Legend manipulation

One of the most useful things you can do is to manipulate the legend. You can rename items in the legend, restyle it, move it around, or just delete it altogether.

p + theme(legend.position = "none") #remove the legend

p + theme(legend.position = c(.1, .75)) #move the legend to a different position

p + theme(legend.title = element_blank()) #remove the title of the legend

Axes manipulation

First of all, it’s about time we do something about our ugly data. One of our main issues is that the country areas (on the X axis) should be log-transformed. We can transform our data using a simple line of code, and some functions from the scales package. Here are some examples, but make sure to check the trans() function documentation and pick the type of transformation that fits your data!

p + scale_x_continuous(trans = log10_trans()) #log10 transformation

p + scale_x_continuous(trans = sqrt_trans()) #square-root transformation

p + scale_x_continuous(trans = reverse_trans()) #reverse transformation

And to make this look a bit better, let’s change the tick labels to have actual numbers. We’ll do this with the trans_format() function. We tell the function which transformation to apply to the numbers, and the format in which to write the labels.

p <- p + scale_x_continuous(trans = log10_trans(),
                            labels = trans_format("log10", math_format(10^.x)))
p

You may also want to change the titles of your axes - this is always useful because usually your data will have column names that are short and easy to type, but less informative for a plot you want to publish. The expression() function (from base R) is useful for this, because it allows you to add special characters, as well as create super- and subscripts, italics, etc.

p <- p + labs(x = expression("Area (km"^"2"~")"),
              y = expression("Consumption (l"/"y"~")"))
p

High quality files

As you’ve probably noticed, while the PDF file looks ok, the JPG and PNG are really poor quality. However, we can fix this. For instance, we can create a TIFF file with a higher dpi and larger dimensions. Also, we can change the size of the PDF, or export to other, different file formats.

tiff("myplot.tiff", height = 20, width = 20, units = "cm", res = 300)
p3
dev.off()

## quartz_off_screen 
##                 2

pdf("myplot.pdf", height = 20, width = 20)
p3
dev.off()

## quartz_off_screen 
##                 2

Tip: pdf() is super useful! The function has arguments to e.g. change the font family, combine different plots to a single file, choose the target paper size, etc. It’s highly recommended to check the function documentation!

Introduction

plotly is a great cross-platform package, originating in JavaScript (JS) form, that allows a lot of options for plotting interactive plots. We will use the R version of the package. It can be used in two ways. The simplest usage of the package is to use ggplotly(), which transforms ggplot objects to interactive visualizations. This is done with the code

#create a ggplot object
p <- ggplot(aes(x = SQKM_CNTRY, y = total, colour = Continent), data = data_vis) + geom_point() + ggtitle("Per capita alcohol consumption") + scale_colour_manual(values = witness.me) + scale_x_continuous(trans = log10_trans())
p

#transform into an interactive object
ggplotly(p)

The second option is to transfer data into an interactive object. This is done with the plot_ly() function. This function was originally designed to create plots that were not available with ggplot such as 3D and mash plots. However, today you can create any plot with this function, and the majority of ggplot objects can be transformed into a plotly object.

We will learn a bit about the plot_ly() syntax for you to be able to use the power of the JS library in R.

plot_ly() syntax

This function is extremely versatile and it allows you to control and change every small thing on the plot. The plot_ly() function uses the pipe symbol %>% to add traces, which are equivalent to geom in ggplot and describe the type of plot you want, and layers which are anything you want to change or add to the the type of plot. Scatter is the default type. Let’s look at the most basic plot_ly() plot: Litres of beer consumed as a function of country area

plot_ly(data_vis,x=~log10(SQKM_CNTRY),y=~beer, type = 'scatter')

Adding color

Now lets add a discrete variable to use as color variable in the plot. We’ll use island (yes or no) as a factor

plot_ly(data_vis,x=~log10(SQKM_CNTRY),y=~beer, type = 'scatter',color =~ ISLAND)

We can change the colors to what we want

plot_ly(data_vis,x=~log10(SQKM_CNTRY),y=~beer, type = 'scatter',color =~ISLAND,colors = "Set1")

