Saving scripts in R

Generate lecture code examples

Maps in ggplot2

Exercises (maps & data manipulation)

Use the knitr package

# install.packages("knitr")
library("knitr")
purl("https://raw.githubusercontent.com/sebastianbarfort/sds/gh-pages/_slides/lecture3.Rmd")

There are many ways to make maps in R

1. easy: use an existing package
2. hard: learn how to work with shapefiles (we don't have time for this today, but I strongly recommend reading these notes on the topic)

Today focus is on 1.

There are many useful packages for making maps in R

• maps: all kinds of maps
• ggcounty: generate United States county maps
• ggmap: extends ggplot2 for maps
• mapDK: maps of Denmark

Let's return to our marijuana price data

library("readr")
df = read_csv("https://raw.githubusercontent.com/sebastianbarfort/sds/master/data/marijuana-street-price-clean.csv")

Generate yearly state level means

library("lubridate")
library("dplyr")
df$year = year(df$date)

df = df %>% 
  group_by(State, year) %>%
  summarise(
    m.price = mean(HighQ, na.rm = TRUE)
  ) %>%
  mutate(
    region = tolower(State)
  )

The maps package has geographic information on all U.S states

library("maps")
library("ggplot2")
us.states = map_data("state")
head(us.states)

##        long      lat group order  region subregion
## 1 -87.46201 30.38968     1     1 alabama      <NA>
## 2 -87.48493 30.37249     1     2 alabama      <NA>
## 3 -87.52503 30.37249     1     3 alabama      <NA>
## 4 -87.53076 30.33239     1     4 alabama      <NA>
## 5 -87.57087 30.32665     1     5 alabama      <NA>
## 6 -87.58806 30.32665     1     6 alabama      <NA>

Merge the data

df.merge = left_join(df, us.states)

Plotting the dataframe is easy in ggplot2

p = ggplot(df.merge, aes(x = long, y = lat, group = group)) + 
  geom_polygon(aes(fill = m.price)) + 
  facet_wrap( ~ year, ncol = 1) + 
  expand_limits() + 
  theme_minimal()

The ggcounty package provides data at the U.S county level

# devtools::install_github("hrbrmstr/ggcounty")
library("ggcounty")
data(population) # built-in US population by FIPS code data set
population$brk <- cut(population$count, 
                      breaks=c(0, 100, 1000, 10000, 100000, 1000000, 10000000), 
                      labels=c("0-99", "100-1K", "1K-10K", "10K-100K", 
                               "100K-1M", "1M-10M"),
                      include.lowest=TRUE) # define appropriate (& nicely labeled) population breaks
us <- ggcounty.us()
gg <- us$g # start the plot with our base map
gg <- gg + geom_map(data=population, map=us$map,
                    aes(map_id=FIPS, fill=brk), 
                    color="white", size=0.125) # add a new geom with our population (choropleth)
gg <- gg + scale_fill_manual(values=c("#ffffcc", "#c7e9b4", "#7fcdbb", 
                                      "#41b6c4", "#2c7fb8", "#253494"), 
                             name="Population")

ggmap is a package that uses the ggplot2 syntax as a template to create maps with image tiles taken from map servers such as Google and OpenStreetMap

Let's use some data on benches in Copenhagen

df = read_csv("http://wfs-kbhkort.kk.dk/k101/ows?service=WFS&version=1.0.0&request=GetFeature&typeName=k101:baenk&outputFormat=csv&SRSNAME=EPSG:4326")
names(df)

##  [1] "FID"                "wkb_geometry"       "id"                
##  [4] "vejkode"            "vejnavn"            "park_id"           
##  [7] "bydel"              "distrikt"           "baenk_type"        
## [10] "baenk_tilstand"     "baenk_placering"    "baenk_foto"        
## [13] "baenk_driftsopgave" "baenk_fjernet"      "bemaerkning"       
## [16] "reg_metode"         "reg_dato"           "rettet_dato"

We need to do quite some data cleaning

library("dplyr")
library("stringr")
df = df %>%
  select(wkb_geometry, baenk_tilstand) 

