Peter wrote: > On Wed, 2007-02-07 at 07:07 +0100, Jobst Heitzig wrote: >> where c = 1 / sum_ij (ri-rj)² > Should be c = 1 / sqrt(sum_ij (ri-rj)²)
Ah, now I see. It would have been helpful had I noticed that root symbol of yours :-) So, like Warren, I'm finally convinced, but you could definitely have had that easier... By the way, I find the following way of writing your formula helpful: Probability of option X winning given that voter V's ballot is drawn := 1/N + (V's standardized rating of X) / (N-1)sqrt(N) where "standardized" means the usual thing (expressed rating minus the mean of all of voter V's ratings, divided by the standard deviation of those ratings, using the usual version of the standard deviation for sample data, that is, with denominator N-1). Yours, Jobst ---- election-methods mailing list - see http://electorama.com/em for list info
