It's tough to do any type of analysis on data that is incomplete or
incorrect. I'm sure we've all heard a lot of "raw data" stated as fact as
this issue has been debated. The 1996 number of 14800 disqualified ballots
quoted is one possible example. I have heard recently that this includes
both the double-marked ballots and the unmarked ballots, while the 19000
number for this year is for only the double-marked ballots. Apparently the
total number (over & under) of disqualified ballots this year is ~30000. At
this point, I'm taking all numbers I hear with a grain of salt, since the
data-finding process seems more political than logical.
I usually try not to get involved in discussions regarding data that I
haven't had a chance to personally audit, but this one was too tempting...
- Eric Scharin
-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED]]On Behalf Of Robert J. MacG.
Dawson
Sent: Friday, November 10, 2000 8:02 AM
To: Peter Lewycky
Cc: [EMAIL PROTECTED]
Subject: Re: Stats on Palm Beach votes
Peter Lewycky wrote:
>
> How many may have wanted to vote for Buchanan but got Gore instead?
Assuming that errors take place at the same rates, [rather than making
any unwarranted and invidious assumptions about relative intelligence of
Buchanan
and Gore voters], and assuming that (as the regression model suggests)
about
2500 of the 3407 Buchanan votes were intended for Gore, the answer is
about 1 in 1,000
intended votes - that is to say, 1.
I
> don't have the all the stats at hand but Buchanan got something like
> 3000 votes. Is it the claim by the Democrats that all these votes
> should've gone to Gore?
I do not know what the Democrats claim. My own regression suggests that
about 2500 ofthem should have, if Palm Beach fits the pattern of the
rest
of Florida. According to this morning's paper, it was claimed by
Buchanan - not,
as I understand, a Democrat - that most of them probably should have.
There were 19000 spoiled ballots due to double
> marking this year. In 1996, 14800 ballots rejected for the same reason.
> Where's the evidence that this year's ballot was more difficult?
Right there, I'd say... if we assume no change in difficulty or voter
intelligence the variation would be due only to Poisson sampling
variation.
[I'm assuming equal turnout here.] Best fit would be Poisson(16900), SD
= 130.
That's about 16 sigmas. Enough?
-Robert Dawson
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
