11 min read

Map Census Data with R!

This talk was presented on May 30th, 2019 at Code For Pittsburgh.

Before we dive in, this presentation assumes that the user has basic familiarity with tidyverse, mainly dplyr. Knowing how to use %>% will be very helpful.

How to install packages:

install.packages("package_name")

Get your census API key: https://api.census.gov/data/key_signup.html

Configure environment:

library(tidycensus)
library(tidyverse)
library(sf)
library(tigris)
library(ggmap)
library(janitor)

theme_set(theme_bw())

options(tigris_use_cache = TRUE,
        scipen = 4,
        digits = 3)

Packages

{tidycensus}

tidycensus gives access to the Census API and makes it easy to plot data on a map.

Data

  • Demographics
    • Decennial Census
    • American Community Survey (ACS)
    • error estimates
  • Geometries
    • country
    • county
    • zip code
    • blocks
    • tracts
    • and more

{sf}

simple features makes it easy to work with polygon data in R. It uses the familiar tidyverse framework: everything is a tibble, and it uses %>%.

ggplot2::geom_sf() makes it easy to plot sf polygons.

sf can also do spatial calculations such as st_contains, st_intersects, and st_boundary.

{ggmap}

Uses Google Maps API to get basemaps. The API now requires a credit card, but it has a fairly generous “free” tier.

Using {tidycensus}

census_api_key("your_key_here")

This loads the variables from the Decennial Census in 2010:

variables_dec <- load_variables(year = 2010, dataset = "sf1", cache = TRUE)
## # A tibble: 3,346 x 3
##    name    label                                concept         
##    <chr>   <chr>                                <chr>           
##  1 H001001 Total                                HOUSING UNITS   
##  2 H002001 Total                                URBAN AND RURAL 
##  3 H002002 Total!!Urban                         URBAN AND RURAL 
##  4 H002003 Total!!Urban!!Inside urbanized areas URBAN AND RURAL 
##  5 H002004 Total!!Urban!!Inside urban clusters  URBAN AND RURAL 
##  6 H002005 Total!!Rural                         URBAN AND RURAL 
##  7 H002006 Total!!Not defined for this file     URBAN AND RURAL 
##  8 H003001 Total                                OCCUPANCY STATUS
##  9 H003002 Total!!Occupied                      OCCUPANCY STATUS
## 10 H003003 Total!!Vacant                        OCCUPANCY STATUS
## # ... with 3,336 more rows

This loads the ACS variables for 2017:

variables_acs <- load_variables(year = 2017, dataset = "acs5", cache = TRUE)
## # A tibble: 25,070 x 3
##    name      label                         concept                         
##    <chr>     <chr>                         <chr>                           
##  1 B00001_0~ Estimate!!Total               UNWEIGHTED SAMPLE COUNT OF THE ~
##  2 B00002_0~ Estimate!!Total               UNWEIGHTED SAMPLE HOUSING UNITS 
##  3 B01001_0~ Estimate!!Total               SEX BY AGE                      
##  4 B01001_0~ Estimate!!Total!!Male         SEX BY AGE                      
##  5 B01001_0~ Estimate!!Total!!Male!!Under~ SEX BY AGE                      
##  6 B01001_0~ Estimate!!Total!!Male!!5 to ~ SEX BY AGE                      
##  7 B01001_0~ Estimate!!Total!!Male!!10 to~ SEX BY AGE                      
##  8 B01001_0~ Estimate!!Total!!Male!!15 to~ SEX BY AGE                      
##  9 B01001_0~ Estimate!!Total!!Male!!18 an~ SEX BY AGE                      
## 10 B01001_0~ Estimate!!Total!!Male!!20 ye~ SEX BY AGE                      
## # ... with 25,060 more rows

Map total population in the U.S.

Use View() to browse the variables

variables_dec %>% 
  filter(str_detect(concept, "POPULATION")) %>% 
  View()

P001001 has the data we are looking for.

Query the total population of the continental U.S. states:

states <- get_decennial(geography = "state",
                        variables = c(total_pop = "P001001"),
                        geometry = TRUE,
                        output = "wide")

The states tibble contains the census data and the polygons for the geometries.

