[EM] Munsterhjelm election sims

Kristofer Munsterhjelm km-elmet at broadpark.no
Tue Jun 8 14:50:00 PDT 2010


Warren Smith wrote:
> You have to be clearer.  E.g.
> 
> QUOTE
> If N = total number of voters, then (unless there's a bug somewhere) we get:
> 
> PA_Linear_Range_STV                                 0.29813 0.00031
> 
> If N = the number of voters voting according to the ballot being
> reweighted, then the result is:
> 
> PA_Linear_Range_STV                                 0.20928 0.01991
> END QUOTE.
> 
> I have absolutely no idea what any of that meant.  It is just
> undefined gobbledegook.
> Everything you wrote is meaningless to me.   It would help to have
> your program print out  summary complete with keys, definitions, etc
> so that (1) somebody can understand what you are talking about and (2)
> you do not have to keep telling us over and over again what you (now)
> mean whenever you rerun revised program, the printout will just tell
> us.

See my previous posts and 
http://munsterhjelm.no/km/elections/multiwinner_tradeoffs/ . Very simply 
put, the former number is a measure of proportionality, and the latter 
is a measure of Bayesian regret. They are scaled so that 0 is best and 1 
is worst. To find out the details (proportionality measure, utility 
calc), see the page or previous posts to the list.

As for what N means, I quoted Jameson's formula. Since he says "N = num 
voters", it was ambiguous as to whether the number of voters was the 
number who voted the particular way, or the number of voters in total. 
In other words, if I have an election of the type

10: A (0.9) > B (0.5) > C (0.1)
  5: B (0.7) > C (0.3) > A (0.2)

and I'm processing the first ballot set (the 10: ...), then if N = 
number of voters who voted the particular way, it would be equal to 10. 
Otherwise, it would be equal to 15.

"Linear Range STV" is just the name of the method, and PA is a prefix 
for the stats output (as opposed to, for instance, TPF for social 
optimum Pareto front).


If you think it'll help, I could probably add a header so you get:

NAME                                                PROP.   REGRET
PA_Linear_Range_STV                                 0.20928 0.01991

Even so, you'd have to look at the page and/or previous posts to know 
what those fields *mean*.


More information about the Election-Methods mailing list