Data Visualization and Exploration

maps!

Introduction

Mapmaking

It’s a map, how much more is there to say?

Turns out, a lot.

  • Maps are “familiar”.
  • But, can be misleading.

What are we looking at?

Don’t forget about colors (1/4)

A lightly edited heatmap (shown with simulated protanopia) from <xkcd.com> showing the role of population density. A stick figure points at three maps.
Figure 3: https://xkcd.com/1138/ with simulated protanopia.

Don’t forget about colors (2/4)

A lightly edited heatmap from <xkcd.com> showing the role of population density. A stick figure points at three maps.

https://xkcd.com/1138/
A lightly edited heatmap (shown with simulated deuteranopia) from <xkcd.com> showing the role of population density. A stick figure points at three maps.
Figure 4: https://xkcd.com/1138/ with simulated deuteranopia.

Don’t forget about colors (3/4)

A lightly edited heatmap from <xkcd.com> showing the role of population density. A stick figure points at three maps.

https://xkcd.com/1138/
A lightly edited heatmap (shown with simulated Tritanopia) from <xkcd.com> showing the role of population density. A stick figure points at three maps.
Figure 5: https://xkcd.com/1138/ with simulated tritanopia

Don’t forget about colors (4/4)

A lightly edited heatmap from <xkcd.com> showing the role of population density. A stick figure points at three maps.

https://xkcd.com/1138/
A lightly edited heatmap (shown with simulated typical monochromacy) from <xkcd.com> showing the role of population density. A stick figure points at three maps.
Figure 6: https://xkcd.com/1138/ with simulated typical monochromacy.

Maps

There are a variety of tools available, but a lot of new, discipline-specific terminology to grapple with.

We will look at a few tools, but, depending on your needs, a reasonable approach is to

  • start with an online tutorial,
  • reproduce the results, and
  • slowly modify the input data (and variable names) to match your current needs.

If you find a good tutorial, use it, share it, comment on it, and ideally archive it so it is more likely to be available to future use.

Simple map

To make simple maps, the packages "maps", "mapproj", and "mapdata" may be useful, but, "mapdata" is no longer current.

We will start “simple”, perhaps disappointingly simple, because this gets quite complicated quite quickly!

options(repos=structure(c("https://cloud.r-project.org", "http://www.stats.ox.ac.uk/pub/RWin" ), .Names = c("CRAN", "CRANextra")))
if(!"maps" %in% names(installed.packages)){install.packages("maps")}

The downloaded binary packages are in
    /var/folders/cv/57f7pbds3y7_pq476q9b438r0000gn/T//RtmpZPzQYG/downloaded_packages
if(!"mapproj" %in% names(installed.packages)){install.packages("mapproj")}

The downloaded binary packages are in
    /var/folders/cv/57f7pbds3y7_pq476q9b438r0000gn/T//RtmpZPzQYG/downloaded_packages
if(!"mapdata" %in% names(installed.packages)){install.packages("mapdata")}

The downloaded binary packages are in
    /var/folders/cv/57f7pbds3y7_pq476q9b438r0000gn/T//RtmpZPzQYG/downloaded_packages
## for basic mapping
library(maps)
library(mapproj)
## for too many other things to list
library(ggplot2)

A series of maps

From a world map, to a US state map, to a US county map, to an Oklahoma county map.

par(mfrow = c(2, 2), mar = rep(0, 4))
map()
map(database = 'state')
map(database = 'county')
map(database = 'county', region = 'oklahoma')
Figure 7: A collection of basic maps.

That’s a wrap!

We can change the center of the map by specifying a 360-degree wrapping. In a world map, numbers must represent the full 360-degree range of longitudes.

par(mfrow = c(2, 2), mar = rep(0, 4))
map(); map.axes(las = 1)
map(wrap = c(0, 360)); map.axes(las = 1)
map(wrap = c(-90, 270)); map.axes(las = 1)
map(wrap = c(-180, 180)); map.axes(las = 1)
Figure 8

Map annotation

map(database = 'state')
map.axes(las = 1)
map.cities(pch = 19)

A collection of basic maps.

A collection of basic maps.

Map “calculations”

m <- map(database = 'state', fill = TRUE, plot = FALSE)
area.map(m, ".*virginia")
.*virginia 
  17.06325 
area.map(m, c("West Virginia", "Virginia"))
West Virginia      Virginia 
     6.515311     10.547939 
area.map(m, c("West Virginia", "Virginia"))
West Virginia      Virginia 
     6.515311     10.547939

Sum to 17.06325.

Terminology

Below the color scheme is rather uninformative, but as we learn why we might color maps, we can make it more useful.

chloropleth - regions of maps colored or shaded by value (quantity or intensity)

map('state', resolution = 0, fill = TRUE, col = hcl.colors(50))
Figure 9

cartograms - regions scaled (in size) by value (here area)

map('state.carto', resolution = 0, fill = TRUE, col = hcl.colors(50))
Figure 10

Chloropleths vs Cartograms

Chloropleths can be misleading.

  • At first glance, we conflate area with importance.
  • Choice of geographic boundary can hide or inflate patterns.

