jimjam-slam/absmapsdata

Use ABS ASGS data easily in R

absmapsdata

The absmapsdata package exists to make it easier to produce maps from ABS data in R. The package contains compressed, tidied, and lazily-loadable sf objects that hold geometric information about ABS data structures.

It also contains a vast number of 2016 population-weighted ABS correspondences (the most recent) that you can access with the get_correspondence_absmaps function. The correspondences available can be found at the data.gov.au website.

Before we get into the ‘what problem is this package solving’ details, let’s look at some examples so that you can copy-paste into your own script and replicate out-of-the-box (and impress your friends).

Installation

You can install absmapsdata from github with:

# install.packages("remotes")
remotes::install_github("wfmackey/absmapsdata")

absmapsdata contains a lot of data, so installing using remotes::install_github may fail if the download times out. If this happens, set the timeout option to a large value and try again, i.e. run:

options(timeout=1000)
remotes::install_github("wfmackey/absmapsdata")

Note: to download and read absmapsdata files without installing the whole absmapsdata package, please see strayr::read_absmaps.

The sf package is required to handle the sf objects:

# install.packages("sf")
library(sf)

Maps loaded with this package

Available maps are listed below. These will be added to over time. If you would like to request a map to be added, let me know via an issue on this Github repo.

ASGS Main Structures

• Statistical Area 1 2011: sa12011; 2016: sa12016; and 2021: sa12021.
• Statistical Area 2 2011: sa22011; 2016: sa22016; and 2021: sa22021.
• Statistical Area 3 2011: sa32011; 2016: sa32016; and 2021: sa32021.
• Statistical Area 4 2011: sa42011; 2016: sa42016; and 2021: sa42021.
• Greater Capital Cities 2011: gcc2011; 2016: gcc2016; and 2021: gcc2021.
• Remoteness Areas 2011: ra2011; and 2016: ra2016
• State 2011: state2011; 2016: state2016; and state2021.

ASGS Non-ABS Structures

• Commonwealth Electoral Divisions 2018: ced2018
• State Electoral Divisions 2018:sed2018
• Local Government Areas 2016: lga2016; and 2018: lga2018
• Regions for the Internet Vacancy Index 2008: regional_ivi2008
• Postcodes 2016: postcodes2016
• Census of Population and Housing Destination Zones 2011: dz2011; and 2016: dz2016.

Non-ABS Australian Government Structures

• Employment Regions 2015-2020: employment_regions2015

Just show me how to make a map with this package

Using the package’s pre-loaded data

The absmapsdata package comes with pre-downloaded and pre-processed data. To load a particular geospatial object: load the package, then call the object (see list above for object names).

library(tidyverse)
#> ── Attaching packages ─────────────────────────────────────── tidyverse 1.3.1 ──
#> ✓ ggplot2 3.3.5     ✓ purrr   0.3.4
#> ✓ tibble  3.1.4     ✓ dplyr   1.0.7
#> ✓ tidyr   1.1.3     ✓ stringr 1.4.0
#> ✓ readr   2.0.0     ✓ forcats 0.5.1
#> ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
#> x dplyr::filter() masks stats::filter()
#> x dplyr::lag()    masks stats::lag()
library(sf)
#> Linking to GEOS 3.8.1, GDAL 3.2.1, PROJ 7.2.1
library(absmapsdata)

mapdata1 <- sa32011

glimpse(mapdata1)
#> Rows: 351
#> Columns: 12
#> $ sa3_code_2011   <chr> "10101", "10102", "10103", "10104", "10201", "10202", …
#> $ sa3_name_2011   <chr> "Goulburn - Yass", "Queanbeyan", "Snowy Mountains", "S…
#> $ sa4_code_2011   <chr> "101", "101", "101", "101", "102", "102", "103", "103"…
#> $ sa4_name_2011   <chr> "Capital Region", "Capital Region", "Capital Region", …
#> $ gcc_code_2011   <chr> "1RNSW", "1RNSW", "1RNSW", "1RNSW", "1GSYD", "1GSYD", …
#> $ gcc_name_2011   <chr> "Rest of NSW", "Rest of NSW", "Rest of NSW", "Rest of …
#> $ state_code_2011 <chr> "1", "1", "1", "1", "1", "1", "1", "1", "1", "1", "1",…
#> $ state_name_2011 <chr> "New South Wales", "New South Wales", "New South Wales…
#> $ albers_sqkm     <dbl> 21236.61396, 6511.12140, 14281.83009, 9864.93971, 988.…
#> $ cent_long       <dbl> 149.0763, 149.6013, 148.9416, 149.8063, 151.2182, 151.…
#> $ cent_lat        <dbl> -34.55399, -35.44940, -36.43958, -36.49934, -33.36542,…
#> $ geometry        <MULTIPOLYGON [°]> MULTIPOLYGON (((149.1198 -3..., MULTIPOLY…

