--- In [EMAIL PROTECTED], "John D. Giorgis" <[EMAIL PROTECTED]> wrote:
>Finally - some mention of a story I have not yet dared speak....
>Anybody else find it amazing that Gore picked up an almost unheard of 1,500
>votes in the recount? According to this statistician (a registered
>Democrat, no less) - this is the
>equivalent of being hit by lightning 30 times.
>
>
It�s not amazing at all. I saw the story, and there are many assumptions
underlying that statistical analysis. First of all, it�s the assumption that
there are no mistakes that involve more than one vote at a time involved.
The 300 votes that were missed before the hand recount in Volusia County is
an example of this kind of error. These types of mistakes can sway the vote
by scores or hundreds of votes at a time.
The second assumption is that Gore and Bush voters will be exactly as likely
to have partially punched cards. The demographics for the Democratic vote
indicate to me that their constituency is more likely to have this
difficulty. IIRC, when the Democrats were asked to pick sample precincts
for Dade County, they looked through the results list and found highly
democratic precincts with 25% erroneous ballots. These precincts either had
a significant higher percentage of elderly or a higher percentage of
low-income voters. One could easily understand why these voters,
particularly the elderly, had votes that were almost punched well enough for
the machine to read the first time, and had the chad knocked out of the hole
the second time. It may be worthwhile to point out that, while what I said
is speculative, it is the advancement of a hypothesis to explain data that
already exist. There has been a shift in the votes, and I am offering
hypothesis other than Democrats shifting 5 votes here and 9 votes there
under the table.
The third bit of information is the existence of shifts in recounts in other
states. New Mexico and Iowa come to mind. An Iowa Republican reported a
1000 vote shift to Bush in one county. He claimed no hanky panky, just
honest human error. If this can happen in one county in Iowa, who's total
vote was only 20% of the Florida vote, why couldn't it happen in Florida?
New Mexico also comes to mind, with a hand recount being an important part
of the shift of a several thousand vote margin for Gore to a 15 or so vote
margin for Bush.
So, I would argue that the assumptions underlying the statistical analysis
are unproved, and that data exist that strongly contradicts those
assumptions.
<< Unabashed Plug for Own Profession>>
I think that a scientist may be more able to look for the basic assumptions
underlying any statistical model. I personally think this shows some of the
advantages scientists have over mathematicians in analyzing real world data.
<< End of Plug>>
Finally, there are several things worth noting in the debate on whether a
manual recount will be beneficial. One is the existence nation wide
precedents. Are manual counts done, when possible, after machine counts have
proven to be very close? The answer, as far as I can tell, is that they are
usually done upon request and are considered the final word. An example
that a Democratic Senator raised was his own election in Nevada where the he
joined the Republicans in asking for a hand count, even though he was ahead.
He stated that he agreed to this because it was the right thing to do.
And, of course, Bush signed a Texas law calling for hand counts as the
final, best arbitrator. Yes, our ballots are a bit different, but that still
is a data point.
Finally, if my life depended on getting a count right, I believe being able
to manually inspect the ballots would be extremely worthwhile.
Particularly, since there would be easily available techniques for
highlighting the questionable ballots, as I think they are doing in Florida,
and spending most of one's time on them.
For example, if the machines can separate the cards by presidential vote,
people can know that they have to look long and hard at the machine called
Gore votes before calling them Bush votes, etc. If the machines cannot
physically seperate the cards, the human counters can. After that, the
votes can be sent in batches, and any differences between the human and
machine counts can be highlighted and subjected to fine scrutiny.
The prospects of cheating can be reduced to a very low level by the presence
of observes from both parties. If the count of a ballot is disputed, it can
be segragated for more intense review. I'd argue that the prospects of
catching earlier cheating is as large or larger than introducing new
cheating, as long as both parties are allowed to observe every step.
Dan'm Traeki Ring of Crystallized Knowledge.
Known for calculating, but not known for shutting up
_________________________________________________________________________
Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
Share information about yourself, create your own public profile at
http://profiles.msn.com.
