Introduction

For the community contribution, we want to compare three different ways of drawing census graph for The United States. The three different methods are choroplethr, ggplot2, and tmap. We will use the basic R dataset state.x77 to demonstrate how to draw the the graph using those methods and provide the pros and cons of three methods in our personal experience perspective.

Loading all the libraries.

library(sf)          # classes and functions for vector data
library(REmap) #A package sepcified for China that we think people might be interested in.
library(choroplethr) #package for choroplethr tool
library(maps) # Display of maps, contains map_data function which can "easily turn data from the maps package in to a data frame suitable for plotting with ggplot2". 
library(tmap) #package for tmap graph tool
library(tidyverse) #basic data manipulate

Comparison

Choroplethr

The choroplethr would be considered as the easiest way of drawing the census graph. However, as a trade-off, it loses the flexibility that the other two methods have. For choroplethr, we only need to provide the name of states and the inforamtion we want to draw on the graph.

https://arilamstein.com/documentation/choroplethr/reference/index.html
https://edav.info/maps.html#choropleth-maps

For the package itself contains the population information for us in df_pop_state

data(df_pop_state)
state_choropleth(df_pop_state, title="2012 Population in US by states", legend="Population")

Draw information that we want to know.

choroplethe_Income_data <- state.x77 %>% as.data.frame() %>% 
  rownames_to_column("state") %>%
  transmute(region = tolower(`state`), value = Income)
state_choropleth(choroplethe_Income_data, title="2012 Income in US by states", legend="Income")

You can also draw the graph for country using county_choropleth and the graph for world using country_choropleth. You can also draw graph for a sepcific country as long as it is in the package choroplethrAdmin1 with function admin1_choropleth (Example can be found in: https://rdrr.io/cran/choroplethr/man/admin1_choropleth.html).

Disadvantage

Since all the functions are pre-determined, we cannot add more countries or it might be missing in the data. Compared with ggplot2, we can also find one more disadvantage that ggplot2 works with dataframe, so we don’t need to create an aaddtional dataset for different information.

Example

data(df_pop_country) #get information for country.
country_choropleth(df_pop_country, "2012 World Bank Populate Estimates")

#we get the warning massage: The following regions were missing and are being set to NA: namibia, western sahara, taiwan, antarctica, kosovo

#We find that the country seychelles is missing in the original dataset.
which(df_pop_country$region=="seychelles")
## integer(0)
#Add it to the dataset, population number is found in Wikipedia.
seychelles <- c("seychelles",95843)
df_pop_country_with_seychelles <- rbind(df_pop_country,seychelles)
country_choropleth(df_pop_country_with_seychelles, "2012 World Bank Populate Estimates")

# Get warning information : Your data.frame contains the following regions which are not mappable: seychelles

ggplot2 and tmap

For ggplot2 and tmap, they both need CRS information of the spatial data in order to draw census. They also share similar syntax. What gglot2 can do is also aviliable in tmap, however in ggplot2, you need to specify the parameter as longitude and latitude. Tmap provides a simple way to do interactive graph(by simply changing the model of tmap, you don’t even need to draw a new graph). In ggplot2, you can hardly do it without plotly. ggplot2 works with shp. tmap works with sf data formate (sf is a more organized data format for spatial data). When drawing it with facet, tmap is clearly better under view than ggplot2 since it has a bottom layer of the overall map. Another difference between ggplot2 and tmap is that you need to have the central CRS information for each state in order to draw the bubble graph.

Compare the difference between sph and sf.

ggplot2_states <- map_data("state")
nrow(ggplot2_states)
## [1] 15537
head(ggplot2_states,5)
##        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>
tmap_states <- sf::st_as_sf(maps::map("state", plot = FALSE, fill = TRUE))
nrow(tmap_states)
## [1] 49
head(tmap_states,5)
## Simple feature collection with 5 features and 1 field
## geometry type:  MULTIPOLYGON
## dimension:      XY
## bbox:           xmin: -124.3834 ymin: 30.24071 xmax: -84.90089 ymax: 42.02073
## epsg (SRID):    4326
## proj4string:    +proj=longlat +datum=WGS84 +no_defs
##                         geometry         ID
## 1 MULTIPOLYGON (((-87.46201 3...    alabama
## 2 MULTIPOLYGON (((-114.6374 3...    arizona
## 3 MULTIPOLYGON (((-94.05103 3...   arkansas
## 4 MULTIPOLYGON (((-120.006 42... california
## 5 MULTIPOLYGON (((-102.0552 4...   colorado

ggplot2

ggplot2 provides geom_polygon function to draw census graph.

