Tutorial (vignette) for the eurostat R package

R Tools for Eurostat Open Data

This rOpenGov R package provides tools to access Eurostat database, which you can also browse on-line for the data sets and documentation. For contact information and source code, see the package website.

# Global options
library(knitr)
opts_chunk$set(fig.path="fig/")

Installation

Release version (CRAN):

install.packages("eurostat")

Development version (Github):

library(devtools)
install_github("ropengov/eurostat")
library(eurostat)

Overall, the eurostat package includes the following functions:

cat(paste0(library(help = "eurostat")$info[[2]], collapse = "\n"))

Finding data

Function get_eurostat_toc() downloads a table of contents of eurostat datasets. The values in column 'code' should be used to download a selected dataset.

# Load the package
library(eurostat)
library(rvest)

# Get Eurostat data listing
toc <- get_eurostat_toc()

# Check the first items
library(knitr)
kable(head(toc))

With search_eurostat() you can search the table of contents for particular patterns, e.g. all datasets related to passenger transport. The kable function to produces nice markdown output. Note that with the type argument of this function you could restrict the search to for instance datasets or tables.

# info about passengers
kable(head(search_eurostat("passenger transport")))

Codes for the dataset can be searched also from the Eurostat database. The Eurostat database gives codes in the Data Navigation Tree after every dataset in parenthesis.

Downloading data

The package supports two of the Eurostats download methods: the bulk download facility and the Web Services' JSON API. The bulk download facility is the fastest method to download whole datasets. It is also often the only way as the JSON API has limitation of maximum 50 sub-indicators at a time and whole datasets usually exceeds that. To download only a small section of the dataset the JSON API is faster, as it allows to make a data selection before downloading.

A user does not usually have to bother with methods, as both are used via main function get_eurostat(). If only the table id is given, the whole table is downloaded from the bulk download facility. If also filters are defined the JSON API is used.

Here an example of indicator 'Modal split of passenger transport'. This is the percentage share of each mode of transport in total inland transport, expressed in passenger-kilometres (pkm) based on transport by passenger cars, buses and coaches, and trains. All data should be based on movements on national territory, regardless of the nationality of the vehicle. However, the data collection is not harmonized at the EU level.

Pick and print the id of the data set to download:

# For the original data, see
# http://ec.europa.eu/eurostat/tgm/table.do?tab=table&init=1&plugin=1&language=en&pcode=tsdtr210
id <- search_eurostat("Modal split of passenger transport", 
                         type = "table")$code[1]
print(id)

Get the whole corresponding table. As the table is annual data, it is more convient to use a numeric time variable than use the default date format:

dat <- get_eurostat(id, time_format = "num")

Investigate the structure of the downloaded data set:

str(dat)
kable(head(dat))

Or you can get only a part of the dataset by defining filters argument. It should be named list, where names corresponds to variable names (lower case) and values are vectors of codes corresponding desidered series (upper case). For time variable, in addition to a time, also a sinceTimePeriod and a lastTimePeriod can be used.

dat2 <- get_eurostat(id, filters = list(geo = c("EU28", "FI"), lastTimePeriod=1), time_format = "num")
kable(dat2)

Replacing codes with labels

By default variables are returned as Eurostat codes, but to get human-readable labels instead, use a type = "label" argument.

datl2 <- get_eurostat(id, filters = list(geo = c("EU28", "FI"), 
                                         lastTimePeriod = 1), 
                      type = "label", time_format = "num")
kable(head(datl2))

Eurostat codes in the downloaded data set can be replaced with human-readable labels from the Eurostat dictionaries with the label_eurostat() function.

datl <- label_eurostat(dat)
kable(head(datl))

The label_eurostat() allows conversion of individual variable vectors or variable names as well.

label_eurostat_vars(names(datl))

Vehicle information has 3 levels. You can check them now with:

levels(datl$vehicle)

Selecting and modifying data

EFTA, Eurozone, EU and EU candidate countries

To facilitate smooth visualization of standard European geographic areas, the package provides ready-made lists of the country codes used in the eurostat database for EFTA (efta_countries), Euro area (ea_countries), EU (eu_countries) and EU candidate countries (eu_candidate_countries). These can be used to select specific groups of countries for closer investigation. For conversions with other standard country coding systems, see the countrycode R package. To retrieve the country code list for EFTA, for instance, use:

data(efta_countries)
kable(efta_countries)

EU data from 2012 in all vehicles:

dat_eu12 <- subset(datl, geo == "European Union (current composition)" & time == 2012)
kable(dat_eu12, row.names = FALSE)

EU data from 2000 - 2012 with vehicle types as variables:

Reshaping the data is best done with spread() in tidyr.

library("tidyr")
dat_eu_0012 <- subset(dat, geo == "EU28" & time %in% 2000:2012)
dat_eu_0012_wide <- spread(dat_eu_0012, vehicle, values)
kable(subset(dat_eu_0012_wide, select = -geo), row.names = FALSE)

