Redistricting Metrics.

redistmetrics

redistmetrics is one of the R packages developed and maintained by the ALARM Project. redistmetrics provides the back-end for the computation of summary statistics for a redistricting plan. It provides a more direct access point to use methods in redist without requiring redist objects.

Installation

You can install the stable version of redistmetrics from CRAN with:

install.packages('redistmetrics')

You can install the development version of redistmetrics from GitHub with:

if (!requireNamespace('remotes')) install.packages('remotes')
remotes::install_github('alarm-redist/redistmetrics')

Example

library(redistmetrics)

redistmetrics offers support for 4 common input types and has examples of each, all based on New Hampshire:

data(nh)

This example is based on comp_polsby() for the Polsby Popper compactness, but comp_polsby() can be substituted for any implemented measure!

Single Plan:

For a single plan, we can pass the single plan to the input. We also pass an argument to shp which takes in an sf dataframe. r_2020 here is the Republican proposal for New Hampshire’s congressional districts.

comp_polsby(plans = nh$r_2020, shp = nh)
#> [1] 0.2324375 0.1582763

The output here is a numeric vector, where each entry is the output for a district. The first district here has a compactness of about 0.23 and the second district has a compactness of about 0.16.

Now, if you’re redistricting in R, we recommend using the R package redist. In which case, you would have a redist_map object.

We can load an example here with:

data(nh_map)

For redist maps, the workflow is identical!

comp_polsby(plans = nh_map$r_2020, shp = nh)
#> [1] 0.2324375 0.1582763

Multiple Plans:

For multiple plans, we can pass either a matrix of plans or a redist_plans object to plans. We will still need nh or nh_map to provide the shapes.

If we have a matrix, we can compare with nh_m a matrix of plans, where each column indicates a plan.

data(nh_m)

From there, the process is nearly identical. Here we compute the Polsby Popper compactness for the first two columns:

comp_polsby(plans = nh_m[, 1:2], shp = nh)
#> [1] 0.1844955 0.1796426 0.2324375 0.1582763

Now we got 4 outputs: 1 for each district x 2 for each plan x 2 plans.

If we are using redist, we likely have a redist_plans object which hides the matrix as an attribute to give a more familiar tidy workflow. With that, we can do a very similar process:

First, we load the plans object (included as an example):

data(nh_plans)

The benefit of using a redist_plans object is that we can cleanly mutate into it using the . shortcut:

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
nh_plans <- nh_plans %>% mutate(polsby = comp_polsby(plans = ., shp = nh))
#> Linking to GEOS 3.9.3, GDAL 3.5.2, PROJ 8.2.1; sf_use_s2() is TRUE

Now our values are cleanly held in the redist_plans object:

head(nh_plans)
#> # A tibble: 6 × 4
#>   draw   district total_pop polsby
#>   <fct>     <int>     <dbl>  <dbl>
#> 1 d_2020        1    688739  0.184
#> 2 d_2020        2    688790  0.180
#> 3 r_2020        1    688676  0.232
#> 4 r_2020        2    688853  0.158
#> 5 1             1    688961  0.235
#> 6 1             2    688568  0.349

Detailed information on each measure are contained in the vignettes and references are contained in the function documentation.