Welcome, Guest. Please login or register.
Did you miss your activation email?
December 08, 2016, 04:54:26 am
HomePredMockPollEVCalcAFEWIKIHelpLogin Register
News: Please delete your old personal messages.

Pages: [1] 2 3 4 5 6 ... 10

 1 
 on: Today at 04:36:00 am 
Started by Voter #652 - Last post by Senator Scott
Yes.  Make us proud, buddy.  And be sure to put a link to your Act Blue page in your signature. Smiley

 2 
 on: Today at 04:35:52 am 
Started by Crumpets - Last post by Eraserhead
Clinton and Sanders are the clear favorites. It's just a question of whether either of them want to run again or not.

 3 
 on: Today at 04:34:19 am 
Started by Erc - Last post by jimrtex
Ok, this is going to sound stupid at first glance, but hear me out.

One issue with most any system of drawing districts is that "nice" looking districts are in practice a Republican gerrymander, as you'll often just have handful of strong D urban districts, and a bunch of lean / strong R suburban / rural districts.

One way to fix this could be to require that each district have the same mix of urban/suburban/rural voters as the state as a whole, though then the results would really depend on how you (or the Census) defines this term.

Another, more agnostic, policy, would be to force the districts to have equal area, as well as equal population.  Thus, all the districts will have equal population density (by construction) and should hopefully have a good mix of urban, suburban, and rural populations.  Of course, if drawn by a person, you could still have all sorts of abuses (let me throw this desert or this National Park into this otherwise urban district, etc.), even if you throw out obvious problems (like water area).

But there's a way to do this algorithmically, at least in principle.

Take your state, pick some cardinal direction (let's say north), and divide your state into a northern portion and a southern portion.  Move the dividing line until you've split the state into two halves of equal population.  Now, this division will not (unless you're very lucky) result in halves of equal area.

But we can repeat this for any number of cardinal directions.  Eventually, we can build up a function f(d), the proportion of the state that is on the "d" side of the line when I divide the state into two halves of equal population.   Barring some really weird population distributions, this should be a continuous function of d (in practice if we're working with voting blocks, it won't be of course, but it should be close enough).

By construction, f(N) = 1 - f(S).  That is, if I had picked South instead of North as my cardinal direction, that's essentially equivalent, and I'm going to get the exact same line dividing my state into two portions of equal population.  Obviously, the proportion of the state that's on the north side of that line is just 1 minus the proportion that's on the south side of the line, as all of the state is on one side of the line or the other.

Thus, there's some direction where f(d) is greater than or equal to 1/2, and there's some point where f(d) is less than or equal to 1/2.  Since f(d) is continuous (or close enough, for practical purposes), there's some direction d where f(d) = 1/2, and we can divide the state into two halves that have both equal area and equal population.

There may be multiple directions d (even apart from the trivial 180-degree flip) where this is true; you can pick some other criterion to break those ties (the one with the shortest length of the dividing boundary, for instance).

For states with 2^n districts, as opposed to just 2, you can just iterate this process on each of the districts until you have enough.

Possible issues:

(1) For non-convex states, this process is not guaranteed to give you contiguous districts.  Not sure if this would be a problem in practice, as most states are pretty convex.

(2) This process is decidedly not guaranteed to work if the number of districts needed is not a power of 2.  The proof rested on the fact that north and south are equivalent if I'm dividing the state in half, which is no longer true if I'm dividing it into one-third and two-thirds portions.  It's pretty easy to find a counterexample where the process doesn't work (e.g. a circular state with some radially symmetric but non-uniform population distribution). 

This is a harder problem, but one that has a couple possible solutions:

(A) I can of course approximate 1/p by some binary fraction (e.g. 1/3 \approx 85/256), but (a) you can't slice things too fine without running into continuity problems due to the finite size of voting blocs and (b) there's then a lot of freedom in how to allocate my binary fraction approximation into different districts (e.g. which 85 portions do I put into district A, which 85 portions do I put in district B, and which 86 portions do I put into district C)?

(B) Perhaps there's some way to generalize the algorithm that works to divide a state into some odd prime number of districts, p.  Perhaps picking p-1 different directions?  Though it's not immediately obvious to me how (or if) this would work.

TL;DR In any event, despite these issues, I'd like to have a try at doing this for some different states (e.g. New Hampshire).  Anyone know of a good tool that would work for this? Dave's tool doesn't show surface area, so I expect I'll have to get into the nitty-gritty of the GIS here.
This is grossly similar in concept to the splitline algorithm.

