For now, we’ve stuck to the basics of data viz. However, there’s so much more you can do. One major thing you can do is to start your document with a setup code block. I find it helps me to have a more productive session in R when I start by gathering all my tools and updating my .qmd settings in the very first code block in my Quarto file.
Here’s what my first code block might look like if I were writing a report using some conflict data.
#| include: false
## 1. Set things up the way I like:
knitr::opts_chunk$set(echo = F,
# I don't want to see my code chunks
message = F,
# I don't want messages appearing
warning = F,
# I don't wart warnings appearing
dpi = 500,
# When I knit, I want my figures crisp
out.width = "75%")
# I want my figures to be a reasonable size
## 2. Here are the packages I need in my session:
library(tidyverse) # I want to use tools in the tidyverse
library(coolorrr) # I want to update my colors
library(socsci) # For some additional summarizing functions
library(geomtextpath) # For plotting text on lines
library(peacesciencer)# For my data
# 3. Set my color palette and ggplot theme:
set_palette(
qualitative = c("navy", "gold", "gray", "red", "steelblue"),
sequential = c("white", "navy"),
diverging = c("steelblue", "white", "red3"),
binary = c("steelblue", "red3")
)
set_theme()
## 4. Get the data I need for analysis:
create_stateyears(subset_years = 1945:2007) |>
add_cow_wars(type = "intra") |>
add_sdp_gdp() |>
add_democracy() -> DataThe code block begins with #| include: false so that none of the output of the code appears when I render my report to a final document. It then proceeds in steps. For step 1, I use knitr::opts_chunk$set() to set a few custom features for code blocks. Since in this example I’m writing a report, presumably for some audience that probably doesn’t care about the underlying code I used, I have option echo = F. This ensures that my code will remain hidden in my rendered report. I also have message = F and warning = F to ensure that any function messages or warnings don’t appear in my rendered document. Finally, so that I have crisp looking data visualizations I have set dpi = 500. I also want to ensure that my data visualizations are reasonably sized, so I have set out.width = "80%".
In step 2, I open up the R packages that I need for my analysis. You can look at the comments next to these packages to see why I need them.
In step 3, I update my color palettes for use with {coolorrr}. I also use set_theme() to update the default ggplot() plotting theme.
Finally, in step 4 I bring in my data that I’ll use for my analysis.
I recommend this workflow. The more prep work you do up front, the more productive your overall session is going to be.
You’ll notice that in the above code I used the function set_theme(). This comes from the {coolorrr} package, which we talked about in the previous notes. This is my own personal ggplot theme that I like to use in my R sessions, and it’s available for you to use (if you like it) once you open the {coolorrr} package. It’s based on FiveThirtyEight’s theme but with a few of my own modifications. By running set_theme(), it sets my personal theme globally. That means that all ggplot figures I make will now use this theme rather than the default “gray” theme.
Check it out:
Data |>
group_by(year) |>
summarize(
conflict_rate = mean(cowintraongoing)
) |>
ggplot() +
aes(
x = year,
y = conflict_rate
) +
geom_point(
color = "gray",
alpha = 0.6
) +
geom_smooth(
se = F,
color = "navy"
) +
scale_y_continuous(
labels = scales::percent
) +
labs(
x = NULL,
y = "Countries at War",
title = "Civil wars post-WWII, 1945-2007"
)
If I wanted to go back to the default, I can just use the appropriate theme_*() function. In this case, I want to use theme_gray():
Data |>
group_by(year) |>
summarize(
conflict_rate = mean(cowintraongoing)
) |>
ggplot() +
aes(
x = year,
y = conflict_rate
) +
geom_point(
color = "gray",
alpha = 0.6
) +
geom_smooth(
se = F,
color = "navy"
) +
scale_y_continuous(
labels = scales::percent
) +
labs(
x = NULL,
y = "Countries at War",
title = "Civil wars post-WWII, 1945-2007"
) +
theme_gray()
There are many different themes you can use. Let’s save the original figure as an object called my_plot and then check out a few of the available theme options.
my_plot <- Data |>
group_by(year) |>
summarize(
conflict_rate = mean(cowintraongoing)
) |>
ggplot() +
aes(
x = year,
y = conflict_rate
) +
geom_point(
color = "gray",
alpha = 0.6
) +
geom_smooth(
se = F,
color = "navy"
) +
scale_y_continuous(
labels = scales::percent
) +
labs(
x = NULL,
y = "Countries at War",
title = "Civil wars post-WWII, 1945-2007"
)Here’s an example with theme_bw():
my_plot + theme_bw()
Here’s an example with theme_classic():
my_plot + theme_classic()
Here’s theme_dark():
my_plot + theme_dark()
Here’s theme_light():
my_plot + theme_light()
And on and on it goes…
You can access even more specialized themes using the {ggthemes} package. You can install it by writing and running install.packages("ggthemes") in the R console.