Legend position

We can change the legend position

#legende bottom
plot_ly(data_vis,
        x=~log10(SQKM_CNTRY),
        y=~beer, 
        type = 'scatter',
        color =~ISLAND,
        colors = "Set1") %>% 
   layout(legend = list(orientation = 'h'))

#legend witin plot
plot_ly(data_vis,
        x=~log10(SQKM_CNTRY),
        y=~beer, 
        type = 'scatter',
        color =~ISLAND,
        colors = "Set1") %>% 
layout(legend = list(x = 0.1, y = 0.9))

#no ledgend
plot_ly(data_vis,
        x=~log10(SQKM_CNTRY),
        y=~beer, 
        type = 'scatter',
        color =~ISLAND,
        colors = "Set1") %>% 
  layout(showlegend = FALSE)

Change axis names

plot_ly(data_vis,
        x=~log10(SQKM_CNTRY),
        y=~beer, 
        type = 'scatter',
        color =~ISLAND,
        colors = "Set1") %>% 
layout(legend = list(x = 0.1, y = 0.9),
       xaxis = list(title = "where in the world",
                    titlefont = list(family = "Comic Sans MS",size = 18)),
       yaxis = list(title = "Beer consumption",
                    titlefont = list(family = "Courier New, monospace", size =30)))

A cleaner way to do the same

#first font
f1 = list(family = "Comic Sans MS",size = 18)
#second font
f2 = list(family = "Courier New, monospace", size =30)
#xaxis
x = list(title = "where in the world", titlefont = f1)
#yaxis
y = list(title = "Beer consumption", titlefont = f2)

#make the plot
plot_ly(data_vis,
        x=~log10(SQKM_CNTRY),
        y=~beer, 
        type = 'scatter',
        color =~ISLAND,
        colors = "Set1") %>% 
layout(legend = list(x = 0.1, y = 0.9),
       xaxis = x,
       yaxis = y)

Plot side by side

We can plot side by side each level in the discrete variable. Make sure that the data you want to plot are in a long format.

p <- data_vis %>%
  #transform the variable that will be used to plot to integer type
  transform(id = as.integer(factor(LANDLOCKED))) %>%
  #plot the data where X is the predictor, y is the response, color is thedivider, yaxis is the variable to use to divide the plot
  plot_ly(x = ~total, y = ~beer, color =~LANDLOCKED, colors = "Dark2",
          yaxis = ~paste0("y", id)) %>%
  # define the number of subplots and share the x to plot them together
  subplot(nrows = 2, shareX = TRUE)
p

p <- data_vis %>%
  #transform the variable that will be used to plot to integer type
  transform(id = as.integer(factor(LANDLOCKED))) %>%
  #plot the data where X is the predictor, y is the response, color is thedivider, yaxis is the variable to use to divide the plot
  plot_ly(x = ~total, y = ~beer, color = ~LANDLOCKED, colors = "Dark2",
          yaxis = ~paste0("y", id)) %>%
  #make a line graph instead of just scatterplot
  add_lines() %>%
  # define the number of subplots and share the x to plot them together
  subplot(nrows = 2, shareX = TRUE)
p

Change the size of the dots based on a variable

#change size for all dots
p<- plot_ly(data_vis,
        x=~log10(SQKM_CNTRY),
        y=~wine, 
        type = 'scatter',
        color =~LANDLOCKED,
        colors = "Set1",
        mode = "markers",
        marker = list(size = 20)) %>% 
  layout(showlegend = FALSE)
p

#change size based on a continious factor
p<- plot_ly(data_vis,
        x=~log10(SQKM_CNTRY),
        y=~log10(POP_CNTRY), 
        type = 'scatter',
        mode = "markers",
        size = ~total,
        color =~LANDLOCKED,
        colors = "Set1") %>% 
  layout(showlegend = FALSE)
p

Line plots

plot_ly(data_vis, x = ~log10(POP_CNTRY), y = ~log10(SQKM_CNTRY)) %>%
  add_lines(color = ~LANDLOCKED, colors = "Set1", alpha = 0.9)

