Feature image for post about Ggplot on map of sweden from OpenStreetMap

Simple data mining and plotting data on a map with ggplot2

Introduction

A Facebook group for psychologists in Sweden created a document where each member could type in were they live and what they work with. At the moment 192 psychologists have added their information to the document. This made me think about how to plot this information using R, and especially how to read data in this format. I was also interested in how to plot this information geographically on a map of Sweden and represent the number of individuals by the size of a circle over the corresponding city.

The data

The document with the data was ordered like this:

City
Jane Doe, title, workplace etc. 
John Doe, title, workplace etc. 

City 2
John Doe, title, workplace etc. 
Jane Doe, title, workplace etc. 
…
etc.

Methods

Parsing the data
To get the data into R I copied the Facebook document to a txt-file and used scan() to read it. This saves the text into a vector with each line representing one row. I wrote a function to parse the text and to create a data.frame with the number of people of each city.

Creating the map
To get a map of Sweden I used the OpenStreetMap-package. But since I wanted to plot data on this map I had to grab the coordinates for each city. This was accomplished by using data from the “world.cities”-set which is included in R. However, 5 cities weren’t included in this set so I looked up their coordinates on Google Maps and added them manually. Lastly I converted latitude and longitude into Mercator coordinates since this is the format used when plotting from the OpenStreetMaps-package.

Creating the plot
I used autoplot() to plot the map with ggplot2(). The city data was added as geom_point()-layer with size = number of psychologists in the city. I annotated the the plot using direct.labels().

The resulting plot


Figure 1. Plot of where people live, with data taken from a Facebook document.

The R Code

Sys.setenv(NOAWT=1) #call this before loading OpenStreetMap
library(OpenStreetMap)
library(rgdal)
library(stringr)
library(ggplot2)

setwd("~/Dropbox/R-projekt/Var bor psykologerna")

# Get data for sweden from 'world.cities' dataset
data(world.cities)
sweden.cities <- world.cities[world.cities$country.etc == "Sweden",]

##################
# importing data #
##################

# 'world.cities' didn't have data for all my cities
# so I added long and lat for them in excel and saved
# the new data as city.txt
sweden.cities <- read.delim("city.txt")
# load text file with info about where people live
data.txt <- scan("var.txt", character(0), sep = "\n")
# replace swedish å,ä,ö, since the regular expression
# wouldn't catch them.
data.txt <- gsub('(å|ä)', 'a', data.txt, ignore.case = TRUE)
data.txt <- gsub('ö', 'o', data.txt, ignore.case = TRUE)

###################
# Extracting data #
###################
# Use regular expression to extract rows with only 
# one word, i.e. city headings. Save as boolean. 
txt.pattern <- str_detect(data.txt, "^[[:alpha:]]*$")
# Create a list of city names
cities.txt <- grep("^[[:alpha:]]*$", data.txt, value=T)

# Create empty variables for each city
for (i in 1:length(cities.txt)) {
  assign(cities.txt[i], 0)
}

# Count number of rows after each city heading
# and assign that number to the corresponding city-variable
x <- 0 # row count
y <- 0 # index for cities.txt
for (i in 1:length(data.txt)) {
  if (txt.pattern[i] == TRUE) {
    if (y != 0) assign(cities.txt[y], x)
    x <- 0
    y <- y + 1
  } else x <- x + 1
}

# Combine individual city variables into one data.frame
for (i in 1:length(cities.txt)) {
  cities <- cities.txt[i]
  people <- eval(parse(text = cities))
  if (i == 1) {
    df <- data.frame("cities" = cities, "people" = people)
  } else df <- rbind(data.frame("cities" = cities, "people" = people), df)
}

# remove dummy row 'end'
df <- df[-1,]

# add capital letter for 'gsub'ed words.
df$cities <- as.character(df$cities)
df$cities[1] <- "Ostersund"
df$cities[2] <- "Ornskoldsvik"
df$cities[3] <- "Orebro"

# Get city long/lat data
match <- match(df$cities, sweden.cities$name)
df.coords <- sweden.cities[match,]

# add number of psychologists per city
df.coords <- cbind(df.coords, "people" = df$people)

# convert long/lat to Mercator coordinates
df.coords <- cbind(df.coords,projectMercator(df.coords$lat,df.coords$long))

# clean up workspace
list <- ls()
list <- list[-grep("df.coords", list)] 
rm(list=list)
rm(list)

# get openmap for sweden
mp <- openmap(c(69.2,9.303711), c(54,24.433594), zoom=6, type="osm")

########
# Plot #
########
autoplot(mp) + 
       geom_point(aes(x,y, size=people, color=people),
                  alpha = I(8/10), data=df.coords) +
       opts(axis.line=theme_blank(),axis.text.x=theme_blank(),
       axis.text.y=theme_blank(),axis.ticks=theme_blank(),
       axis.title.x=theme_blank(),
       axis.title.y=theme_blank()) + 
      scale_size_continuous(range= c(1, 25)) +  
       scale_colour_gradient(low="blue", high="red") +
       opts(title="Where do we live?") 
              
# save plot
ggsave("plot.png", dpi=600)

Kristoffer Magnusson

I'm a PhD-student and a clinical psychologist from Sweden with a passion for research and statistics. This is my personal blog about psychological research and statistical programming with R.

Comments (5) Write a comment

  1. Did you make any attempts at trying to use direct.label() with the resulting object from autoplot(mp)? I’m trying to use direct.label to annotate contour lines on top of autoplot(mp) and am having difficulties doing so. Any feedback would be appreciated!

    Reply

  2. Amazing, thank you.
    I should mention that theme_empty() and opts are now deprecated and one can use element_empty() and theme instead.

    Reply

Leave a Reply