Train passengers for selected EU countries in 2000 - 2012

dat_trains <- subset(datl, geo %in% c("Austria", "Belgium", "Finland", "Sweden")
                     & time %in% 2000:2012 
                     & vehicle == "Trains")

dat_trains_wide <- spread(dat_trains, geo, values) 
kable(subset(dat_trains_wide, select = -vehicle), row.names = FALSE)

Visualization

Visualizing train passenger data with ggplot2:

library(ggplot2)
p <- ggplot(dat_trains, aes(x = time, y = values, colour = geo)) 
p <- p + geom_line()
print(p)

Triangle plot

Triangle plot is handy for visualizing data sets with three variables.

library(tidyr)
library(plotrix)
library(eurostat)
library(dplyr)
library(tidyr)

# All sources of renewable energy are to be grouped into three sets
 dict <- c("Solid biofuels (excluding charcoal)" = "Biofuels",
 "Biogasoline" = "Biofuels",
 "Other liquid biofuels" = "Biofuels",
 "Biodiesels" = "Biofuels",
 "Biogas" = "Biofuels",
 "Hydro power" = "Hydro power",
 "Tide, Wave and Ocean" = "Hydro power",
 "Solar thermal" = "Wind, solar, waste and Other",
 "Geothermal Energy" = "Wind, solar, waste and Other",
 "Solar photovoltaic" = "Wind, solar, waste and Other",
 "Municipal waste (renewable)" = "Wind, solar, waste and Other",
 "Wind power" = "Wind, solar, waste and Other",
 "Bio jet kerosene" = "Wind, solar, waste and Other")
# Some cleaning of the data is required
 energy3 <- get_eurostat("ten00081") %>%
 label_eurostat(dat) %>%
 filter(time == "2013-01-01",
 product != "Renewable energies") %>%
 mutate(nproduct = dict[as.character(product)], # just three categories
 geo = gsub(geo, pattern=" \\(.*", replacement="")) %>%
 select(nproduct, geo, values) %>%
 group_by(nproduct, geo) %>%
 summarise(svalue = sum(values)) %>%
 group_by(geo) %>%
 mutate(tvalue = sum(svalue),
 svalue = svalue/sum(svalue)) %>%
 filter(tvalue > 1000) %>% # only large countries
 spread(nproduct, svalue)

# Triangle plot
 par(cex=0.75, mar=c(0,0,0,0))
 positions <- plotrix::triax.plot(as.matrix(energy3[, c(3,5,4)]),
                     show.grid = TRUE,
                     label.points= FALSE, point.labels = energy3$geo,
                     col.axis="gray50", col.grid="gray90",
                     pch = 19, cex.axis=0.8, cex.ticks=0.7, col="grey50")

 # Larger labels
 ind <- which(energy3$geo %in%  c("Norway", "Iceland","Denmark","Estonia", "Turkey", "Italy", "Finland"))
 df <- data.frame(positions$xypos, geo = energy3$geo)
 points(df$x[ind], df$y[ind], cex=2, col="red", pch=19)
 text(df$x[ind], df$y[ind], df$geo[ind], adj = c(0.5,-1), cex=1.5)

Maps

Disposable income of private households by NUTS 2 regions at 1:60mln resolution using tmap

The mapping examples below use tmap package.

library(dplyr)
library(eurostat)
library(sf)
library(tmap)

# Download attribute data from Eurostat
sp_data <- eurostat::get_eurostat("tgs00026", time_format = "raw", stringsAsFactors = FALSE) %>% 
  # subset to have only a single row per geo
  dplyr::filter(time == 2010, nchar(geo) == 4) %>% 
  # categorise
  dplyr::mutate(income = cut_to_classes(values, n = 5))

# Download geospatial data from GISCO
geodata <- get_eurostat_geospatial(output_class = "sf", resolution = "60", nuts_level = 2, year = 2013)

# merge with attribute data with geodata
map_data <- inner_join(geodata, sp_data)

Construct the map

map1 <- tmap::tm_shape(geodata) +
  tmap::tm_fill("lightgrey") +
  tmap::tm_shape(map_data) +
  tmap::tm_grid() +
  tmap::tm_polygons("income", title = "Disposable household\nincomes in 2010",  
                    palette = "Oranges")
print(map1)  

Interactive maps can be generated as well

# Interactive
tmap_mode("view")
map1

# Set the mode back to normal plotting
tmap_mode("plot")
print(map1)

Disposable income of private households by NUTS 2 regions in Poland with labels at 1:1mln resolution using tmap

library(eurostat)
library(dplyr)
library(sf)
library(RColorBrewer)

# Downloading and manipulating the tabular data
print("Let us focus on year 2014 and NUTS-3 level")
euro_sf2 <- get_eurostat("tgs00026", time_format = "raw",
                         stringsAsFactors = FALSE,
             filter = list(time = "2014")) %>% 