Their initial algorithm would subdivide an area with N districts into two areas of

[N/2] and [(N+1)/2] districts, so 27 would divide into 13 and 14, and 28 would divide into 14 and 14.

The basic premise of using the shortest line is that it would lead the smallest internal boundary length. The problem is that recursive subdivision does not necessarily lead to that result collectively. At one time they discussed that a state could hold a contest where the winner would have the smallest internal boundaries, and splitline would simply set an upper limit. While someone might produce a shorter length, it would be unlikely to have a direct political bias.

My conjecture is that maintaining a more balanced number of internal vertices would result in a better global solution (e.g. districts which are pentagons have an average internal angle of 108 degrees, while triangles average 60 degrees, pentagons will tend to be more circle-like).

Splitline sometimes favors creation of needle-like triangles. A median line will tend to pass through higher density areas. In Colorado, the initial split is a north-south line through the Denver area. But a second split is not an east-west split, but rather a needle-like triangle from the Wyoming border down into the Denver area, with much of the population concentrated in the tip.

In Illinois you get a number of cuts chopping off pieces of Chicago. Eventually you will get some larger divisions down-state but mostly you have districts aligning in Chicago with a northwest-southeast orientation, reflecting some of the earlier cuts.

I had thought about a division of an area with N districts into two areas of M and N-M districts, where
1 <= M < N, and choosing the division that produced the best areal balance. Perhaps instead of area, you could use some geometric compactness measure.

A weakness of shortest line is that it may focus too much on local shapes. The shortest cut across California begins at Santa Monica because it the closest point to the Nevada border (and you would expect that a shortest line be normal to the Nevada border), plus there are lots of people in the area.

Imagine that you are applying your algorithm to California. Your first cut will be in the LA area to get enough population on both sides of the line, and will tilt northward to get enough area. Southern California is wider than northern California, so it might not have to reach to Oregon.

Perhaps an east-west line between San Francisco and Los Angeles will work for a second division. But then the northern coastal areas are going to be split into long then strips.

Back to mechanics. You will want to use a gnomic projection, which ensures that straight lines on your map correspond to great circles on a sphere.

If your atoms are census blocks, each has:
(1) population, which may be 0;
(2) an internal point, which is typically the centroid of the block, but may be shifted to be internal to the census block, and for your purposes is sufficient to represent the population;
(3) a land and water area.

It is relatively trivial for an initial cardinal angle of North, to sort the blocks by a y coordinate (after the gnomic projection) and find the population median, and calculate the total area of the subareas.

For a dividing line of angle d, you don't have to do a full rotation. You can use:

x' = x
y' = y + x sin(d)

For a given angle, sin(d) is a constant.

In effect for a given angle, you don't have to slide the dividing line parallel to the initial line, but can slide it somewhat obliquely, since that line will also be parallel, and you aren't determining distance, but merely whether a point is to the left or right of the line.

You might be able to simply step d by some increment, and then perhaps refine near cross-over points.

I am not confident that f(d) is continuous.

A problem with using census blocks is that there is a ginormous amount of them. Larger states have more than one million census blocks.

An alternative would be to use block groups, which are collections of census blocks with a couple of useful features:

(1) They are roughly 1000 persons in size, so they can aggregate lots of census blocks in sparsely populated areas;
(2) They may have some degree of demographic homogeneity, since they are intended to produce useful statistical data from the ACS, thus districts may have a little bit of COI coherence.

If you use block groups, you probably want to use the center of population (perhaps this is already done by the Census Bureau) for location.

QGIS is quite useful for converting spreadsheet data to maps.

Regarding the power of 2 problem:

I had at one time thought at one time apportioning representatives to whole units like counties, and then using some automated process to divide the residual population.

Think of California, and let's say Los Angeles County is equivalent to 12.437 districts. We apportion 12 districts to Los Angeles, and then assign the residual population of 0.437 districts throughout the county (0.437/12.437) * population of each atom.

Then divide the residual from all counties on a statewide basis.

But let's say we create 1024*53 districts in California. We can apportion those to each county, and perhaps a secondary apportionment to individual cities. Then create districts within each city and county.

And finally aggregate the districts pairwise in 10 rounds to create the final 53 districts.


