See notes.R to follow along in RStudio.
Also see the R input from the lecture.
Assignment 4 is now posted.
A short peer review for assignment 3 will be posted this evening and due in one week.
The Craigslist vehicles data has a structure similar to the Craigslist apartments data on assignment 4.
Each row is one post. For the vehicles data, the posts are about vehicles for sale in the San Francisco Bay Area.
v = readRDS("cl_vehicles.rds")
names(v)
head(v[-2])
v[1, 2]A key feature of this data set is that it includes latitude and longitude, so we can make maps!
There are a lot of packages for making maps in R. Here's a few:
| Package | Description |
|---|---|
| maps | make simple static maps |
| RgoogleMaps | make static maps with backgrounds from Google maps |
| choroplethr | make choropleths (advanced features) |
| cartography | make static maps (advanced features) |
| leafletR | interactive maps with the Leaflet JavaScript library |
| rMaps | interactive maps (still in development) |
Let's start simple.
# install.packages("maps")
library("maps")The map() function is the starting point.
?map
map()The database parameter controls which map is drawn.
The regions parameter controls which regions are drawn.
Use the namesonly parameter to get region names for a particular map. You
might also want to set the plot parameter to FALSE for this.
map("state", "California")
map("county")
map("county", namesonly = TRUE, plot = FALSE)
map("county", "california,sacramento")A map is just a set of shapes (also called polygons).
A region is one or more shapes in a map. Each shape in a region has a
name in the format region:shape or region,shape.
map("county", "california")Use map.text() to add labels. You can use the default labels, or chose
your own.
map.text("county", "california")The add parameter lets you add to the current map.
map("county")
map.text("county", "california", add = TRUE)How can you visualize where the car posts came from?
map("state")
points(v$longitude, v$latitude)
map("county", "california")
points(v$longitude, v$latitude)
map("county", "california")
smoothScatter(v$longitude, v$latitude, 1000, add = TRUE)
map("county", "california", add = TRUE)How can you zoom in on the San Francisco Bay area?
One way is to set xlim and ylim to an appropriate longitude/latitude.
Another way:
map("county", c("california,san francisco", "california,san mateo"))How can you automate typing the county names?
The paste() function glues strings together. Its sep parameter controls
the characters separating each string.
paste("hello", "world", sep = ",")
sfbay_counties = c("san francisco", "san mateo", "marin", "alameda", "contra
costa", "napa", "sonoma", "santa clara", "solano", "yolo", "sacramento")
sfbay = paste("california", sfbay_counties, sep = ",")
map("county", sfbay)
smoothScatter(v$longitude, v$latitude, 3000, add = TRUE)
map("county", sfbay, add = TRUE)A larger nbin setting for smoothScatter() results in a smoother plot,
but also takes more time to draw.
How can you visualize the price of vehicles by county?
First, compute a price for each county!
nrow(v)
by_county = split(v$price, v$shp_county)
sapply(by_county, mean, na.rm = TRUE)
# Same thing:
prices = tapply(v$price, v$shp_county, mean, na.rm = TRUE)Color shapes in a map by assigning a vector to the col parameter.
You also need to set fill to TRUE.
map("county", sfbay, fill = TRUE, col = c("red", "blue"))
counties = map("county", sfbay, namesonly = TRUE, plot = FALSE)You need to match the average prices to map regions.
The names in our price vector don't match!
prices
names(prices)How can you convert the shp_state and shp_county variables to a region
name?
v$map_region = paste(v$shp_state, v$shp_county, sep = ",")
head(v$map_region)
tolower("HELLO world")
v$map_region = tolower(v$map_region)
head(v$map_region)Now compute the average prices again.
prices = tapply(v$price, v$map_region, mean, na.rm = TRUE)
prices
prices = prices[counties]Great, but how can you assign a color to each number?
map("county", sfbay, fill = TRUE, col = prices)Factors behave like colors.
cut(prices, 4)
price_cat = cut(prices, 4, dig.lab = 10)
typeof(price_cat)
levels(price_cat)
map("county", sfbay, fill = TRUE, col = price_cat)
legend("bottomleft", levels(price_cat), fill = 1:4, cex = 1)But now there's two new problems!
# install.packages("viridis")
library("viridis")
pal = magma(4)
# Use the levels (1-4) of the factor to select the colors.
names(price_cat) = sfbay
levels(price_cat) = c("cheap", "mid-cheap", "mid-expensive", "expensive")
price_cat
colors = pal[price_cat]
names(colors) = sfbay
map("county", sfbay, fill = TRUE, col = pal[price_cat])
legend("bottomleft", levels(price_cat), fill = pal, cex = 1)There's a lot of useful functions in the maps package:
| Function | Description |
|---|---|
| map | make a map |
| map.text | make a map with labels |
| map.axes | add axes to a map |
| map.scale | add a scale to a map |
| map.where | identify which map region a point falls into |
| smooth.map | alternative to smoothScatter |
| identify.map | identify map regions by clicking |
The Rgooglemaps package can fetch images from Google Maps or OpenStreetMap.
library("RgoogleMaps")Download a map with the GetMap() function.
The center parameter specifies where the map is centered.
The zoom parameter specifies the zoom level as an integer.
m = GetMap(c(37.78, -122.42), zoom = 8)Use the PlotOnStaticMap() function to display a map.
PlotOnStaticMap(m, v$latitude, v$longitude, cex = 0.15)What if you want to color the points by county? This is a silly example, but it shows how to color the points.
# First subset to the counties of interest.
sf_v = subset(v, map_region %in% sfbay)
# Now convert map_region to a categorical variable.
sf_v$map_region_cat = factor(sf_v$map_region)
levels(sf_v$map_region_cat)
# Now get a color palette with one color for each category.
length(levels(sf_v$map_region_cat))
pal = viridis(11)
# Finally, assign a color to each post, then plot the posts.
sf_v$color = pal[sf_v$map_region_cat]