## Simple feature collection with 52 features and 3 fields
## geometry type:  MULTIPOLYGON
## dimension:      XY
## bbox:           xmin: -179 ymin: 17.9 xmax: 180 ymax: 71.4
## epsg (SRID):    4269
## proj4string:    +proj=longlat +datum=NAD83 +no_defs
## # A tibble: 52 x 4
##    GEOID NAME     total_pop                                        geometry
##    <chr> <chr>        <dbl>                              <MULTIPOLYGON [°]>
##  1 01    Alabama    4779736 (((-85 31, -85 31, -85 31, -85 31, -85 31, -85~
##  2 02    Alaska      710231 (((-165 54.1, -165 54.1, -165 54.1, -165 54.1,~
##  3 04    Arizona    6392017 (((-109 37, -109 37, -109 37, -109 36.9, -109 ~
##  4 05    Arkansas   2915918 (((-94.6 36.5, -94.5 36.5, -94.4 36.5, -94.1 3~
##  5 06    Califor~  37253956 (((-122 37.9, -122 37.9, -122 37.9, -122 37.9,~
##  6 22    Louisia~   4533372 (((-88.9 29.8, -88.9 29.7, -88.9 29.7, -88.9 2~
##  7 21    Kentucky   4339367 (((-83.7 36.6, -83.7 36.6, -83.7 36.6, -83.7 3~
##  8 08    Colorado   5029196 (((-102 37, -102 37, -102 37, -102 37, -102 37~
##  9 09    Connect~   3574097 (((-71.9 41.3, -71.9 41.3, -71.9 41.3, -71.9 4~
## 10 10    Delaware    897934 (((-75.6 39.6, -75.6 39.6, -75.6 39.6, -75.6 3~
## # ... with 42 more rows

Make a bar graph with the data:

states %>% 
  mutate(NAME = fct_reorder(NAME, total_pop)) %>% 
  ggplot(aes(NAME, total_pop)) +
    geom_col() +
    coord_flip()

Plot the same data on a map:

states %>% 
  filter(NAME != "Alaska",
         NAME != "Hawaii",
         !str_detect(NAME, "Puerto")) %>% 
  ggplot(aes(fill = total_pop)) +
    geom_sf() +
    scale_fill_viridis_c("Total Population")

Pull the total population of each county in PA and plot it:

pennsylvania <- get_decennial(geography = "county",
                           variables = c(total_pop = "P001001"),
                           state = "PA",
                           geometry = TRUE,
                           output = "wide")
pennsylvania %>% 
  ggplot(aes(fill = total_pop)) +
    geom_sf() +
    scale_fill_viridis_c()

ggplot2 intelligently handles cases when we don’t have data for a certain polygon:

pennsylvania %>% 
  mutate(total_pop = case_when(NAME == "Allegheny County, Pennsylvania" ~ NA_real_,
                               NAME != "Allegheny County, Pennsylvania" ~ total_pop)) %>% 
  ggplot(aes(fill = total_pop)) +
    geom_sf() +
    scale_fill_viridis_c()

We can stack multiple polygons in the same graph to highlight Allegheny County:

allegheny <- pennsylvania %>% 
  filter(str_detect(NAME, "Allegheny"))

pennsylvania %>% 
  ggplot() +
    geom_sf(aes(fill = total_pop)) +
    geom_sf(data = allegheny, color = "white", linetype = 2, size = 1, alpha = 0) +
    scale_fill_viridis_c()

We can also use tidycensus to download demographic data for census tracts.

Set the variables we want to use first:

racevars <- c(White = "P005003", 
              Black = "P005004", 
              Asian = "P005006", 
              Hispanic = "P004003")
#note that this data is long, not wide
allegheny_tracts <- get_decennial(geography = "tract", variables = racevars, 
                        state = "PA", county = "Allegheny County", geometry = TRUE,
                        summary_var = "P001001") 

Calculate as a percentage of tract population:

allegheny_tracts <- allegheny_tracts %>% 
  mutate(pct = 100 * value / summary_value)

Facet by variable and map the data:

allegheny_tracts %>% 
  ggplot(aes(fill = pct)) +
    geom_sf(color = NA) +
    facet_wrap(~variable) +
    scale_fill_viridis_c()

We can overlay the boundaries of Pittsburgh over the same graph.

Download the boundary shapefile and use sf::st_read to read it into R:

city_pgh <- st_read("data/Pittsburgh_City_Boundary/Pittsburgh_City_Boundary.shp")
## Reading layer `Pittsburgh_City_Boundary' from data source `C:\Users\Conor\Documents\github\blog\content\post\data\Pittsburgh_City_Boundary\Pittsburgh_City_Boundary.shp' using driver `ESRI Shapefile'
## Simple feature collection with 8 features and 6 fields
## geometry type:  POLYGON
## dimension:      XY
## bbox:           xmin: -80.1 ymin: 40.4 xmax: -79.9 ymax: 40.5
## epsg (SRID):    4326
## proj4string:    +proj=longlat +datum=WGS84 +no_defs
allegheny_tracts %>% 
  ggplot() +
    geom_sf(aes(fill = pct), color = NA) +
    geom_sf(data = city_pgh, color = "white", linetype = 2, size = 1, alpha = 0) +
    facet_wrap(~variable) + 
    scale_fill_viridis_c()