Or

mapdata2 <- sa22016

glimpse(mapdata2)
#> Rows: 2,310
#> Columns: 15
#> $ sa2_code_2016   <chr> "101021007", "101021008", "101021009", "101021010", "1…
#> $ sa2_5dig_2016   <chr> "11007", "11008", "11009", "11010", "11011", "11012", …
#> $ sa2_name_2016   <chr> "Braidwood", "Karabar", "Queanbeyan", "Queanbeyan - Ea…
#> $ sa3_code_2016   <chr> "10102", "10102", "10102", "10102", "10102", "10102", …
#> $ sa3_name_2016   <chr> "Queanbeyan", "Queanbeyan", "Queanbeyan", "Queanbeyan"…
#> $ sa4_code_2016   <chr> "101", "101", "101", "101", "101", "101", "101", "101"…
#> $ sa4_name_2016   <chr> "Capital Region", "Capital Region", "Capital Region", …
#> $ gcc_code_2016   <chr> "1RNSW", "1RNSW", "1RNSW", "1RNSW", "1RNSW", "1RNSW", …
#> $ gcc_name_2016   <chr> "Rest of NSW", "Rest of NSW", "Rest of NSW", "Rest of …
#> $ state_code_2016 <chr> "1", "1", "1", "1", "1", "1", "1", "1", "1", "1", "1",…
#> $ state_name_2016 <chr> "New South Wales", "New South Wales", "New South Wales…
#> $ areasqkm_2016   <dbl> 3418.3525, 6.9825, 4.7634, 13.0034, 3054.4099, 13.6789…
#> $ cent_long       <dbl> 149.7932, 149.2328, 149.2255, 149.2524, 149.3911, 149.…
#> $ cent_lat        <dbl> -35.45508, -35.37590, -35.35103, -35.35520, -35.44408,…
#> $ geometry        <MULTIPOLYGON [°]> MULTIPOLYGON (((149.7606 -3..., MULTIPOLY…

The resulting sf object contains one observation per area (in the following examples, one observation per sa3). It stores the geometry information in the geometry variable, which is a nested list describing the area’s polygon. The object can be joined to a standard data.frame or tibble and can be used with dplyr functions.

Creating maps with your sf object

We do all this so we can create gorgeous maps. And with the sf object in hand, plotting a map via ggplot and geom_sf is simple.

map <-
sa32016 %>%
  filter(gcc_name_2016 == "Greater Melbourne") %>%   # let's just look Melbourne
  ggplot() +
  geom_sf(aes(geometry = geometry))  # use the geometry variable

map

The data also include centroids of each area, and we can add these points to the map with the cent_lat and cent_long variables using geom_point.

map <- sa32016 %>%
  filter(gcc_name_2016 == "Greater Melbourne") %>%   # let's just look Melbourne
  ggplot() +
  geom_sf(aes(geometry = geometry)) +   # use the geometry variable
  geom_point(aes(cent_long, cent_lat))  # use the centroid long (x) and lats (y)

map

Cool. But this all looks a bit ugly. We can pretty it up using ggplot tweaks. See the comments on each line for its objective. Also note that we’re filling the areas by their areasqkm size, another variable included in the sf object (we’ll replace this with more interesting data in the next section).

map <- sa32016 %>%
  filter(gcc_name_2016 == "Greater Melbourne") %>%   # let's just look Melbourne
  ggplot() +
  geom_sf(aes(geometry = geometry,  # use the geometry variable
              fill = areasqkm_2016),     # fill by area size
          lwd = 0,                  # remove borders
          show.legend = FALSE) +    # remove legend
  geom_point(aes(cent_long,
                 cent_lat),        # use the centroid long (x) and lats (y)
             colour = "white") +    # make the points white
  theme_void() +                    # clears other plot elements
  coord_sf()

map

Joining with other datasets

At some point, we’ll want to join our spatial data with data-of-interest. The variables in our mapping data—stating the numeric code and name of each area and parent area—will make this relatively easy.

