Warren says:
Far as I can see, the simple method that starts out
http://rangevoting.org/NewAppo.html is identical with the method for
apportionment that Ossipoff now proposes ("What I mean by Weighed Webster")
I reply:
I guess its necessary to take your word for that, because didnt try to
decypher the method and derivation at your website. If you had briefly and
concisely stated your method and its derivation, as I had requested, then
Id have acknowledged it, assuming that it was really up at your website at
that time.
Warren continues:
glad to see Ossipoff has now come around to my point of view
I reply:
Warren must be referring to the point of view that unbias is best achieved
by making each cycles s/q = 1. But wait, Warren only expressed disagreement
with that approach.
But did Warren post Weighted Webster at his website, while expressing
disagreement with the unbias approach on which it is based? Who knows.
Warren continues:
Note, the parameters Ossipoff is calling A and B can be deduced from the
number of states and total country population
I reply:
Yesterday I told how to do that. And I told why least squares would almost
surely be more accurate.
Warren continues:
in which case his formula should simplify to become my formula. Warren D
Smith http://rangevoting.org
Some details:
Ossipoff: R = (-1/A)*ln[(-1/A)*{exp(-A*b)-exp(-A*a)}] R is the rounding
point between consecutive integers a & b. I hope that that expression can
be simplified. --
response, use b=a+1
I reply:
Of course that should be done.
, then {exp(-A*b)-exp(-A*a)} becomes exp(-A*a) * {exp(-A) - 1} now note R-a
is going to come out depending only on A, B and not on a,b just as in
http://rangevoting.org/NewAppo.html ----
I reply:
So it does. A fixed rounding point (the same in all the cycles). I couldnt
ask for more simplification than that.
By the way, I would propose that A be determined, for and from each census,
by the most accurate method available, which is probably least-squares.
Mike Ossipoff
----
election-methods mailing list - see http://electorama.com/em for list info
