186 2.5% 7.88E-10 1,267,605,172
2.0% 2.265E-14 44,152,937,523,240
= NORMDIST(0.493,0.57,0.01,TRUE)
|