Agree, but I am more interested in similarity of voters' ballots than similarity of candidates.
Not sure what you mean by similarity of ballots.
I was referring to the thing I mentioned in a previous post, which is "similarity of platform as perceived by the voters" of the candidates, which should be completely independent of "popularity of the candidate", which is what, in essense, both the pairwise matrix and your matrix measure (albeit in different ways).
Maybe that is an oversimplification, and speaking of "platform" is not exactly a mathematical term (or one that generalizes to non-political voting), but hopefully you see what I am talking about. To put it another way, I was looking for a dimension that that is more-or-less perpendicular to that which the pw matrix measures.
This is straying pretty far from my original question, but I find it interesting nonetheless.
---- election-methods mailing list - see http://electorama.com/em for list info
