The article begins below the line. This stuff above the line is just related junk.

Hyperlinks: google underlines them, but of course you can change the style to whatever you want, the underline is

not intended by me per se.

About Sources: Official election #s in Ohio from the Ohio Secy of State website.

Fitrakis pro-choice abortion stand - please confirm/deny that yourself!

Abortion numbers - Alan Guttmacher institute.

Ohio gerrymandering: /Ohio06b.html . Iraq war polls:

54% say was mistake to send troops to Iraq (CNN/USA Today/Gallup) CNN story 25 June 2004

56% say Iraq war a mistake not worth fighting: Washington Post-ABC News poll Washington post 21 Dec 2004,

55% in CBS news story 10 Oct 2005,

55% in CNN poll+story 12 June 2006,

Gallup poll 2-4 Mar 2007 (press story 6 Mar): 58% want troops to be withdrawn within 12 months, 59% say war=mistake,

and only 13% support sending more troops.

Vietnam war:

Gallup polls, starting in August-September 1968, found a majority of Americans said the war was a mistake,

and the percentage basically kept rising with time, reaching 69% in 2000.

Nader 90% switchover: NES data (National Election Study, Univ. of Michigan).

US elections 98% predictable: /OneParty.html see re Ron Faucheux predictions database.

Lieberman run illegal in 46/50 states & ballot access laws & 2500 threshold: Richard Winger at Ballot Access News,

see also /BallAccess.html .

Range voting (and IRV) properties and computer simulations (and honeybees) and

shortest splitline algorithm: .

Study by Brams & Fishburn = chapter 9 of their book "Approval Voting".

Columbus Dispatch poll - am going by a report in the Free Press (you might want to make it a hyperlink),

I did not see original Dispatch article, and I used standard error not 95% confidence window.

I worry that the Free Press report on this seemed deceptive because

there were many other polls, I think all predicting 52-62% for Strickland, and it did not mention them. See

http://en.wikipedia.org/wiki/Ohio_gubernatorial_election,_2006 for polls in Oct+Nov predicting

55+-5, 59+-2, 59, 62+-4, 59, 53, 52+-4, 60+-4, and 52 which actually would mean Strickland did better than expected.

Checks on the Dispatch's math:

If N independent people are found in a poll to cast pN votes for Strickland,

then the standard error (std. deviation) is sqrt(Np(1-p)) which

expressed as a fraction of the total number N of votes, is sqrt(p(1-p)/N).

Here, N=1541 and p=0.67 yielding sqrt(.67*.33/1541) = 1.198% as the std error;

the 95% confidence interval is then 1.96*stdDev = 2.35.

Sp it should have said (67 +- 2.35)% as the 95% confidence interval, not 67 +- 2.2.

A slight error.

The idealization of independent people is definitely false for exit polls (and you need to multiply

the standard error by a fudge factor), but for random telephone polls of people

who say they are "likely voters," it should be accurate (albeit then you are sampling

from the wrong distribution).

The discrepancy between 67 and 60.5 is 6.5, which is 6.5/1.198 = 5.43 standard errors.

The chance of a deviation this large or larger is about 20 billionths. If by 67 they really meant 67.5

then it's 5.84 standard errors and the chance is about 2 billionths; if by 67 they really meant 66.5

then it's 5.00 standard errors and the chance is 287 billionths.

Author Bio: Warren D. Smith is a math PhD and co-founder of .

Currently 1368 words including title: