In article <90dpba$47r$[EMAIL PROTECTED]>,
Gene Gallagher <[EMAIL PROTECTED]> wrote:
> Hengartner's simple analysis shows that after the machine recount in
>Broward County, the percentage of Bush to Gore votes was 31% to 69%.
>The additional votes, after the hand recount and assignment of some
>of the "undervote", were in the percentage of Bush 34% to Gore 66%.
>Hengartner simply stated that the direction of the change didn't even
>warrant a test for bias in favor of Gore. Very clear analysis.
>From your description (I haven't seen the original analysis), it
doesn't seem at all clear. Suppose that there were NO legitimate
votes missed in the machine recount - that is, suppose that the manual
recount is just manufacturing votes by divining "voter intent" from
what are in reality chance imperfections in the ballots. An unbiased
procedure for manufacturing such votes would lead to a 50-50 split (or
maybe to a 10-10-10-10-10-10-10-10-10-10 split, if you consider that
there were actually more than two candidates). The actual 34-66 split
would then certainly be a sign of bias.
Of course, there surely must have been some legitimate votes missed in
the machine recount. But there also surely must be some "votes" being
counted that do not reflect the actual intent of the voter, and it
seems quite likely that the number of such mis-assigned votes is
greater for the manual recount of ambiguous ballots than for the
machine count of ballots that seemed unambiguous to the machine. So
you would *expect* the split to be closer to 50-50 if there is no
bias. The question is whether the amount by which it has moved closer
to 50-50 is suspiciously small. I think this can be answered only by
some procedure that brings in additional information, such as an
independent audit of a random sample of the recounted ballots.
----------------------------------------------------------------------------
Radford M. Neal [EMAIL PROTECTED]
Dept. of Statistics and Dept. of Computer Science [EMAIL PROTECTED]
University of Toronto http://www.cs.utoronto.ca/~radford
----------------------------------------------------------------------------
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