Barplot

p2 <- data_vis %>%
  dplyr::count(Continent) %>%
  plot_ly(x = ~Continent, y = ~n) %>% 
  add_bars()
p2

Proportion barplot

# number of wines by continent and whether its and island (n)
cc <- count(data_vis, Continent, ISLAND)
# number of diamonds by Continent (nn)
cc2 <- left_join(cc, count(cc, Continent, wt = n))
cc2 %>%
  mutate(prop = n / nn) %>%
  plot_ly(x = ~Continent, y = ~prop, color = ~ISLAND) %>%
  add_bars() %>%
  layout(barmode = "stack")

Boxplot

#Standard boxplot
p <- plot_ly(data_vis, y = ~total, color = I("darkgreen"), 
             boxpoints = "suspectedoutliers")%>% 
  add_boxplot(x = ~Continent)
p

#faceted boxplot
p<- plot_ly(data_vis, x = ~wine, y = ~interaction(ISLAND, Continent)) %>%
  add_boxplot(color = ~ISLAND) %>%
  layout(yaxis = list(title = ""), margin = list(l = 100))
p

Exporting locally as html widget

This can be used in a website or presentation

htmlwidgets::saveWidget(p, "my_plot.html")

Export static

To export a high quality static plot from plot_ly() you need to install an additional package for your OS names orca. Follow this github to unstall it: https://github.com/plotly/orca#installation Then write

# orca(p, file='image.png',width = 10,height = 10)

To summarize plotly

Pros: A strong tool for interactive ploting especially if you want to use Shiny. Also has the ggplotly option that converts all ggplots to interactive ones. Cons: Syntax is not intuitive and needs some time to get used to if you don’t know JS

Scatterplot

We’ll start with a scatterplot

p<- data_vis %>% 
  ggvis(x =~log10(POP_CNTRY),y =~spirits) %>% 
  layer_points()
p

Adding variables to plot

You can add more variables to the plot by mapping them to other visual properties: fill = color stroke= color palette size = size of the symbols shape= shape of the symbols

data_vis %>% 
  ggvis(x =~log10(POP_CNTRY),
        y =~spirits,fill=~Continent,size =~wine,shape =~ISLAND) %>% 
  layer_points()

To use a fixed color or size instead of a variable, use :=

data_vis %>% 
  ggvis(x =~log10(POP_CNTRY),y =~spirits,fill:="red",size :=30,shape :="square") %>% 
  layer_points()

Add grouping with group_by()

data_vis %>% 
ggvis(x = ~wine, y = ~beer) %>% 
  layer_points(fill = ~Continent) %>%
  group_by(Continent) %>% 
  layer_model_predictions(model = "lm",se = T)

## Guessing formula = beer ~ wine

Fix legend

data_vis %>% 
ggvis(x = ~wine, y = ~beer) %>% 
  layer_points(fill = ~Continent,size =~total) %>%
  group_by(Continent) %>% 
 add_legend(c("size", "fill"),
            orient = "left")

Histogram

data_vis %>% 
  ggvis(~spirits,fill:="darkgreen") %>% 
  layer_histograms()

## Guessing width = 0.2 # range / 34

Barplot

data_vis %>% 
  ggvis(~log10(POP_CNTRY), ~wine) %>%
  layer_bars(width = 0.04)

Line plot

data_vis %>% 
  ggvis(~log10(POP_CNTRY), ~log10(SQKM_CNTRY)) %>%
  layer_lines()

Boxplot

data_vis %>% 
  ggvis(x=~Continent, y=~wine,fill:="darkblue") %>% 
  layer_boxplots()

Make the plot interactive

ggvis gives an option of interactive plotting. It is different from the interactive plotting in plotly but maybe useful for other things.

