redist: Markov Chain Monte Carlo Methods for Redistricting Simulation

R package for simulating redistricting plans via Markov chain Monte Carlo by Ben Fifield (bfifield@princeton.edu), Alex Tarr (atarr@princeton.edu), Michael Higgins (mjh5@princeton.edu), and Kosuke Imai (Imai@Harvard.Edu). Maintainer is Ben Fifield.

Installation Instructions

The package is available on CRAN and can be installed using:

install.packages("redist")

Users can also install the most stable development release of the redist package using the install_github() function in the devtools package.

library(devtools)
install_github("redistricting/redist")

We hope the following guide will be of help to users who want to take a look at the original redist source code:

sw_mh_alg.cpp: Contains the swMH() function, which conducts Markov chain Monte Carlo simulation of redistricting plans.
sw_mh_helper.cpp: A series of functions to aid in simulating redistricting plans.
make_swaps_helper.cpp: A series of functions to propose and make swaps of geographic units in the primary redistricting algorithm.
constraint_calc_helper.cpp: Functions to calculate the strength of certain implemented constraints such as population and compactness requirements.
rsg.cpp: An implementation of the random seed-and-grow algorithm described in detail in Chen and Rodden (2013).
check_contiguity.cpp: A contiguity check for the implementation of the Chen and Rodden (2013) algorithm in rsg.cpp.
redist_analysis.cpp: Functions to aid in analysis of simulated redistricting plans.
enumerate.cpp: Functions called by enumerate.R that allow users to fully enumerate all valid, contiguous redistricting plans for a given set of geographic units.

Flip sign for cold temperatures (currently fed in as negative values, should be positive to fit with paper)
Feed in betaweights as a numeric argument with an exponential sequence
Add summary function that calculates acceptance probability, the function call with parameters, and the distribution of beta values for tempering
Make q and lambda dynamic - for example, check acceptance probability every 50 iterations. If too high, increase lambda or decrease q