Here’s the FiveThirtyEight theme:
library(ggthemes)
my_plot + theme_fivethirtyeight()
The Wallstreet Journal theme:
my_plot +
theme_wsj() +
## I need to adjust the title
labs(
title = "Civil wars post-WWII",
subtitle = "1945-2007"
)
The Economist theme:
my_plot + theme_economist()
There’s even a Tufte theme:
my_plot + theme_tufte()
Remember that Tufte was the guy who said we should minimize our ink to data ratio. His theme certainly does.
The set_theme() function I used in the first code chunk set my own person theme globally. You can set your own theme globally, too.
Say you like the clean theme from {ggthemes}. Just run:
theme_set(theme_clean())Notice that this function is theme_set() rather than set_theme(). How confusing! But that’s how it goes with R sometimes.
After you run the above, any new plots you make will follow the theme you set without having to add an extra line of code to your plot:
my_plot
I’m partial to my own, so I’m going to set it back with set_theme() from {coolorrr}:
set_theme()Once you master settings like themes and colors, it’s time to start thinking about how to build up to more complicated visuals.
Here’s an example. Let’s start by filtering our data down to the year 1990:
Data |>
filter(year == 1990) -> data_90Next, let’s build a base plot—a scatter plot showing GDP per capita on the x-axis and quality of democracy on the y-axis:
p <- ggplot(data_90) +
aes(
x = exp(wbgdppc2011est),
y = v2x_polyarchy
) +
geom_point(
alpha = 0.6,
color = "gray"
) +
scale_x_log10(labels = scales::dollar) +
labs(
x = "GDP per capita",
y = "Quality of Democracy",
title = "Economic Health, Democracy, and Civil War",
subtitle = "Data for 1990",
caption = "Source: {peacesciencer}"
)
p # print
Note that I saved the plot as an object called p.
Next, let’s use color to highlight countries experiencing civil war:
p2 <- p +
geom_point(
data = data_90 |>
filter(cowintraongoing == 1),
color = "red",
size = 2
)
p2
In the next layer, we’ll add some text to tell us which countries experience conflict in 1990. We can use the geom_text_repel() function from {ggrepel}:
library(ggrepel)
p3 <- p2 +
geom_text_repel(
data = data_90 |>
filter(cowintraongoing == 1),
aes(label = statenme),
size = 3,
nudge_y = .1,
min.segment.length = unit(0, "inches")
)
p3
That’s not too bad, but I want to add one more element to the plot. Countries experiencing conflict are highlighted using their names and the color red. I’d be nice to have an efficient way to tell people what red means in the plot. There’s a great package called {ggtext} that makes it really easy to edit and customize text in labels using HTML code. To install it, just write install.packages("ggtext") in the R console.
Check out how we can use it to really spice up our data viz below. By incorporating some HTML code, I’ve given R instructions to make “Civil War” in the title red, bold, and italic. This way viewers can use context clues (the color Civil War matches the color of certain highlighted data points) to understand that the countries being singled out are countries experiencing an ongoing civil war.
## open the package
library(ggtext)
## create a title object to keep things tidy
## and use paste0() to separate the normal text
## from the HTML part of the text
new_title <- paste0(
"Economic Health, Democracy, and ",
"<span style = 'color:red'>***Civil War***</span>"
)
## update the plot with the new title and then
## tell ggplot how to read the HTML code
p3 +
labs(
title = new_title
) +
theme(
plot.title = element_markdown()
)
My goal with the above examples wasn’t to cover everything you can possibly do to refine your plots. That would be impossible because your options are practically limitless. Instead, I wanted to leave you with an impression of just how many elements can be updated as you design your data visualizations. We haven’t even talked about the theme() function which gives you even more control over the look of your data viz.
You won’t master this in a single class, or a single semester for that matter. The key with all of this is to practice.