#add sliders to the size and the opacity
data_vis %>% 
  ggvis(x =~log10(POP_CNTRY),
        y =~spirits,
        fill:="red",
        size := input_slider(10, 300),
        opacity := input_slider(0, 1)
        ) %>% 
  layer_points()

## Warning: Can't output dynamic/interactive ggvis plots in a knitr document.
## Generating a static (non-dynamic, non-interactive) version of the plot.

model_type <- input_checkbox(label = "Use flexible curve",
  map = function(val) if(val) "loess" else "lm")
ggvis(data = data_vis,x=~total, y=~spirits) %>%
  layer_model_predictions(model = model_type)

Summary

ggvis is a good tool to create plots in R and is in some ways similar to ggplot2 Pros: Uses less lines than ggplot to create the same plot Cons: Has less documentation and forums which makes it a bit harder to learn

Here is a good comparison review of ggplot2 vs. ggvis and the difference between them https://stats.stackexchange.com/questions/117078/for-plotting-with-r-should-i-learn-ggplot2-or-ggvis/117080

Here is a good comparison between the different available packages for plotting in R http://ouzor.github.io/blog/2014/11/21/interactive-visualizations.html

Manipulating the data

To plot geographical data with ggplot you first need to change the structure of the data. We will use the shapefile data_vis.sp that we created previously.

To use the shapefile in ggplot2 we have to first fortify the spatial attribute of our data and then join the metadata back to the new spatial data

#save the metadata in a new object
data_vis.sp_d<- data_vis.sp@data
#fortify the data use the ID column as region
data_vis.sp_f <- fortify(data_vis.sp, region = "ID") %>% 
  mutate(id = as.numeric(id))
# now we join the thematic data
map_data <- data_vis.sp_f %>% left_join(data_vis.sp_d, by = c("id" = "ID"))

Basic map

p <- ggplot() +
    # municipality polygons
    geom_polygon(data = map_data, aes(fill = beer, 
                                      x = long, 
                                      y = lat, 
                                      group = group)) +
    # municipality outline
    geom_path(data = map_data, aes(x = long, 
                                   y = lat, 
                                   group = group), 
              color = "white", size = 0.1) +
    coord_equal() +
    # add the previously defined basic theme
    theme_map() +
    labs(x = NULL, 
         y = NULL, 
         title = "Beer consumption around the world", 
         subtitle = "Litres consumed per capita"
         )
p

#change the color to have a nicer color scheme
p+scale_fill_viridis(option = "magma", direction = -1)

Change the legend position

q <- p +
  # this is the main part
  theme(legend.position = "bottom") +
  scale_fill_viridis(
    option = "magma", 
    direction = -1,
    name = "Litres per country",
    # here we use guide_colourbar because it is still a continuous scale
    guide = guide_colorbar(
      direction = "horizontal",
      barheight = unit(2, units = "mm"),
      barwidth = unit(50, units = "mm"),
      draw.ulim = F,
      title.position = 'top',
      # some shifting around
      title.hjust = 0.5,
      label.hjust = 0.5
  ))
q

Plot a descrete variable

pp <- ggplot() +
    # municipality polygons
    geom_polygon(data = map_data, aes(fill = Continent, 
                                      x = long, 
                                      y = lat, 
                                      group = group)) +
    # municipality outline
    geom_path(data = map_data, aes(x = long, 
                                   y = lat, 
                                   group = group), 
              color = "white", size = 0.1) +
    coord_equal() +
    theme_map() +
    theme(legend.position = "bottom") +
    labs(x = NULL, 
         y = NULL, 
         title = "The continents of the world")
brks_scale <- levels(map_data$Continent)
labels_scale <- rev(brks_scale)
qq <- pp +
    # now we have to use a manual scale, 
    # because only ever one number should be shown per label
    scale_fill_manual(
          # in manual scales, one has to define colors, well, manually
          # I can directly access them using viridis' magma-function
          values = rev(magma(6)),
          breaks = rev(brks_scale),
          name = "Continents",
          drop = FALSE,
          labels = labels_scale,
          guide = guide_legend(
            direction = "horizontal",
            title.position = 'top',
            # I shift the labels around, they should be placed 
            # exactly at the right end of each legend key
            title.hjust = 0.5,
            label.hjust = 1,
            nrow = 1,
            byrow = T,
            # also the guide needs to be reversed
            reverse = T,
            label.position = "bottom"
          )
      )

qq

A link to more things you can add to the plot https://timogrossenbacher.ch/2016/12/beautiful-thematic-maps-with-ggplot2-only/

