Allowing hard constraint for splitting counties/administrative entities?

[martial00120]

Hi
How about adding the possibilty to have a hard constraint for splitting counties? In Germany, we have a higher tolerance for population deviation in the US (+/- 25%) but we optimize more in the direction of not splitting counties/county-equivalent areas. So how about opening up the possibility that at max .... counties can be split or that a county can at max be split k times? Then your project could be more helpful for the German application.

Sebastian Goderbauer, a Operations Research scientiest has published a proposal for redistricting in Germany, which optimizes mainly on this constraint, but his Code is unfortunately not available (and his thesis is in German).

[CoryMcCartan]

We do have a hard split constraint! It can be accessed with the county= argument to redist_smc(). That will guarantee fewer splits than the number of districts minus 1 (though in practice you will often get many samples with fewer splits). If you need even fewer splits, you have two options:

• Combine county= with a soft county constraint (see https://alarm-redist.org/redist/reference/constraints.html)
• Merge all the units (e.g. precincts) in one or more counties together before running the sampler. The extreme version of this is to do your sampling at the county level, which will guarantee zero splits. See https://alarm-redist.org/redist/articles/map-preproc.html#merging for an example of this technique.

Hope that helps.

[martial00120]

Ok, thank you! I could not try out your solutions yet.