  # Subset to NUTS-3 level
  dplyr::filter(grepl("PL",geo)) %>% 
  # label the single geo column
  mutate(label = paste0(label_eurostat(.)[["geo"]], "\n", values, "€"),
         income = cut_to_classes(values))

print("Download geospatial data from GISCO")
geodata <- get_eurostat_geospatial(output_class = "sf", resolution = "60", nuts_level = 2, year = 2013)

# Merge with attribute data with geodata
map_data <- inner_join(geodata, euro_sf2)

# plot map
map2 <- tm_shape(geodata) +
  tm_fill("lightgrey") +
  tm_shape(map_data, is.master = TRUE) +
  tm_polygons("income", title = "Disposable household incomes in 2014",
              palette = "Oranges", border.col = "white") + 
  tm_text("NUTS_NAME", just = "center") + 
  tm_scale_bar()
map2

Disposable income of private households by NUTS 2 regions at 1:10mln resolution using spplot

library(sp)
library(eurostat)
library(dplyr)
library(RColorBrewer)
dat <- get_eurostat("tgs00026", time_format = "raw", stringsAsFactors = FALSE) %>% 
  # subsetting to year 2014 and NUTS-2 level
  dplyr::filter(time == 2014, nchar(geo) == 4) %>% 
  # classifying the values the variable
  dplyr::mutate(cat = cut_to_classes(values))

# Download geospatial data from GISCO
geodata <- get_eurostat_geospatial(output_class = "spdf", resolution = "10", nuts_level = 2, year = 2013)

# merge with attribute data with geodata
geodata@data <- left_join(geodata@data, dat)

# plot map
sp::spplot(obj = geodata, "cat", main = "Disposable household income",
       xlim = c(-22,34), ylim = c(35,70), 
           col.regions = c("dim grey", brewer.pal(n = 5, name = "Oranges")),
       col = "white", usePolypath = FALSE)

Disposable income of private households by NUTS 2 regions at 1:60mln resolution using ggplot2

Meanwhile the CRAN version of ggplot2 is lacking support for simple features, you can plot maps with ggplot2 by downloading geospatial data as data.frame with output_class argument set as df.

library(eurostat)
library(dplyr)
library(ggplot2)
dat <- get_eurostat("tgs00026", time_format = "raw", stringsAsFactors = FALSE) %>% 
  # subsetting to year 2014 and NUTS-2 level
  dplyr::filter(time == 2014, nchar(geo) == 4) %>% 
  # classifying the values the variable
  dplyr::mutate(cat = cut_to_classes(values))

# Download geospatial data from GISCO
geodata <- get_eurostat_geospatial(resolution = "60", nuts_level = "2", year = 2013)

# merge with attribute data with geodata
map_data <- inner_join(geodata, dat)

ggplot(data=map_data) + geom_sf(aes(fill=cat),color="dim grey", size=.1) + 
    scale_fill_brewer(palette = "Oranges") +
  guides(fill = guide_legend(reverse=T, title = "euro")) +
  labs(title="Disposable household income in 2014",
       caption="(C) EuroGeographics for the administrative boundaries 
                Map produced in R with a help from Eurostat-package <github.com/ropengov/eurostat/>") +
  theme_light() + theme(legend.position=c(.8,.8)) +
  coord_sf(xlim=c(-12,44), ylim=c(35,70))

SDMX

Eurostat data is available also in the SDMX format. The eurostat R package does not provide custom tools for this but the generic rsdmx and rjsdmx R packages can be used to access data in that format when necessary:

library(rsdmx)

# Data set URL
url <- "http://ec.europa.eu/eurostat/SDMX/diss-web/rest/data/cdh_e_fos/..PC.FOS1.BE/?startperiod=2005&endPeriod=2011"

# Read the data from eurostat
d <- readSDMX(url)

# Convert to data frame and show the first entries
df <- as.data.frame(d)

kable(head(df))

Further examples

For further examples, see the package homepage.

Citations and related work

Citing the data sources

Eurostat data: cite Eurostat.

Administrative boundaries: cite EuroGeographics

Citing the eurostat R package

For main developers and contributors, see the package homepage.

This work can be freely used, modified and distributed under the BSD-2-clause (modified FreeBSD) license:

citation("eurostat")

Related work

This rOpenGov R package is based on the earlier CRAN packages statfi and smarterpoland.

The independent reurostat package develops related Eurostat tools but seems to be in an experimental stage at the time of writing this tutorial.

The more generic quandl, datamart, rsdmx, and pdfetch packages may provide access to some versions of eurostat data but these packages are more generic and hence, in contrast to the eurostat R package, lack tools that are specifically customized to facilitate eurostat analysis.

Contact

For contact information, see the package homepage.

Version info

This tutorial was created with

sessionInfo()


Try the eurostat package in your browser

Any scripts or data that you put into this service are public.

eurostat documentation built on Sept. 21, 2018, 6:25 p.m.