Working with other data

WPRDC 311 data and city wards

We can also download the shapefile for the City of Pittsburgh wards. The 311 dataset is tagged with the ward the request originated from, so we can use that to aggregate and map the total number of 311 requests per ward.

df_311 <- read_csv("https://data.wprdc.org/datastore/dump/76fda9d0-69be-4dd5-8108-0de7907fc5a4") %>% 
  clean_names()
request_id created_on request_type request_origin status department neighborhood council_district ward
203364 2017-12-15 14:53:00 Street Obstruction/Closure Call Center 1 DOMI - Permits Central Northside 1 22
200800 2017-11-29 09:54:00 Graffiti Control Panel 1 Police - Zones 1-6 South Side Flats 3 16
201310 2017-12-01 13:23:00 Litter Call Center 1 DPW - Street Maintenance Troy Hill 1 24
200171 2017-11-22 14:54:00 Water Main Break Call Center 0 Pittsburgh Water and Sewer Authority Banksville 2 20
193043 2017-10-12 12:46:00 Guide Rail Call Center 1 DPW - Construction Division East Hills 9 13
196521 2017-10-31 15:17:00 Guide Rail Call Center 1 DPW - Construction Division East Hills 9 13
193206 2017-10-13 09:18:00 Curb /Broken/Deteriorated Call Center 1 DOMI - Permits Mount Washington 2 19
195917 2017-10-27 10:23:00 Manhole Cover Call Center 0 DOMI - Permits Bluff 6 1
179176 2017-08-14 14:00:00 Neighborhood Issues Control Panel 0 NA Middle Hill 6 5
190422 2017-09-29 11:46:00 Mayor’s Office Website 1 311 North Oakland 8 4
## Simple feature collection with 35 features and 7 fields
## geometry type:  POLYGON
## dimension:      XY
## bbox:           xmin: -80.1 ymin: 40.4 xmax: -79.9 ymax: 40.5
## epsg (SRID):    4326
## proj4string:    +proj=longlat +datum=WGS84 +no_defs
## First 10 features:
##    objectid wards wards_id ward wardtext shape_are shape_len
## 1         2     3        0   19       19 0.0011446    0.2389
## 2         3     4        0   24       24 0.0001901    0.0734
## 3         4     5        0   24       24 0.0000210    0.0274
## 4         5     6        0   20       20 0.0011755    0.2888
## 5         6     7        0   21       21 0.0002074    0.0619
## 6         7     8        0   23       23 0.0001031    0.0480
## 7         8     9        0   22       22 0.0001967    0.0593
## 8         9    10        0   26       26 0.0009079    0.1915
## 9        10    11        0   27       27 0.0005849    0.1173
## 10       11    12        0   27       27 0.0000634    0.0400
##                          geometry
## 1  POLYGON ((-80 40.4, -80 40....
## 2  POLYGON ((-80 40.5, -80 40....
## 3  POLYGON ((-80 40.5, -80 40....
## 4  POLYGON ((-80 40.4, -80 40....
## 5  POLYGON ((-80 40.5, -80 40....
## 6  POLYGON ((-80 40.5, -80 40....
## 7  POLYGON ((-80 40.5, -80 40....
## 8  POLYGON ((-80 40.5, -80 40....
## 9  POLYGON ((-80 40.5, -80 40....
## 10 POLYGON ((-80 40.5, -80 40....

Plot the ward polygons:

wards %>% 
  ggplot() +
  geom_sf()

Calculate the center of each ward. We will use this to label the wards on the map:

ward_labels <- wards %>% 
  st_centroid() %>% 
  st_coordinates() %>% 
  as_tibble() %>% 
  clean_names()
x y
-80 40.4
-80 40.5
-80 40.5
-80 40.4
-80 40.5

Count the number of requests per ward:

df_311_count <- df_311 %>% 
  count(ward, sort = TRUE)

Use left_join and bind_cols to join the count data with the coordinates:

