In article <[EMAIL PROTECTED]>,
[EMAIL PROTECTED] wrote:
<Snip>
> I'd disagree with the "some must get counted as votes" point. Some
may, but it
> would depend on people's capacity to discriminate between them. If d'
is good
> then few votes should be missassigned.
> Thom
There is a very interesting analysis on Bruce Hansen's U. Wisconsin web
page describing the yield and error rates in the analysis of undervotes:
http://www.ssc.wisc.edu/~bhansen/vote/vote.html
The link is to his web page with several articles, the lead article,
‘Recount in Dade County?' derives an 8% misclassification rate based on
the 20% of Dade County precincts that were analyzed. Hansen estimates
that a full Miami-Dade recount of the undervote would net Gore 254 votes
with 95% CI (172, 336).
Hansen's study is cited in a 12/1 article in CNN, critical of a
simplistic extrapolation (5*177) of the potential number of Gore votes
in Miami-Dade:
http://www.cnn.com/2000/ALLPOLITICS/stories/12/01/jackson.factcheck/
This CNN article cites an analysis by Peyton Young, which is not yet
available on his web page (but two other election articles are):
http://www.econ.jhu.edu/people/young/
My question is: "Why, oh why, didn't the Gore legal and statistical team
publish an expected number of votes WITH 95% confidence limits that
would result from recounts of Palm Beach and Miami-Dade counties?"
Hansen's article certainly illustrates that relatively clear-cut methods
exist for obtaining these estimates and propagating the error. The CNN
article cites Peyton Young's 1 in 3 odds of a Gore victory, but even
Judge Sauls would have to conclude that 1 in 3 is a "reasonable
probability."
--
Eugene D. Gallagher
ECOS, UMASS/Boston
Sent via Deja.com http://www.deja.com/
Before you buy.
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
