<table cellspacing="0" cellpadding="0" border="0" ><tr><td valign="top" style="font: inherit;">Hello. It's been quite a while since<br>
I posted here. I have a question: does<br>
anyone have any good pointers<br>
to material on metrics on elections?<br>
<br>
A "metric" is as usual, and an<br>
"election" would be simply an<br>
election profile, that is you have<br>
some set S of permitted ballot types,<br>
and so many ballots of type 1, so many of<br>
type 2, ...<br>
<br>
Though these sound like subsets-of-S-with-<br>
repetition-allowed, you can view them<br>
as functions from S to N (N=0,1,2,...),<br>
where for x in S, f(x) is the number of<br>
ballots of type x.<br>
<br>
So metrics on the set N^S are what is<br>
of interest.<br>
<br>
One obvious way to do this is simply to<br>
take metrics on R^#S (R=reals), restricted<br>
to N^S. However S itself has a metric,<br>
so I was really after metrics on N^S<br>
which reflect the metric on S,<br>
rather like the Hausdorff metric gives<br>
a metric on the finite subsets of an<br>
arbitrary metric space. <br>
<br>
[I've appended a summary of the<br>
Hausdorff construction.]<br>
<br>
In practice, of course, S is a finite set<br>
but I've already found a couple of<br>
constructions on N^S, incorporating<br>
a metric on S, which<br>
turn out to be themselves metrics<br>
on elections under certain more general<br>
conditions (than when S is finite).<br>
<br>
The fact that they actually are metrics<br>
matters because it simplifies the<br>
calculations quite a bit.<br>
<br>
I don't currently have access to a<br>
university library, so I'd prefer<br>
some specialised on-line resources.<br>
<br>
TIA<br>
Stephen Turner<br>
--------------<br>
Hausdorff metric: let M be a metric space<br>
with metric d:MxM->R. If x is in M and<br>
A,B are two non-empty finite subsets of M, we define<br>
the distance from x to B as usual, namely<br>
<br>
d(x,B) := min d(x,y)<br>
<br>
where the minimum is taken over all y in B.<br>
Then we could define f(A,B) by<br>
<br>
f(A,B) := max d(x,B)<br>
<br>
where the maximum is taken over all x in A.<br>
Finally we define<br>
d(A,B) = max (f(A,B),f(B,A)), and this<br>
is called the Hausdorff metric on<br>
the set of (non-empty) finite subsets of M.<br>
-----------------------------------------------------</td></tr></table><br>