[EM] Another alternative to Tideman's estimating method

[MIKE OSSIPOFF]

[I tried to post this about a week ago, but it was returned from EM,
and so I'm re-posting it]


Suppose there are no 2 candidates who seem almost sure to be the top
2, but that it isn't a 0-info election. You don't have the information
needed for the "best frontrunner" strategy, or the extensions of it
that I suggested.

One could use the
election-utility strategy, voting for those candidates who are so
good that you'd like to have them as president instead of holding
the election.

Or you could use Tideman's estimating method, using
the probability that candidate i will win for estimating Pi, the
probability that if, when all the votes but yours are counted,
if some 2 candidates are in a tie for 1st, or are the top 2, differing
by only one vote, then i will be one of those 2. It's also assumed
that the probability of candidate i being the upper in that near-tie
is equal to his probability of being the lower in that near-tie.

Pij can then be estimated by assuming that:

Pij = Pji = Pi*Pj.

But it seems to me to be easier to estimate the probability that
some particular candidate will be the villain that you'll have
a chance to make someone better beat. And the probability that some
other candidate will be the hero that you can make beat him.

One could make those estimates exhaustively to estimate the Pij
directly, or could use them to get estimates for the Pi.

This is vaguely described, because there are probably various ways
these additional estimates could be made, and because I haven't
tried it yet.

Mike Ossipoff




_________________________________________________________________
Join the world’s largest e-mail service with MSN Hotmail. 
http://www.hotmail.com

----
For more information about this list (subscribe, unsubscribe, FAQ, etc), 
please see http://www.eskimo.com/~robla/em