https://socviz.co/maps.html
https://cran.r-project.org/web/packages/tmap/tmap.pdf
https://geocompr.robinlovelace.net/adv-map.html
https://www.jstatsoft.org/article/view/v084i06
https://cran.r-project.org/web/packages/sf/vignettes/sf1.html
#get data for states
statedata <- state.x77 %>% as.data.frame() %>% 
  rownames_to_column("region")
#change state name to capital so it works with inner_join
ggplot2_states$region <- str_to_title(ggplot2_states$region)
ggplot2_statesdata <- inner_join(ggplot2_states,statedata,"region")
region_data <- data.frame(region=state.name,area = state.region)
ggplot2_statesdata <- inner_join(ggplot2_statesdata,region_data,"region")

#basic census graph.
g1 <- ggplot(data = ggplot2_statesdata,
  mapping = aes(x = long, y = lat, group = group,fill=Income))+
  scale_fill_gradient(low = "azure", high = "dodgerblue3") +
  geom_polygon(color="gray90",size=0.1)
g1

#draw it with facet that based on what part of us the state is in.
g1+facet_wrap(.~area,nrow = 2)

#draw the bubble plot

#this is what you shouldn't do
ggplot(data = ggplot2_statesdata)+
  geom_polygon(aes(x=long, y=lat, group=group), colour = "grey90", fill="white", size=0.1) +
  geom_point(aes(x=long, y=lat,size=Frost))

#seems we need the central point spatial information for each state to draw the tmap bubble function.
#get the idea from https://stackoverflow.com/questions/14773477/putting-interest-rates-on-us-states-map-in-r
state.location <- data.frame ("region" = state.name,
                              "Longitude" = state.center$x,
                              "Latitude" = state.center$y)

ggplot2_statesdata <- inner_join(ggplot2_statesdata,state.location,"region")


ggplot(data = ggplot2_statesdata)+
  geom_polygon(aes(x=long, y=lat, group=group), colour = "grey90", fill="white", size=0.1) +
  geom_point(aes(x=Longitude, y=Latitude,size=Frost),colour = "lightblue")

tmap

https://cran.r-project.org/web/packages/tmap/vignettes/tmap-getstarted.html
#create the dataset that contains all the information we need.
colnames(tmap_states)[colnames(tmap_states)=="ID"] <- "region"
tmap_states$region <- str_to_title(tmap_states$region)
tmap_states <- inner_join(tmap_states,statedata,"region")
tmap_states <- inner_join(tmap_states,region_data,"region")


#plot model gives us the plot
tmap_mode("plot")
t1 <- tm_shape(tmap_states) +
  tm_polygons("Income")
t1

#create census with facet
t2 <- t1+tm_facets(by = "area")

#view mode gives us interactive plot

tmap_mode("view")
t1
t2
#Bubble graph
tm1 <- tm_shape(tmap_states) + tm_polygons("Frost")
tm2 <- tm_shape(tmap_states) + tm_bubbles(size = "Frost",col = "lightblue")

tmap_arrange(tm1, tm2)

Additional information for people who want to draw the census graph for China. Since China spatial data is not avilible in the package mentioned above.

https://github.com/Lchiffon/REmap
https://blog.csdn.net/wzgl__wh/article/details/53108754?from=singlemessage&isappinstalled=0
#there is an additional map package that provide china spatial information
#however, hard to understand the data.
library(mapdata)
maps::map("china")

china <- map_data("china")
tail(china,1)
##           long      lat group order region subregion
## 94999 112.2986 4.250258   712 94999    712      <NA>
#china has no province information and nearly 200 region, hard to figure out what it is represent.
##An alternative would be REmap
library(REmap)
#plot it with random number assigned to province.
data = data.frame(country = mapNames("china"),
                   value = 5*sample(34)+200)
out = remapC(data,maptype = "china",color = 'skyblue')
plot(out)
## Save img as: /var/folders/8y/jc48yh4x797_lhk5gvktpfl40000gn/T//RtmpsJIrhK/ID_20190330014036_1917.html

Conclusion

In summary, we provide packages that can be used to draw census graphs in R.The Choroplethr is really easy and time-saving. Eventhough it sacrifies the ability to add more area/country we may want, we don’t need to do much data manipultion and don’t need to provide the spatial data. Choroplethr should be considered as a fairly good choice. For ggplot2 and tmap, personally, I would recommand tmap. Eventhough you can do the same thing with ggolot2, it is more complicated and requires more data. Tmap provides a simpler way to generate interactive graph and the graph tends to be more good-looking.