README

WeLoveDataScience 2017-07-28

geo_facet

The geofacet R package provides a way to flexibly visualize data for different geographical regions by providing a ggplot2 faceting function facet_geo() which works just like ggplot2’s built-in faceting, except that the resulting arrangement of panels follows a grid that mimics the original geographic topology as closely as possible.

For more information on this exciting package, see https://hafen.github.io/geofacet/.

This repository is a quick demonstration on how to use geo_facet with a specific map created by WeLoveDataScience for Belgium (provinces + Brussels level).

We take as example data real estate prices evolution over more than 40 years in Belgium (open data provided by Belgium government).

Belgium map

Our point of comparison

geofacet map

Here is what we propose as corresponding geo-facet map:

library(geofacet)

## Warning: package 'geofacet' was built under R version 3.3.3

library(ggplot2)
bemap <- read.csv("geofacet_be_map.csv")
grid_preview(bemap,label="name_EN")

## You provided a user-specified grid. If this is a generally-useful
##   grid, please consider submitting it to become a part of the
##   geofacet package. You can do this easily by calling:
##   grid_submit(__grid_df_name__)

Example of usage

Let's import some open data from Belgium government. Those are data about real estates prices evolution over more than 40 years in Belgium (houses, flats...).

library(rio)
library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

# Import data at Region level (for Brussels Capital region)
t1=tempfile(fileext=".zip")
data1.link= "http://statbel.fgov.be/fr/binaries/immo_by_region_tcm326-283164.zip"
download.file(data1.link,t1,mode="wb",quiet=TRUE)
unzip(t1)
data1 <- import("immo_by_region.txt")
file.remove(t1)

## [1] TRUE

# filter region data: keep only Brussels and Yearly data
data1 <- filter(data1,CD_PERIOD=="Y" & CD_REFNIS=="4000")


# Import data at Region level (for Brussels Capital region)
t2=tempfile(fileext=".zip")
data2.link= "http://statbel.fgov.be/fr/binaries/immo_by_province_tcm326-283148.zip"
download.file(data2.link,t2,mode="wb",quiet=TRUE)
unzip(t2)
data2 <- import("immo_by_province.txt")
file.remove(t2)

## [1] TRUE

# filter province  data: keep only  Yearly data
data2 <- filter(data2,CD_PERIOD=="Y")

# Append data
data <- rbind(data1, data2)
data$code_NIS5 = data$CD_REFNIS
data$MS_MEAN_PRICE_1K = round(data$MS_MEAN_PRICE/1000,1)

Average houses prices evolution per province

Let's focus on houses

houses <- filter(data, CD_TYPE_FR=="maisons d'habitation")

g<-ggplot(houses, aes(x=CD_YEAR, y=MS_MEAN_PRICE_1K))+
    geom_smooth(color="steelblue")
g <- g+facet_geo(~code_NIS5,grid=bemap, label="name_EN")

## You provided a user-specified grid. If this is a generally-useful
##   grid, please consider submitting it to become a part of the
##   geofacet package. You can do this easily by calling:
##   grid_submit(__grid_df_name__)

g <- g+labs(title = paste("Houses prices yearly evolution (in thousands €) - ",paste(range(houses$CD_YEAR),collapse="-")),
    caption = "Data Source: statbel.be",
    x = "Year",
    y = "Price (1.000 €)") +
  theme_bw()

g

## `geom_smooth()` using method = 'loess'