I disagree with those who would use a binomial model for the
overall vote totals to describe the uncertainty in the Florida vote
count. (This constitutes the Type III error discussed in another
thread--the right answer to the wrong question.) I have a bit more
sympathy for those like Paulos who point out that the error in the
vote counting process is larger than the currently claimed Bush
margin
of victory, but only a bit. With academic detachment, they do not
acknowledge that their claim, although correct, is irrelevant under
current law and doesn't help resolve the current situation.
While I am not a lawyer, my understanding of an election is that
the winner is the individual with the most votes at the final tally.
Historically, courts have decided and statutes have specified that
this includes the possibility of inherent measurement error, which
falls in the category of "those are the breaks".
The purpose of the courts is to remind us of what the process is
(which votes should be counted, which should be excluded) and what,
in
the eyes of the law, constitutes inherent measurement error. We have
been told, for example, that in many elections failure to punch a
clean hole in a ballot was not viewed as inherent measurement error.
Hand inspection was used to reduce the number of votes that went
uncounted.
The courts should interpret the relevant statutes to decide how
the votes are to be counted. If the election is a "statistical tie",
the vote totals themselves are a randomizing device and should be
used
to settle the election. An outside randomizing device such as a
coin
flip should be used to determine a winner only when the totals are
identical. Otherwise, we would be faced with the task of defining a
statistical tie and having candidates with lower vote totals
repeating
everything that is now occurring in Florida--not to establish a win,
but to establish a statistical tie and a chance at a coin flip.
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