Cartograms can be confusing.

  • They represent space, but aren’t “maps”.
  • May require more accompanying text to guide interpretation.

Map projections

See https://en.wikipedia.org/wiki/List_of_map_projections among other good sources.

par(mfrow = c(2, 2), mar = rep(0, 4))
map(database = "state", projection = "mercator")#; map.axes(las = 1)
map(database = "state", projection = "gnomonic")#; map.axes(las = 1)
map(database = "state", projection = "orthographic")#; map.axes(las = 1)
map(database = "state", projection = "albers", par=c(39, 45))#; map.axes(las = 1)
Figure 11: Map projections: Mercator, Gnomonic, Orthographic, Albers.

Computation on maps

  • Suppose we were interested in extracting simple data from maps.
  • We will get warnings as these are not true areas (making round things flat).
m <- map(database = 'state', fill = TRUE, plot = FALSE)
area.map(m)
      . 
816.295 
area.map(m, ".*virginia")
.*virginia 
  17.06325 
area.map(m, c("West Virginia", "Virginia"))
West Virginia      Virginia 
     6.515311     10.547939

All things considered

Not all “spatial” data needs to be shown as a map!

dat <- read.delim("./data/drink-names.csv", header = TRUE, sep = '\t')
dat <- dat[-1, ]
ggplot(data = subset(dat, Percent > 1), 
       mapping = aes(x = Percent, y = reorder(Region, Percent))) + geom_point()
Figure 12: A dotplot of percent of population to respond.

Data on maps

Suppose we wanted to color states by their response rate in the “pop”-“soda”-“coke” survey.

We will use a US-“standard” Albers map projection.

Turns out due to the complexity of map data, this is very challenging in base R.

To ggplot

We will begin with map_data() a ggplot2 function.

usStates <- map_data("state")
head(usStates, n = 2)
       long      lat group order  region subregion
1 -87.46201 30.38968     1     1 alabama      <NA>
2 -87.48493 30.37249     1     2 alabama      <NA>
dim(usStates)
[1] 15537     6

What is all of that stuff?!?

To ggplot

Having read the data, we can make a plain old “base R” plot.

par(mar = c(4.1, 4.1, 0.1, 0.1))
plot(usStates[, c("long", "lat")])
Figure 13: a map

A map is just a bunch of lines.

Coloring regions

Below there is no rationale to the color scheme.

ggplot(data = usStates, 
       mapping = aes(x = long, y = lat,  
       group = group, fill = region)) + 
       geom_polygon(color = "gray90") + 
       guides(fill = FALSE)
Figure 14: a map

Improving projection

Below there is no rationale to the color scheme.

ggplot(data = usStates, 
  mapping = aes(x = long, y = lat,  
  group = group, fill = region)) + 
  geom_polygon(color = "gray90") + 
  coord_map(projection = "albers", lat0 = 39, lat1 = 45) + 
  guides(fill = FALSE)
Figure 15: a map

Coloring by a data value

Below there is no rationale to the color scheme.

dat$Region <- tolower(dat$Region)
usStates <- merge(usStates, dat, by.x = "region", by.y = "Region")
ggplot(data = usStates, 
  mapping = aes(x = long, y = lat,  
  group = group, fill = Percent)) + 
  geom_polygon(color = "gray90", linewidth = 0.1, show.legend = TRUE) + 
  coord_map(projection = "albers", lat0 = 39, lat1 = 45) + 
  guides(fill = FALSE)
Figure 16: a map

Binned percents as colors

We can bin percents in to one-percent-width bins. It is worth exploring what cut() does out of context of this application.

usStates$bin <- as.numeric(cut(usStates$Percent, c(0, 1, 2, 3, 4, 5, 100)))
ggplot(data = usStates, 
  mapping = aes(x = long, y = lat,  
  group = group, fill = bin)) + 
  geom_polygon(color = "gray90", linewidth = 0.1, show.legend = TRUE) + 
  coord_map(projection = "albers", lat0 = 39, lat1 = 45) + 
  guides(fill = FALSE)
Figure 17: a map

Back to our comic…

ggplot(data = usStates, 
  mapping = aes(x = long, y = lat,  
  group = group, fill = Total)) + 
  geom_polygon(color = "gray90", linewidth = 0.1, show.legend = TRUE) + 
  coord_map(projection = "albers", lat0 = 39, lat1 = 45) + 
  guides(fill = FALSE)
Figure 18: a map

Back to our comic…

ggplot(data = usStates, 
  mapping = aes(x = long, y = lat,  
  group = group, fill = bin)) + 
  geom_polygon(color = "gray90", linewidth = 0.1) + 
  coord_map(projection = "albers", lat0 = 39, lat1 = 45) + 
  guides(fill = FALSE)
Figure 19: a map

Challenges

  • Create new variables to represent the choice of beverage vernacular (e.g., fraction “pop”).
  • Search the internet for similar regional datasets (US or abroad).
  • Compute the state centroid (manually or cleverly) and plot the percent to respond.
  • Consider faceting by beverage type (might require work).
  • Figure out the darn legend bar.