Add bubble plots

#create the data for the bubbles
ggplot() +
  geom_polygon(data = map_data, aes(x=long, 
                                    y = lat, 
                                    group = group), 
               fill="grey", alpha=0.3)+
   geom_path(data = map_data, aes(x = long, 
                                   y = lat, 
                                   group = group), 
              color = "black", size = 0.1)+
  geom_point(data=map_data, aes(x=x, y=y, size=total, color = total))+
  scale_size_continuous(range=c(0.3,5)) +
  scale_color_viridis(option = "viridis",direction = -1)+
  theme_void()

EXCERCISE

Plot each one of the alcohol consumption types data (beer, wine, spirits) on a map and put them next to eachother in a grid plot of 2X2.

Basic example

Lets put a random point on a default map

m <- leaflet() %>%
  addTiles() %>%  # Add default OpenStreetMap map tiles as a background
  addMarkers(34.807371, 32.115358, popup="Where we are in the world")
m

Backgrounds

We can change the basemap we are using to either one from the internet or a shapefile we have. We will do this using addProviderTiles() that hosts many free third-party maps.

For example a black and white version

m %>% addProviderTiles(providers$Stamen.Toner)

Or a grey version

m %>% addProviderTiles(providers$CartoDB.Positron)

Popups

# create a string with the adrdess and the link to the location
content <- paste(sep = "<br/>",
  "<b><a href='https://smnh.tau.ac.il/'>The Steinhardt Museum of Natural History</a></b>",
  "Klausner St 12",
  "Tel Aviv-Yafo, IL"
)

# plot a default map
leaflet() %>% addTiles() %>%
  #add popup GPS coordinates, content variable allows the popup to appear
  addPopups(34.807371, 32.115358, content)

You can use this website to find GPS coordinates for the address you are interested in https://www.gps-coordinates.net/

Let’s put popups in the center of all our countries with their names

#create the data that we will pass to the popup layer.
df<- data_vis.sp@data[,c("CNTRY_NAME","x","y")]

leaflet(df) %>% addTiles() %>%
  addMarkers(~x, ~y, popup = ~htmlEscape(CNTRY_NAME))

Labels and Markers

Labels will make the information in the popup markers appear when you hover over them

leaflet(df) %>% addTiles() %>%
  addMarkers(~x, ~y, label = ~htmlEscape(CNTRY_NAME))

leaflet(data_vis.sp) %>% addTiles() %>% addMarkers(~x,~y,
  clusterOptions = markerClusterOptions()
)

#create the data
df2<- filter(data_vis.sp@data, CNTRY_NAME %in% c("United States","Russia","Papua New Guinea"))
#control the looks of each label
leaflet(df2) %>% addTiles() %>%
  addMarkers(~x[1], ~y[1], label = ~htmlEscape(CNTRY_NAME[1]),
             #this is a use of CSS programming language to change the parameters of the label
             labelOptions = labelOptions(noHide = T,
                                         direction = "bottom",
                                         style = list(
                                           "color" = "red",
                                           "font-family" = "serif",
                                           "font-style" = "italic",
                                           "box-shadow" = "3px 3px rgba(0,0,0,0.25)",
                                           "font-size" = "12px",
                                           "border-color" = "rgba(0,0,0,0.5)"
                                           ))) %>%
    addMarkers(~x[2], ~y[2], label = ~htmlEscape(CNTRY_NAME[2]),
             labelOptions = labelOptions(noHide = T, textsize = "15px"
                                           )) %>%
  addMarkers(~x[3], ~y[3], label = ~htmlEscape(CNTRY_NAME[3]),
             labelOptions = labelOptions(noHide = T, textOnly = TRUE
                                           ))

