On Wed, 15 Nov 2000 16:35:33 GMT, "Robert Chung" <[EMAIL PROTECTED]>
wrote:
> <[EMAIL PROTECTED]> wrote in message 8ut1je$aef$[EMAIL PROTECTED]">news:8ut1je$aef$[EMAIL PROTECTED]...
> > i tell you want I find disturbing:
> > the "chad undercount error" that was discovered in the Volusia
> > county complete hand count went 62% to Gore and 38% to Bush.
> > However, as a whole, Volusia was only 53% Gore and 45% Bush.
> > Since when do chads play favorites, or is this entirely realistic
> > is one were to model chad failure as a Poisson process?
> >
> > mlewis
> > ut southwestern medical center at dallas
>
> Well, maybe they do, maybe they don't. I don't know that answer,
> but I have a subtler point that may apply to statistical education
> (hey, isn't it nice to get to an educational point rather than just
> to political points??). As teachers of statistics, it's an interesting
> issue to raise with students. We often presume independence
> because we can't think of any reason *for* dependence
> (or because we don't have the machinery to handle the
> subtleties of dependence even if we knew that it was there).
Volusia is not one of the huge ones.
What are the numbers behind 62/38% and 53/45% ?
Are they different enough to deserve any curiosity at all?
Also, I think Volusia's numbers, updated from first-to-end, refers to
a lost ballot boxes, etc., in addition to chad.
>
> But, why should we believe that the distribution of
> partially detached chads is independent of voting patterns?
> Perhaps blunt punch awls weren't distributed randomly. Perhaps
> clean punches vary with the angle of the punch, perhaps the
> angle of the punch varies with voter height, and perhaps height
> is correlated with voting patterns. Perhaps the force needed
Oh! that's interesting. I was picturing the *cards* as the source of
variance.
1) If it is cheaper, un-maintained equipment, that might explain why
there were a couple of African-American precincts (1000 votes, 90+%
Gore) with 20+% invalidation -- though what I read did not mention the
reason for invalidation. If these were Palm Beach, then presumably
the bulk were invalid because of two punches, not *none*.
The 10K votes with no-votes were only 2% of all of Palm Beach County,
with about 1% being the norm elsewhere; and about half the 2% will
remain as no-votes even after manual inspection (I presume).
2) The cards are ordered from different manufacturers, I suppose, by
different counties. Or by precincts. They are pre-perforated -- with
how much precision and accuracy? Even if manufacturing control is
good, I bet that a dry-and-crisp card is voted with fewer errors than
a card exposed for long to Florida's 100% humidity.
> to make a clean hole varies with age, and perhaps age
> is correlated with voting patterns. Each of these effects may
> be quite subtle, but then the differences are pretty small,
> too.
What differences are small? Actually some contrasts are pretty big,
when you get the narrow focus.
Investment tip. "Shoup" is a company that makes electronic and
op-scan voting equipment. Homepage claims 100% accuracy.
Does the nation use about 1 voting machine per 100? 200? 500? voters?
How many will be replaced before the next election?
> Alternatively, maybe there really *is* cheating going on. This
> could be part of a discussion with students, too.
No one has suggested much of that so far, where I have noticed.
A statewide recount would bring out gross accusations, I am sure.
And (probably) gross cheating.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
