"Jerry Dallal" <[EMAIL PROTECTED]> wrote in message
[EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> The design of butterfly ballot still seems to be generating the most
> press. What's more interesting to me is the request for recount by
> hand.
>From a statistical standpoint, I think we ought to be interested in
both. I think Jerry's point about hand recounts is correct, but I've
yet to see discussion about the relative precision of hand counts
versus machine counts. I would think that depends on what the
cause of error would be (for example, partially punched holes), and
that would depend on the mechanism by which the original
vote was taken. That gets us back to ballot design.
As statisticians, we need to make certain assumptions about
the underlying error model in order to make estimates. In this
case, the underlying error model is clearly not as simple as
the usual sampling or measurement error models.
> Comment on the FL vote threads: The statistical insights are valuable.
> I'm surprised by the level of the off-topic, heated political debate
> that same [sic] have tried to introduced.
> But, if we treat the votes like any other dataset and go
> about our business, it'll all get taken care of.
I agree. The majority of posts have tried to stick to statistical
insight which (I hope) is our strength. It is not up to us (as
statisticians; perhaps it is as individuals) to determine how things
will be resolved in Florida. It *is* however, fair for us to ask
questions about the purported error and to make estimates
about its magnitude.
An interesting discussion of ballot usability issues can be found
at http://www.bricklin.com/log/ballotusability.htm
At the site http://cuwu.editthispage.com/2000/11/08
is a nice little interactive graph that lets you see not only
the county total votes for Bush, Gore, and Buchanan,
but also the votes for Nader. The Buchanan-Nader graph
looks particularly linear, and has close to the same scale.
--Robert Chung
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