#create an object with custom markers
leafIcons <- icons(
  iconUrl = ifelse(df2$SQKM_CNTRY < 10000000,
    "http://leafletjs.com/examples/custom-icons/leaf-green.png",
    "http://leafletjs.com/examples/custom-icons/leaf-red.png"
  ),
  iconWidth = 38, iconHeight = 95,
  iconAnchorX = 22, iconAnchorY = 94,
  shadowUrl = "http://leafletjs.com/examples/custom-icons/leaf-shadow.png",
  shadowWidth = 50, shadowHeight = 64,
  shadowAnchorX = 4, shadowAnchorY = 62
)

leaflet(data = df2) %>% addTiles() %>%
  addMarkers(~x, ~y, icon = leafIcons)

url.r<- "https://toppng.com/public/uploads/preview/vladimir-putin-11530963554dkmmv4ldms.png"
url.us<- "https://toppng.com/public/uploads/preview/donald-trump-11530966029o6kuaps0i2.png"
url.png<- "https://tinyurl.com/ybehw368"

oceanIcons <- icons(iconUrl = c(url.png,url.r,url.us),
                    iconWidth = 40, iconHeight = 40)

leaflet(df2) %>% addTiles() %>%
  # Select from oceanIcons based on df$type
  addMarkers(~x,~y,icon = ~oceanIcons)

Polygons and Lines

Let’s color only few countries from our polygon

#subset the data. Use the first 10 countries
data_vis.sp_sub<- data_vis.sp[1:10,]

#pass the data to leaflet
leaflet(data_vis.sp_sub) %>%
  #add a base background
  addTiles() %>%
  #add the polygon with colors based on the size cut-off.
  #The first color is passed to the borders of the polygons
  addPolygons(color = "#444444", weight = 1, smoothFactor = 0.5,
    opacity = 1.0, fillOpacity = 0.5,
    #fill color determines what to color based on a condition (SQKM_CNTRY) and makes sure that it is indeed colored
    fillColor = ~colorQuantile("YlOrRd", SQKM_CNTRY)(SQKM_CNTRY),
    #highlight is met to define the border
    highlightOptions = highlightOptions(color = "white", weight = 2,
      bringToFront = TRUE))

Circles on maps

Lets add circles with changing size based on the total alcohol consumption

# We'll use the data from the shapefile we created previously
df3<- data_vis.sp_sub@data

leaflet(df3) %>%
  addTiles() %>%
  addCircles(lng = ~x, lat = ~y, weight = 1,
    radius = ~sqrt(total) * 90000, popup = ~CNTRY_NAME)

Choropleths maps

Choropleths refers to maps in which the polygons are colored based on a continuous or a discrete variable. This is similar to what we did with ggplot2

Lets start with a basic one, color the countries based on spirit consumption

# create the color palette with viridis, let it decide the bins by itself
pal <- leaflet::colorBin(viridis_pal(option = "A",direction = -1)(5), domain = data_vis.sp$wine)
# this comment pal object is using colorBrewer in case you want a different method to create your colors
# pal <- colorBin("RdYlBu", domain = data_vis.sp@data$spirits)
# plot the data
m <- leaflet(data_vis.sp) %>%
  setView(0, 37.8, 1.5) %>%
  addProviderTiles(providers$Esri.WorldStreetMap)
m %>% addPolygons(
  fillColor = ~pal(wine),
  weight = 1,
  opacity = 0.5,
  color = "black",
  dashArray = "3",
  fillOpacity = 0.8)

m<- m %>% addPolygons(
  fillColor = ~pal(wine),
  weight = 1,
  opacity = 1,
  color = "black",
  dashArray = "3",
  fillOpacity = 0.9,
  highlight = highlightOptions(
    weight = 3,
    color = "#666",
    dashArray = "",
    fillOpacity = 0.5,
    bringToFront = TRUE))
m

Add a legend

Basic legend

m %>% addLegend(pal = pal, values = ~wine, opacity = 0.7, title = NULL,
  position = "bottomleft")

Adding some stuff to the legend

m %>% addLegend("bottomright", pal = pal, values = ~wine,
    title = "Wine liters per person",
    labFormat = labelFormat(suffix = " Liters"),
    opacity = 1
  )

Add a minimap to see where we are in the world

m %>%
  addProviderTiles(providers$Esri.WorldStreetMap) %>%
  addMiniMap()