Leaflet map

m <- leaflet() |>
  addTiles()  # Add default OpenStreetMap map tiles
m

Adding points

m <- leaflet() %>%
  addTiles() %>%  # Add default OpenStreetMap map tiles
  addMarkers(lng=-63.5932, lat=44.63697, popup="Math & Stats, Dalhousie U")
m

Generate some data at random

N <- 10 # no more than 26 please
cities <- tibble(
  lat = runif(N, -80, 80),
  lng = runif(N, -180, 180),
  city = sample(LETTERS, N)
)
m2 <- leaflet(data = cities) %>%
  addTiles() %>%
  addCircleMarkers(~ lng, ~lat, label = ~city)
m2

Heatmaps

Heatmaps can be used to show three variables. If the heatmap is to replace a map, typically one is geographic (shown on the x- or y-axis), one is another categorical variable (such as age group), and the third is a quantitative variable shown in colour brightness.

Here I’ll take the census data by county from the USA in the socviz package (county_data).

I’ll aggregate the population by country region (north, south, etc) and the party that won the vote in the 2016 region.

library(socviz)
s1 <- county_data |> filter(!is.na(census_region), !is.na(partywinner16)) |>
  group_by(partywinner16, census_region) |> summarize(pop = sum(pop))
## `summarise()` has regrouped the output.
## ℹ Summaries were computed grouped by partywinner16 and census_region.
## ℹ Output is grouped by partywinner16.
## ℹ Use `summarise(.groups = "drop_last")` to silence this message.
## ℹ Use `summarise(.by = c(partywinner16, census_region))` for per-operation
##   grouping (`?dplyr::dplyr_by`) instead.
s1 |> ggplot(aes(y = census_region, x = partywinner16, fill = log10(pop))) + geom_tile()

More experiments with the same data.

s2 <- county_data |> filter(!is.na(census_region)) |>
  group_by(pop_dens, census_region) |> summarize(pop = sum(pop), n = n(), land = sum(land_area))
## `summarise()` has regrouped the output.
## ℹ Summaries were computed grouped by pop_dens and census_region.
## ℹ Output is grouped by pop_dens.
## ℹ Use `summarise(.groups = "drop_last")` to silence this message.
## ℹ Use `summarise(.by = c(pop_dens, census_region))` for per-operation grouping
##   (`?dplyr::dplyr_by`) instead.
s2 |> ggplot(aes(y = census_region, x = pop_dens, fill = log10(pop))) + geom_tile() 

s2 |> ggplot(aes(y = census_region, x = pop_dens, fill = n)) + geom_tile() 

s2 |> ggplot(aes(y = census_region, x = pop_dens, fill = land)) + geom_tile() 

s2 |> group_by(census_region) |> mutate(percent = n/sum(n)*100) |>
  ggplot(aes(y = census_region, x = pop_dens, fill = percent)) + geom_tile() 

s2 |> group_by(census_region) |> mutate(percent_pop = pop/sum(pop)*100) |>
  ggplot(aes(y = census_region, x = pop_dens, fill = percent_pop)) + geom_tile()

And more…

s3 <- county_data |> filter(!is.na(census_region), !is.na(partywinner16)) |>
  group_by(partywinner16, state) |> summarize(pop = sum(pop))
## `summarise()` has regrouped the output.
## ℹ Summaries were computed grouped by partywinner16 and state.
## ℹ Output is grouped by partywinner16.
## ℹ Use `summarise(.groups = "drop_last")` to silence this message.
## ℹ Use `summarise(.by = c(partywinner16, state))` for per-operation grouping
##   (`?dplyr::dplyr_by`) instead.
s3 |> ggplot(aes(y = fct_reorder(state, pop), x = partywinner16, fill = log10(pop))) + geom_tile()

USA state map

Many true geographic maps of the USA are strongly influenced by population density:

Sometimes a different geographic scale is warranted such as state or county level political influence.

library(statebins)
s2 <- county_data |> filter(!is.na(state), !is.na(per_dem_2012), !is.na(pop)) |> 
  mutate(state = as.character(state)) |>
  group_by(state) |> 
  summarize(pop_sum = sum(pop),
            percent_dem_2012 = weighted.mean(per_dem_2012, pop)) 
m4 <- s2 %>% ggplot(aes(state = state, fill = log10(pop_sum))) +
  geom_statebins() +
  theme_statebins() +
  labs(fill="log10 population")
m4

m5 <- s2 %>% ggplot(aes(state = state, fill = percent_dem_2012)) +
  geom_statebins() +
  theme_statebins() +
  labs(fill="Percent Democratic vote in 2012")
m5

Maps can easily distort your data and their interpretation. They are best used when the underlying information is truely related to geography and area. Always consider the possible interpretations (and misinterpretations) of a map you draw.

Maps for other regions