# cleaning  
df$wkb_geometry = gsub("\\(|\\)", "", df$wkb_geometry) 
df$wkb_geometry = str_extract(df$wkb_geometry, "[0-9].+")
x = str_split(df$wkb_geometry, pattern  = " ")
x = do.call(rbind.data.frame, x)
df = bind_cols(df, x)
names(df) = c("wbk_geometry", "baenk_tilstand", "lat", "lon")
df$lon = as.numeric(as.character(df$lon))
df$lat = as.numeric(as.character(df$lat))

library("ggmap")
qmplot(lat, lon, zoom = 15, data = df, 
       maptype = "toner-background", color = I("red"))

qmplot(lat, lon, zoom = 15, data = df, 
       maptype = "toner-lite", geom = "density2d", color = I("red"))

A package for making maps of Denmark at different levels of aggregation

Functions

The package currently only has two functions:

• mapDK - makes the map

• getID - prints keys in case you run into merge problems

only accepts one argument: detail

library(mapDK)
args(getID)

## function (detail = "municipal") 
## NULL

getID(detail = "municipal")[1:10]

##  [1] "aabenraa"    "aalborg"     "aeroe"       "albertslund" "alleroed"   
##  [6] "aarhus"      "assens"      "ballerup"    "billund"     "bornholm"

getID(detail = "region")

## [1] "hovedstaden" "midtjylland" "nordjylland" "sjaelland"   "syddanmark"

mapDK takes the following arguments

args(mapDK)

## function (values = NULL, id = NULL, data, detail = "municipal", 
##     show_missing = TRUE, sub = NULL, guide.label = NULL, map.title = NULL) 
## NULL

• For basic maps you really only need detail, sub and map.title

• If you want to do choropleth maps you need to specify

  • data: A data frame of values and ids
    
  • values, id: String variables specifying names of value and id columns in the dataset

• returns a ggplot2 object you can modify if you like

• You control the level of aggregation using the detail argument
  • municipality - plots Denmark's 98 municipalities
  • region - plots Denmark's 5 regions
  • rural - plots Denmark's 11 rural areas
  • zip - plots Denmark's 598 zip code areas
  • polling - plots Denmark's 1385 polling places (as of 2015)
  • parish - plots Denmark's 1931 parishes
• the sub argument takes a vector of strings specifying subregions to be plotted

mapDK()

mapDK(detail = "parish")

mapDK(values = "stemmer", id = "id", 
  data = subset(votes, navn == "socialdemokratiet"),
  detail = "polling", show_missing = FALSE,
  guide.label = "Stemmer \nSocialdemokratiet (pct)")

Putting it all togother…

library("mapproj")
library("ggmap")
df = mapDK::polling
df.votes = mapDK::votes
df = df %>% filter(KommuneNav == "koebenhavn")
df.t = left_join(df, df.votes)
cph.map = ggmap(get_map(location = c(12.57, 55.68), 
                       source = "stamen", 
                       maptype = "toner", crop = TRUE,
                       zoom = 13))
p = cph.map + 
  geom_polygon(data = subset(df.t, navn == "socialdemokratiet"), 
                       aes(x = long, y = lat,
                           group = group, fill = stemmer),
                       alpha = .75) 

For this exercise we will work with data on GDP per capita at the country level. You can download the data using the WDI package as shown below

# install.packages("WDI")
library("WDI")
library("dplyr")

df = WDI(indicator = "NY.GDP.PCAP.KN" ,
         start = 2010, end = 2010, extra = F)
df = df %>% filter(!is.na(NY.GDP.PCAP.KN))

Question 1: use the map package and the GDP data to make a world map of GDP per capita.

Question 2: install the package countrycode and use the countrycode function to add a region indicator to the dataset. Create a world map faceted by your region indicator.

In this exercise you will work with data on votes for the Danish general election from 2011. You can read the data using the following piece of code

df = mapDK::votes

Question 1: use the mapDK package to make a map of votes (in pct) for the Conservative Party ("detkonservativefolkeparti") at the polling place level.

Question 2: read up on the documentation for the dplyr package to aggregate the data into votes (in pct) for the Conservatives at the municipal level. Plot the data using mapDK

Question 3: Repeat question 2 but only for the municipalities "Aarhus" and "Koebenhavn".

For this exercise we will work with Facebook data from the Danish parliamentary election 2015 kindly provided by 56 north.

Load the data by running

library(readr)
df = read_csv("https://raw.githubusercontent.com/sebastianbarfort/sds/master/data/FV15_data.csv")

Question 1: Use the dplyr package to aggregate the number of likes by party and "storkreds"

Question 2: Plot the data (do you need to facet?) on a map using the mapDK package.

Question 3: Use the dplyr package to sort the dataset according to the number of likes. Which candidate in the data is most popular? Create a dataset with only the most popular candidate by "storkreds".