For example: suppose we had a simple dataset of median income by SA3 over time.

# Read data in some data
income <- read_csv("https://raw.githubusercontent.com/wfmackey/absmapsdata/master/img/data/median_income_sa3.csv")
#> Rows: 2148 Columns: 3
#> ── Column specification ────────────────────────────────────────────────────────
#> Delimiter: ","
#> chr (2): sa3_name_2016, year
#> dbl (1): median_income
#> 
#> ℹ Use `spec()` to retrieve the full column specification for this data.
#> ℹ Specify the column types or set `show_col_types = FALSE` to quiet this message.
head(income)
#> # A tibble: 6 × 3
#>   sa3_name_2016       year    median_income
#>   <chr>               <chr>           <dbl>
#> 1 Queanbeyan          2010-11         51858
#> 2 Snowy Mountains     2010-11         35884
#> 3 South Coast         2010-11         30908
#> 4 Goulburn - Mulwaree 2010-11         38269
#> 5 Young - Yass        2010-11         39489
#> 6 Gosford             2010-11         38189

This income data contains a variable sa3_name_2016, and we can use dplyr::left_join() to combine with our mapping data.

combined_data <- left_join(income, 
                           sa32016, 
                           by = "sa3_name_2016")

Now that we have a tidy dataset with 1) the income data we want to plot, and 2) the geometry of the areas, we can plot income by area:

map <- combined_data %>%
  filter(gcc_name_2016 == "Greater Melbourne") %>%   # let's just look Melbourne
  ggplot() +
  geom_sf(aes(geometry = geometry,   # use the geometry variable
              fill = median_income), # fill by unemployment rate
          lwd = 0) +                 # remove borders
  theme_void() +                     # clears other plot elements
  labs(fill = "Median income")

map

Get correspondence files

You can use the get_correspondence_absmaps function to get population-weighted correspondence tables provided by the ABS. Note that while there are lots of correspondence tables, not every combination is available.

For example:

get_correspondence_absmaps("cd", 2006,
                           "sa1", 2016)
#> # A tibble: 92,336 × 5
#>    CD_CODE_2006 SA1_MAINCODE_2016 SA1_7DIGITCODE_2016  ratio PERCENTAGE        
#>    <chr>        <chr>             <chr>                <dbl> <chr>             
#>  1 1010101      10902117908       1117908             0.477  47.705709900000002
#>  2 1010101      10902117909       1117909             0.486  48.579130499999998
#>  3 1010101      10902117910       1117910             0.0372 3.7151597000000001
#>  4 1010102      10902117907       1117907             0.210  21.012930999999998
#>  5 1010102      10902117908       1117908             0.281  28.062155199999999
#>  6 1010102      10902117910       1117910             0.509  50.924913799999999
#>  7 1010103      10902117907       1117907             1      100               
#>  8 1010104      10902117901       1117901             0.510  51.007496400000001
#>  9 1010104      10902117907       1117907             0.490  48.992503599999999
#> 10 1010105      10902117907       1117907             1      100               
#> # … with 92,326 more rows

Why does this package exist?

The motivation for this package is that maps are cool and fun and are, sometimes, the best way to communicate data. And making maps is R with ggplot is relatively easy when you have the right object.

Getting the right object is not technically difficult, but requires research into the best-thing-to-do at each of the following steps:

• Find the ASGS ABS spatial-data page and determine the right file to download.
• Read the shapefile into R using one-of-many import tools.
• Convert the object into something usable.
• Clean up any inconsistencies and apply consistent variable naming/values across areas and years.
• Find an appropriate compression function and level to optimise output.

For me at least, finding the correct information and developing the best set of steps was a little bit interesting but mostly tedious and annoying. The absmapsdata package holds this data for you, so you can spend more time making maps, and less time on Stack Overflow, the ABS website, and lovely-people’s wonderful blogs.

Comments/complaints/requests/THOUGHTS

Fair enough! The best avenue is via a Github issue at wfmackey/absmapsdata/issues. This is also the best place to request data that isn’t yet available in the package.