Choroplethr v3.6.0 is now on CRAN

April 17, 2017
/ Blog Open Data Open Source
/ By Ari Lamstein

Choroplethr version 3.6.0 is now on CRAN. This version adds functionality for getting and mapping demographics of US Census Tracts. You can install it from the R console as follows:

install.packages("choroplethr") 
packageVersion("choroplethr") 
[1] ‘3.6.0’

To use this functionality you will need an API key from the US Census Bureau. You can learn more about that here.

[content_upgrade cu_id=”5608″]Bonus: Download the code in this post![content_upgrade_button]Click Here[/content_upgrade_button][/content_upgrade]

Getting Tract Maps

Historically choroplethr has had limited support for Census Tracts. This is because the US Census Bureau releases tract maps on a per-state basis, and it wasn’t feasible to create a separate package for each state.

Choroplethr now uses the tigris package to download Tract maps from the Census Bureau on demand. The function to retrieve a map is get_tract_map. Here’s an example of retrieving and rendering a map of Tracts in New York:


library(choroplethr)
library(ggplot2)

?get_tract_map
ny = get_tract_map("new york")
ggplot(ny, aes(long, lat, group=group)) + geom_polygon()

Getting Tract-Level Demographics

Choroplethr contains both maps and interesting data to map. If you want to explore the demographics of US Census Tracts then use get_tract_demographics:


# see help for extra options
?get_tract_demographics

ny_stats = get_tract_demographics("new york")
head(ny_stats)

region      total_population percent_white percent_black percent_asian percent_hispanic per_capita_income median_rent median_age
36001000100 2163             19            57            2             19               19065             596         36.9
36001000200 5335             9             72            0             13               15376             501         27.8
36001000300 6077             35            44            3             17               20804             743         31.0
36001000401 2380             88            7             2             2                39574             1198        65.5
36001000403 4338             65            19            11            5                32397             859         41.9
36001000404 4932             69            12            7             9                2479              NA          19.6

Creating Choropleth Maps

Now that we have a map and spatial data, we can create a choropleth map with the function tract_choropleth.

Recall that all choroplethr functions require a dataframe where one column is called “region” and one column is called “value”. get_tract_demographics returns a dataframe with a “region” column and eight demographic values. We still need to create a “value” column. Let’s go with median_rent:


ny_stats$value = ny_stats$median_rent

?tract_choropleth
tract_choropleth(ny_stats, "new york", title = "2013 Median Rent\nCensus Tracts", legend="Dollars")

People not familiar with New York might see the above map and not know where major landmarks are. To solve this problem, all choroplethr functions have a “reference_map” parameter, which puts a google map underneath the choropleth:


tract_choropleth(ny_stats, "new york", title = "2013 Median Rent\nCensus Tracts", legend="Dollars", reference_map = TRUE)

Zooming in

Tract maps of an entire state are hard to view because the tracts are so small. This is why all tract-related functions in choroplethr allow you to zoom by county.

In addition to being useful for viewing maps, the county-zoom option is useful for get_tract_demographics because getting tract-level demographics for an entire state is slow.

Note that counties must be specified by their county FIPS code. Here’s an example of zooming in on Manhattan (FIPS code 36061):


# 36061 is the FIPS code for New York county (i.e. Manhattan)
manhattan_2010 = get_tract_demographics("new york", county_fips=36061, endyear = 2010, span = 5)
manhattan_2010$value = manhattan_2010$median_rent

m1 = tract_choropleth(manhattan_2010, "new york", legend = "Dollars", county_zoom = 36061)
m2 = tract_choropleth(manhattan_2010, "new york", legend = "Dollars", county_zoom = 36061, reference_map = TRUE)

?double_map
double_map(m1, m2, "2010 Median Rent\nManhattan Census Tracts")

New function: double_map

As the above example shows, v3.6.0 introduces a new function: double_map.

In my own work I find that frequently switch between viewing a pure choropleth (where the color contrast is strong) and a choropleth + reference map (where the reference map helps you understand what you’re looking at). double_map simply encapsulates some code for putting two maps side-by-side

Caveats

There are a few things to keep in mind when using this functionality:

get_tract_demographics returns data from the American Community Survey (ACS). The ACS provides estimates with a margin of error (MOE). In the case of tracts, the MOE can be quite large. tract_choropleth does not attempt to visualize the MOE.
get_tract_demographics sometimes produces the warning “NAs introduced by coercion”. I assume that this occurs when the ACS simply does not return an estimated value for a tract. (E.g. a sparsely populated, or unpopulated tract). However, I have not investigated this.
tract_choropleth sometimes produces the warning “Your data.frame contains the following regions which are not mappable”. This is a bit of a mystery for me. It indicates that get_tract_demographics (which gets data from the Census API) contains regions which are not part of a map (that also comes from the Census API).
tract_choropleth sometimes produces the warning “The following regions were missing and are being set to NA”. Presumably these regions are the same regions as in (2).