[EM] [CES #8791] Upper-Bucklin naming (was: Median systems, branding....)

[Jameson Quinn]

Two points:

1.

I chatted with Rob Brown about the "upper Bucklin naming" question. His
votes were:

IRAV: F
DAT: B
Median Ranking: A
Median Rating: A
Median Grade: A
Cumulative Best Approval (CBA): B

I myself would give those latter four options C, C, B, and A respectively.

Here are my votes on Andy's proposals. I think his point about two words is
well-taken, but I'm not going to change my existing votes. Also, I think an
appropriate enough acronym could allow 3 letters/words.

-Majority Approval Voting: A
-Delayed Approval Voting: D
-Approval Level Voting: D
-Delayed Support Voting: C
-Majority Support Voting: B
-Support Level Voting: F
-Gradual Support Voting: C
-Gradual Approval Voting: B

In the spirit of his two-level-only dictum, here are a few more ideas:

-Cumulative Approval Voting: A
-Cumulative Support Voting: A
-Cumulative Majority Voting: B (But "CMV" rings a bell, and I don't think
it's just for cytomegalovirus; is there already a CMV voting system
proposal?)
-As above, but replace "Cumulative" with "additive": 1 grade lower.
-ABC (Approval-Based Cumulative) voting: C (I like the acronym, especially
if we're using letter grades; but I am not satisfied with this backronym.
Anyone else have ideas? Approval, additive; building, based, best, biggest;
cumulative, cutoff, classify... )

So currently Descending Approval Threshold (DAT) is in the lead with a
median of B. Please add your votes before Wednesday; until then, I'll just
use my favorite of whatever terms currently lead, but wrap it in "¿?"
question marks.

2.

I was thinking about how to give a GMJ-like single number for reporting a
candidate's results under ¿DAT?, and I realized that the GMJ formula itself
could work with some adjustments. The formula is:

Median + (V> - V<) / (2 * V=)

Where V>, V<, and V= are votes above, below, and at the median.

¿DAT? can use the same formula as long as you replace V< with some number
that's constant across candidates for a given election and median, and
replace V= with (Vtot - (V> + V<)). (It could also in principle work for a
constant V= if that was large enough, but I don't like that idea as much.)

So what should we use for the fake V< for reporting? Using the average (or
even better, geometric mean) of the real V< numbers for that election and
median would give the most-realistic numbers. But even a simple constant,
like 1/(2*number of grades)=10% wouldn't be too bad.

Anyway, the point is that you could pretty clearly find a way to report
¿DAT? results using one number per candidate, which removes one of my last
good reasons to prefer GMJ. And that way comes from GMJ, so my work on GMJ
isn't a total loss, which removes one of my last bad reasons to prefer GMJ
:).

So, pending naming, I think ¿DAT? is the future of Bucklin systems.

Jameson
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.electorama.com/pipermail/election-methods-electorama.com/attachments/20130617/daf86c75/attachment-0004.htm>