ward_311 <- wards %>% 
  left_join(df_311_count) %>%
  bind_cols(ward_labels)
## Simple feature collection with 35 features and 10 fields
## geometry type:  POLYGON
## dimension:      XY
## bbox:           xmin: -80.1 ymin: 40.4 xmax: -79.9 ymax: 40.5
## epsg (SRID):    4326
## proj4string:    +proj=longlat +datum=WGS84 +no_defs
## First 10 features:
##    objectid wards wards_id ward wardtext shape_are shape_len     n
## 1         2     3        0   19       19 0.0011446    0.2389 27136
## 2         3     4        0   24       24 0.0001901    0.0734  5311
## 3         4     5        0   24       24 0.0000210    0.0274  5311
## 4         5     6        0   20       20 0.0011755    0.2888 17263
## 5         6     7        0   21       21 0.0002074    0.0619  3801
## 6         7     8        0   23       23 0.0001031    0.0480  4996
## 7         8     9        0   22       22 0.0001967    0.0593  4606
## 8         9    10        0   26       26 0.0009079    0.1915 10984
## 9        10    11        0   27       27 0.0005849    0.1173 10101
## 10       11    12        0   27       27 0.0000634    0.0400 10101
##                          geometry   x    y
## 1  POLYGON ((-80 40.4, -80 40.... -80 40.4
## 2  POLYGON ((-80 40.5, -80 40.... -80 40.5
## 3  POLYGON ((-80 40.5, -80 40.... -80 40.5
## 4  POLYGON ((-80 40.4, -80 40.... -80 40.4
## 5  POLYGON ((-80 40.5, -80 40.... -80 40.5
## 6  POLYGON ((-80 40.5, -80 40.... -80 40.5
## 7  POLYGON ((-80 40.5, -80 40.... -80 40.5
## 8  POLYGON ((-80 40.5, -80 40.... -80 40.5
## 9  POLYGON ((-80 40.5, -80 40.... -80 40.5
## 10 POLYGON ((-80 40.5, -80 40.... -80 40.5

Plot the data:

ward_311 %>% 
  ggplot() +
    geom_sf(aes(fill = n), color = NA) +
    geom_label(aes(x, y, label = ward), size = 3) +
    scale_fill_viridis_c("Number of 311 requests")

WPRDC overdose data

We can use the census data to adjust other data for per capita rates. For example, the WPRDC’s overdose data has the zip code that the overdose occurred in.

First, download the overdose dataset and pull the population data for each zip code:

df_overdose <- read_csv("https://data.wprdc.org/datastore/dump/1c59b26a-1684-4bfb-92f7-205b947530cf") %>% 
  clean_names() %>% 
  mutate(incident_zip = str_sub(incident_zip, 1, 5))

all_zips <- get_acs(geography = "zip code tabulation area",
                    variables = c(total_pop = "B01003_001"),
                    geometry = TRUE,
                    output = "wide")

Then, aggregate the overdose data to the zip code and join the datasets:

df_overdose <- df_overdose %>% 
  count(incident_zip, sort = TRUE)

attempt1 <- all_zips %>%
  semi_join(df_overdose, by = c("GEOID" = "incident_zip")) %>% 
  left_join(df_overdose, by = c("GEOID" = "incident_zip"))

attempt1 %>% 
  ggplot() +
    geom_sf()

Unfortunately the data is kind of messy and includes zip codes that aren’t in Allegheny County.

We can use st_intersection to exclude all of the zip code polygons that do not fall within the allegheny county tibble we made earlier:

allegheny %>% 
  ggplot() +
    geom_sf()

Then, join the aggregated overdose data with left_join:

df_allegheny_overdose <- st_intersection(allegheny, all_zips) %>% 
  left_join(df_overdose, by = c("GEOID.1" = "incident_zip"))

Now we can calculate the per 1,000 overdose rate and plot the data:

df_allegheny_overdose %>% 
  filter(total_popE >= 400) %>% 
  mutate(overdoses_per_capita = n / total_popE * 1000) %>% 
  ggplot(aes(fill = overdoses_per_capita)) +
    geom_sf(color = NA) +
    scale_fill_viridis_c()

{ggmap} basemaps

We can use ggmap to request a basemap from the Google Maps API. Get your API key here

register_google(key = "Your key here")
pgh_map <- get_map(location = "Pittsburgh, PA", zoom = 12)

ggmap(pgh_map)

There are multiple basemap styles available:

get_map(location = "Pittsburgh, PA", zoom = 12, maptype = "satellite", source = "google") %>% 
  ggmap()

get_map(location = "Pittsburgh, PA", zoom = 12, maptype = "roadmap", source = "google") %>% 
  ggmap()

get_map(location = "Pittsburgh, PA", zoom = 12, maptype = "watercolor", source = "google") %>% 
  ggmap()

get_map(location = "Pittsburgh, PA", zoom = 12, maptype = "toner", source = "stamen") %>% 
  ggmap()

Combining maps from different systems requires us to use the same map projection. Google uses 4326. Use coord_sf to set the projection:

ggmap(pgh_map) +
  geom_sf(data = ward_311, aes(fill = n), inherit.aes = FALSE, color = NA, alpha = .7) +
  geom_label(data = ward_311, aes(x, y, label = ward), size = 3) +
  coord_sf(crs = st_crs(4326)) +
  scale_fill_viridis_c("Number of 311 requests")