 4 
 on: Today at 04:29:57 am 
Started by DPKdebator - Last post by BlueSwan
Yes. He benefits from the most ridiculously low expectations that any world leader has ever had. Any time he does something that is not completely batsh*t crazy, everybody is fawning over him as if he is the second coming of christ. Also he benefits from nobody taking any of his campaign promises seriously and/or actually hoping that he will do any of it. This means that if for instance he abandons the wall, or abandons the muslim ban, or abandons the prosecution of Clinton, or abandons the teare-up of the Iran deal, or abandons withdrawing from the Paris treaty, or abandons ending free trade, or abandons any of his other moronic ideas, then people will be relieved, whereas most politicans would instead face stern criticism for misleading the public.

 5 
 on: Today at 04:27:52 am 
Started by pppolitics - Last post by President Blair
Romney was very much the stereotype for the typical wall street republican; made his wealth from venture capitalism, governed as a moderate only to slam to the right.

Tbh though Bobby Jindal fits the bill pretty well

 6 
 on: Today at 04:23:25 am 
Started by Jante's Law Revivalist - Last post by TN Volunteer
Context: speaking of Mike Pence:

Why would he leave a ticket that is almost certain to win?

Ha!

 7 
 on: Today at 04:20:59 am 
Started by Alpha - Last post by Old School Republican

With this, I remain a believer in the concept of the Jewish, Israeli state, but it should've been socialist, actual egalitarian socialism (mapam), and Israel wouldn't be as such segregated and discriminatory to the Palestinian People.

(Ignoring the normal, ugly, Israel vs Palestinians arguments in this thread)

I understand your point of view as a Socialist, but Mapam was at times a low point for Israeli democracy (though the current government is trying really hard to compete). Their almost absolute power harmed the freedom of speech and created a nest of corruption and discrimination towards Jewish immigrants from Muslim countries. In the end, their slight authoritainism sewed the seeds of some characteristics of current Israeli society that I dislike, like the sanctifying of the army and anything related to it, the corrupt primaries in the major parties, that are controlled by special interest groups, and some generally corrupt and disfunctional systems (like the Broadcasting Authority and the main labour unions).

They were the only one's that had Palestinian members, had a policy for co-existence, and didn't support the expulsion of Palestinian from their lands, causing the problems they had, a lot of which Mapai did or didn't support. I believe if Mapam had been elected, Israel and Palestine would've been a been a better place, though more soviet-orientated it would have been. I really don't understand how a party that only once got second place had absolute control

Oh, I thought you argued for Mapai, not Mapam. Sorry.
In that case, I'm of the opinion that it would've been even worse. It's a fact that the Arabic countries at the time weren't willing to allow Israel to exist, so while Mapam might have been more compassionate for the Palestinians, they would still have to fight for survival. And if they got close to the Soviets, they would probably lose, because the Soviets favoured the more numerous Arabic countries while America was willing to support Israel. Not gonna delve into who's right and who's wrong, but the US was absolutely vital for the survival of Israel in the October War.

If you had a socialist Israel, that allowed the support for the Palestinian people and had USSR support, would the Arab nations attacked it? Mapam were the only one fighting for a egalitarian, equal Israel, while other parties were fine in treating Arabs as treating second class citizens. Initially when there looked to be a socialist Isreal, Stalin was fine in supporting the establishment and supported such a state. I'm not supporting the USSR, and Stalin was of course horrific, but I believe Israel under socialism would be much more equal and accepting of Palestinian rights and would be a more peaceful religion, than it is now. That's unless a communist dictatorship is established, but I doubt that'd happen.

Um that did happen, and Isreal was socialist for some time and probably still is compared to America.

 8 
 on: Today at 04:20:29 am 
Started by riceowl - Last post by President Johnson
So, bets on next year?

Mike Pence Tongue

 9 
 on: Today at 04:18:25 am 
Started by ERM64man - Last post by RFayette
The support of partial birth abortion should disqualify someone from public office.

Thinking things I don't like should be a crime.

Murder should be a crime.

Well, she also supports eugenics, so...

"Well, you know what's wrong with the world today
People done gone and put their Bibles away
They're living by the law of the jungle not the law of the land"

https://www.youtube.com/watch?v=x_1H2njznEU

 10 
 on: Today at 04:17:42 am 
Started by ERM64man - Last post by Intell
Awful (Not Insane).

Pages: [1] 2 3 4 5 6 ... 10


Login with username, password and session length

Logout

Powered by SMF 1.1.21 | SMF © 2015, Simple Machines