Link Search Menu Expand Document
LOST

Introduction

Graph themes help enhance an already informative line, scatter, bar, or boxplot. With the implementation of useful themes, graphs become aesthetically pleasing; with improper implementation of themes, graphs become muddled and confusing.

Keep in Mind

  • When adding themes to a graph, ensure that the theme being used is appropriate for the data that is represented. For example, a scatterplot works best with graph themes that include gridlines while geometric data graphs look better without any lines.
  • While there are a whole host of existing themes that can improve the appearance of graphs, modifying or creating your own themes can be the most effective way to visualize data. Though seemingly complicated, customizing your own graph theme can be easy depending on the graphing package used, and can be rewarding.

Also Consider

  • Not everyone can see the entire color spectrum. When using certain themes or arguments within themes, bear in mind that it can be hard to distinguish between similar colors or other color combinations for those with color-vision deficiency.

Implementations

R

In this R example, we will be using the ggplot2 package and the theme function to add themes to graphs. The dataset we are using comes built into ggplot2 and is known as msleep. The msleep dataset includes the sleep patterns of 83 different animals, as well as characteristics like the genus, brain weight, and whether the animal is a carnivore, herbivore, or other -vore.

# Load in necessary packages
library(ggplot2)
library(ggthemes)
library(hrbrthemes)

# Load in desired data (msleep)
data(msleep)

After viewing the data, we can create a scatterplot with the y-axis representing the total amount of hours the animal is asleep, the x axis representing the weight of the animal’s brain, and the dots of the scatterplot colored by the -vore type of the animal.

#basic scatterplot of sleep against brain weight
sleep_plot <- ggplot(data=msleep, aes(x = sleep_total, y = brainwt, color = vore)) +
  geom_point()
sleep_plot

Now that we have this basic plot, we can use the theme function to customize it. The ggplot2 package comes with some built-in theme options which can be explored at the Complete themes page of the ggplot2 documentation provided by tidyverse.

Since we are using a scatterplot in this example, theme_bw is a simple and elegant way to display the data. Here is an example of what a theme_bw graph looks like:


#Here is the animal sleep plot we made earlier with all the ggplot2 theme_bw() added to it
bw = sleep_plot + theme_bw()
bw

This is a nice, clean look, with gridlines, borders, and clearly defined axes. While theme_bw does an excellent job displaying the data, there are some themes that do not lend themselves easily to scatterplots. One such theme is theme_void, shown below:

# Animal sleep plot with theme_void() from ggplot2
void = sleep_plot + theme_void()
void

This theme removes all gridlines, borders, and axes, leading to a very confusing image of floating colored points and a legend. This theme is useful for geometric data, flowcharts, or other kinds of visualizations that are clearer without any axes or background. For scatterplots it is less than ideal.

These are very basic themes to use, but they can clean up a graph in a pinch. However, there are other packages that contain their own prepackaged themes. A popular theming package is ggthemes. To check out multiple examples of ggthemes, visit the ALL YOUR FIGURE ARE BELONG TO US page of the official ggthemes website. And for more helpful insight into ggthemes and its arguments, visit the Introduction to ggthemes site by Jeffrey B. Arnold.

While most of the ggthemes themes are great, some really stand out. One popular option is theme_tufte, which is a very minimal theme following the principles of prominent data visualization thinker Edward Tufte. Three other themes (theme_economist, theme_fivethirtyeight, and theme_wsj) all mimic the graph styles of major media/news outlets. If you really want to sell your graph and look like mainstream media, here are some examples:

# Theme following Edward Tufte
tufte = sleep_plot + theme_tufte()
tufte

# theme_economist mimics graphs from The Economist magazine
economist = sleep_plot + theme_economist()
economist

# theme_fivethirtyeight mimics graphs from FiveThirtyEight, a political/sports/statistics blogging site
fivethirtyeight = sleep_plot + theme_fivethirtyeight()
fivethirtyeight

# theme_wsj mimcs graphs from The Wall Street Journal newspaper
wsj = sleep_plot + theme_wsj()
wsj

The data we plotted is rather simple and plain; these themes would look better on a more refined graph, one with proper names, clear positioning, and distinguished data points or lines, but these graphs still look better than the original.

Another great repository of several ggplot2 themes is the hrbrthemes package. Themes in hrbrthemes change the formatting up a bit, so don’t be surprised if your axis titles move a bit. For this example, theme_modern_rc is used, as it makes the colored points of the graph pop off the screen. You can take your graph up another notch by adding one of the color options that hrbrthemes has to offer, in this case scale_color_ft:

# hrbrthemes graph with scale_color_ft and theme_modern_rc 
hrbr = sleep_plot + scale_color_ft() + theme_modern_rc()
hrbr

There is much more that hrbrthemes has to offer beyond just theme_ functions, including other scaling options, font choices, and utilities. For more about hrbrthemes, check out the hrbrthemes page.

If none of these existing themes are your cup of tea, you can try to create your own theme or modify one of those listed above to suit your tastes. To do this, you will use the theme function and include element_xx objects as arguments. Element objects that can be used are listed below:

  • element_line() - can add color, linetype, and size arguments to line elements of a theme, like axis lines
  • element_text() - can add color, face, angle, justification, margins, and size arguments to text elements of a theme, like titles
  • element_rect() - can add color, fill, and size arguments to rectangular elements of a theme, like the panel window
  • element_blank() - can remove any element from a theme

Please take note that customizing themes in this way will not do everything. It will not allow you to change aesthetic properties of your graph geometry, such as the different colors you assign with aes(color=). But it does allow you to make your graphs more aesthetically pleasing by changing font size or color, adding shapes around legends, and filling the backgrounds of graphs with colors and gridlines.

For ways to customize your own theme in ggplot2, check out the R Graphics Cookbook by Winston Chang. Chang offers an excellent section on modifying graph themes in Chapter 9.4; at the bottom of the page is a helpful table that outlines each argument that can be used by theme, a description of what each does, and the element_xx to specify when using an argument.

For a simple example of creating a theme, run the following code:

# This code modifies the legend of the graph
legend = sleep_plot +
  theme(
    legend.background = element_rect(fill = "white", color = "dodgerblue", size = 1),
    legend.title = element_text(color = "brown", face = "bold", size = 18),
    legend.text = element_text(color = "brown", face = "bold", size = 10),
    legend.key = element_rect(color = "dodgerblue", size = 0.5)
  )
  
legend

# We use blue and brown here, as most color-blind people can distinguish these two colors
# element_rect changes the box around the legend and the boxes around the colors for -vore
# element_text changes the text color, size, and face within the legend

You can also modify an already existing theme using the same method as above. Here is what it looks like to modify the axis titles of the hrbrthemes graph we created earlier:

modified = hrbr + theme(axis.title.x = element_text(colour = "yellow", size = 12, face = "bold"), 
  axis.title.y = element_text(colour = "yellow", size = 12, face = "bold"))
modified

This is just scratching the surface of graph themes in R. Even more theme packages exist, like ggpubr, with the excellent publication-ready theme_pubr(), and wesanderson; these won’t be explored on this page, but if you’re interested check out this ggpubr page and this wesanderson page.