[EM] Simulations

[Richard Moore]

Forest Simmons wrote:

> Why not estimate who the front runners would be based on which of the
> candidates have the largest Dirichlet regions in policy space?
> 
> The Dirichlet region (also known as the Voronoi polygon) for a candidate
> is the set of points (potential voters in this case) that are closer (in
> whatever metric) to that candidate than to any other candidate.

That might be a valid approach, but I don't plan to
go back and redo the sims.


> Alternately, you could randomly choose two of the three candidates with
> the largest Dirichlet regions, to take into account that popularity is
> determined by money, press, personality, etc. as much as by stance on the
> issues.

Yes, some random factors definitely should be included.

 
> If calculation of the Dirichlet regions is too time consuming, you could
> take a random poll of the voters, and determine their responses by which
> candidate they are the closest to.  If the poll sample is small, this
> would throw in the random factor that we want, but bias it towards the
> candidates with the largest Dirichlet regions.

Or (very easy method) do it by random ballot, a poll
with a sample size of 1. The sim shows that this
method has a pretty small amount of top bias, which
is probably just about the effect that's needed.

Richard