Bivarite Mapping with ggplot2

Thematic choropleth maps are used to display quantities of some variable within areas, such as mapping median income across a city’s neighborhoods. However, we often think in bivariate terms - “how do race and income vary together?”. Maps that captures this, known as bivariate choropleth maps, are often perceived as difficult to create and interpret. The goal of biscale is to implement a consistent approach to bivariate mapping entirely within R. The package’s workflow is based on a 2019 tutorial written by Timo Grossenbacher and Angelo Zehr. biscale also contains a set of bivariate mapping palettes, including Joshua Stevens’ classive color schemes.

Getting Started

Installation

The easiest way to get biscale is to install it from CRAN:

install.packages("biscale")

Alternatively, the development version of biscale can be accessed from GitHub with remotes:

# install.packages("remotes")
remotes::install_github("chris-prener/biscale")

Installing Suggested Dependencies

Since the package does not directly use functions from sf, it is a suggested dependency rather than a required one. However, the most direct approach to using biscale is with sf objects, and we therefore recommend users install sf. Windows and macOS users should be able to install sf without significant issues unless they are building from source. Linux users will need to install several open source spatial libraries to get sf itself up and running.

The other suggested dependency that users may want to consider installing is cowplot. All of the examples in the package documentation utilize it to construct final map images that combine the map with the legend. Like sf, it is suggested because none of the functions in biscale call cowplot directly.

If you want to use them, you can either install these packages individually (faster) or install all of the suggested dependencies at once (slower, will also give you a number of other packages you may or may not want):

## install just cowplot and sf
install.packages(c("cowplot", "sf"))

## install all suggested dependencies
install.packages("biscale", dependencies = TRUE)

Functions

All functions within biscale use the prefix bi_ to leverage the auto-completion features of RStudio and other IDEs.

Data

biscale contains a data set of U.S. Census tracts for the City of St. Louis in Missouri. Both median income and the percentage of white residents are included, both of which can be used to demonstrate the package’s functionality.

Preparing Data

Once data are loaded, bivariate classes can be applied with the bi_class() function:

# load dependencies
library(biscale)

# create classes
data <- bi_class(stl_race_income, x = pctWhite, y = medInc, style = "quantile", dim = 3)

The dim argument is used to control the extent of the legend - do you want to produce a two-by-two map (dim = 2), a three-by-three map (dim = 3), or a four-by-four map (dim = 4)? Note that support for four-by-four mapping is new as of v1.0.0!

Classes can be applied with the style parameter using four approaches for calculating breaks: "quantile" (default), "equal", "fisher", and "jenks". The default "quantile" approach will create relatively equal “buckets” of data for mapping, with a break created at the median (50th percentile) for a two-by-two map or at the 33rd and 66th percentiles for a three-by-three map. For a four-by-four map, breaks are created at the 25th, 50th (median), and 75th percentiles.

With the sample data, this creates a very broad range for the percent white measure in particular. Using one of the other approaches to calculating breaks yields a narrower range for the breaks and produces a map that does not overstate the percent of white residents living on the north side of St. Louis:

Users should therefore choose methods for calculating breaks carefully as well as the number of dimensions mapped.

As of v1.0, biscale now supports larger dimension maps as well as custom breaks for maps. For an overview of these options, see the “Options for Breaks and Legends” vignette.

Bivariate Mapping with Biscale

Creating Maps

Once breaks are created, we can use bi_scale_fill() as part of our ggplot() call:

# create map
map <- ggplot() +
  geom_sf(data = data, mapping = aes(fill = bi_class), color = "white", size = 0.1, show.legend = FALSE) +
  bi_scale_fill(pal = "GrPink", dim = 3) +
  labs(
    title = "Race and Income in St. Louis, MO",
    subtitle = "Gray Pink (GrPink) Palette"
  ) +
  bi_theme()

This requires that the variable bi_class, created with bi_class(), is used as the fill variable in the aesthetic mapping. We also want to remove the legend from the plot since it will not accurately communicate the complexity of the bivariate scale.

The dimensions of the scale must again be supplied for bi_scale_fill() (they should match the dimensions given for bi_class()!), and a palette must be given. For reference, the top image above uses the "DkBlue" palette, the map images in the breaks section above use the "DkViolet" palette, and the map constructed in this section (and displayed below) uses the "GrPink" palette. Note that, as of v1.0.0, the number of options for palettes has been expanded and there is increased support for custom palettes. See the “Bivariate Palettes” vignette or ?bi_pal for more details.

If you are mapping point data, there is an alternative function bi_scale_color() that works the same way as bi_scale_fill().

The example above also includes bi_theme(), which is based on the theme designed by Timo Grossenbacher and Angelo Zehr. This theme creates a simple, clean canvas for bivariate mapping that removes any possible distracting elements.

Creating Legends

We’ve set show.legend = FALSE so that we can add (manually) our own bivariate legend. The legend itself can be created with the bi_legend() function:

legend <- bi_legend(pal = "GrPink",
                    dim = 3,
                    xlab = "Higher % White ",
                    ylab = "Higher Income ",
                    size = 8)

The palette and dimensions should match what has been used for both bi_class() (in terms of dimensions) and bi_scale_fill() (in terms of both dimensions and palette). The size argument controls the font size used on the legend. Note that plotmath is used to draw the arrows since Unicode arrows are font dependent. This happens internally as part of bi_legend() - you don’t need to include them in your xlab and ylab arguments!

Creating a Final Output

With our legend drawn, we can then combine the legend and the map with a package like cowplot. The values needed for this stage will be subject to experimentation depending on the shape of the map itself.

# combine map with legend
finalPlot <- ggdraw() +
  draw_plot(map, 0, 0, 1, 1) +
  draw_plot(legend, 0.2, .65, 0.2, 0.2)

This approach allows us to customize legend’s placement and size to suit different map layouts. All of the maps shown as part of this vignette were produced using this approach. There are other approaches you could take as well that do not use cowplot. Note that, because cowplot requires R v3.5, it is not included as a suggested package (to maintain backward compatibility). Beginning with v1.0.0, there are additional options for constructing legends with biscale. Please see the “Options for Breaks and Legends” vignette for more details!

The completed map, created with the sample code in this vignette, looks like this:

Getting Help

If you are new to R itself, welcome! Hadley Wickham’s R for Data Science is an excellent way to get started with data manipulation in the tidyverse, which biscale is designed to integrate seamlessly with.
If you are new to spatial analysis in R, we strongly encourage you check out the excellent new Geocomputation in R by Robin Lovelace, Jakub Nowosad, and Jannes Muenchow.
If you have questions about using biscale, you are encouraged to use the RStudio Community forums. Please create a reprex before posting. Feel free to tag Chris (@chris.prener) in any posts about biscale.
If you think you’ve found a bug, please create a reprex and then open an issue on GitHub.

Suggesting Features or Changes

If you have features or suggestions you want to see implemented, please open an issue on GitHub (and ideally created a reprex to go with it!). Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.

Christopher Prener, Ph.D.

2022-09-15