distribution, and I'm still debugging, so these are very preliminary
results (and there are only 500 random trials so there's still a significant
error margin). But a funny thing happened when I did the L2 sims.
The success rate for both Plain Condorcet and SSD went from about
96% to 100%. Borda and Approval also improved, while IRV and
Plurality got worse.
I suspect that the improvement results from a lower incidence of tied
rankings with L2 distances. This is because the population of voters
and candidates both have integer policy values, so L1 distances are
always integers. Restricting the distances to integers makes tied ratings,
and hence tied rankings, much more likely. I suspect if I rounded the
L2 distances to integers I would see the success rates go back down.
I still don't know if L1 or L2 is more typical of how voters think, but
I'm leaning towards L1. If you disagree with a candidate only on one
issue, that's one strike against the candidate. If you disagree with him
on three issues, that's three strikes. But I also think there's a point of
saturation, where if you disagree with him on nine issues, a tenth isn't
going to make a big difference. So for a small number of issues, I
still prefer L1.
I didn't see any difference between PC and SSD, even with L1 distances.
I suspect I haven't run enough trials for a case of different results to turn
up. Non-uniform distributions might differentiate these too, as well,
but I can't really predict what effect they will have.
It shouldn't be surprising that Condorcet methods do so well, since
the standard being used here, majority potential, is a majoritarian
standard (another difference from SU).
-- Richard
Anthony Simmons wrote:
[EMAIL PROTECTED]">I think what you're more likely to find is factors that are
correlated, but not perfectly. I once heard a sociologist
say that if you're doing sociology, always be sure to include
socio-economic class as one of the variables, because it's
correlated with everything. In an election for something
like legislative office, I'd expect everything to be
correlated with political party. Problem is, they won't be
colinear, and they won't be independent.
