Writing narratives with data: An introduction to ggplot!
Katie Jolly
November 11, 2017
Setting up our project
I’m a huge fan of nonfiction books and documentaries. I’ve never loved a lot of the big Hollywood movies, but I do enjoy biopics. I find that they give me a good amount of the nonfiction that I enjoy and the drama that makes a blockbuster hit. But, I often find myself a little disappointed in the breadth of biopics available. Most feature a pretty narrow demographic group: straight, white men. Fivethirtyeight did a nice story on this pattern and published the data they used. So, we are going to use that data to write our own story with ggplot! We will start by recreating the graphics from the article ‘Straight Outta Compton’ is the rare biopic not about white dudes and then make some of our own!
Assumed knowledge
In this activity I’ve assumed a few things about your comfort level with R and RStudio, although I’ve designed it to be accessible to everyone. I assume you’ve at least seen the following things before:
RMarkdown (for presenting your work)
Packages (for a wider selection of functions)
If these are new to you, I’d suggest taking a look at the RMarkdown cheat sheet to learn more about the markdown structure and this R Packages article to learn about using packages.
There are times when I mention things about data wrangling (or preparing your data for analysis). I don’t assume knowledge of dplyr so I’ve provided code for you to use. If you’re interested in learning more, try reading this data wrangling cheat sheet or trying this dplyr tutorial for beginners. If you want to test your use of dplyr, try checking out my dplyr tutorial with bikeshare data!
Packages needed
First, we’ll need to install a few packages. Thetidyverse package includes both readr and ggplot2, two of the packages we will be using.
library(tidyverse)
library(ggthemes)Data from the web
Fivethirtyeight publishes a lot of the data they use on github! We’re going to use their biopics data. Fivethirtyeight data is also available in an R package call fivethirtyeight but I learned by using their github repo, so that’s the method I prefer. Either way will get you the same data!
# read in the data from github
biopics <- read_csv("https://raw.githubusercontent.com/fivethirtyeight/data/master/biopics/biopics.csv")# if you want to use the R package
library(fivethirtyeight)
data(biopics)Structure of the data
# I find glimpse to be a more readable version of head/str
glimpse(biopics)## Observations: 761
## Variables: 14
## $ title <chr> "10 Rillington Place", "12 Years a Slave", ...
## $ site <chr> "http://www.imdb.com/title/tt0066730/", "ht...
## $ country <chr> "UK", "US/UK", "US/UK", "Canada", "US", "US...
## $ year_release <int> 1971, 2013, 2010, 2014, 1998, 2008, 2002, 2...
## $ box_office <chr> "-", "$56.7M", "$18.3M", "-", "$537K", "$81...
## $ director <chr> "Richard Fleischer", "Steve McQueen", "Dann...
## $ number_of_subjects <int> 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 1, 3, 3, 3...
## $ subject <chr> "John Christie", "Solomon Northup", "Aron R...
## $ type_of_subject <chr> "Criminal", "Other", "Athlete", "Other", "O...
## $ race_known <chr> "Unknown", "Known", "Unknown", "Known", "Un...
## $ subject_race <chr> NA, "African American", NA, "White", NA, "A...
## $ person_of_color <int> 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0...
## $ subject_sex <chr> "Male", "Male", "Male", "Male", "Male", "Ma...
## $ lead_actor_actress <chr> "Richard Attenborough", "Chiwetel Ejiofor",...Visualizations
Notes about ggplot
ggplot adheres to the “grammar of graphics” (hence the gg!), or the idea that graphics should have a standard structure the same way that language does. The idea is to allow people to use the standard structure of plots to move beyond standard graphics. It was largely popularized by Hadley Wickham after being introduced by Leland Wilkinson in 1999. In 2016, ggplot2 was the fifth most downloaded package for R.
Essentially the grammar of graphics uses layering to create plots. You first start with your mapping of which variables map to which aesthetics on your plot. Then you specify the geom or method of displaying the data, such as a point or bar. After that you get into the design elements. That’s of course a very brief overview, but you’ll see how we can use that structure in this tutorial!
A great resource for ggplot is this Cheat Sheet and as with everything else in R, a little googling will get you a long way.
Looking at race
We will eventually create this graph!
First, we need to decide what variable maps to each aesthetic. In ggplot, aesthetics are essentially how variables are mapped to something visual, such as an axis or the color.
Setting up the axes
To start we need to calculate totals for each racial group in each year. But right now, our subject_race variable doesn’t quite tell us the information we want. We can combine the information from race_known and subject_race to get a variable with three levels: Unknown, White subjects, and Nonwhite subjects.
biopics <- biopics %>% # start with our original data
mutate(race_plotting = ifelse(race_known == "Unknown", "Unknown", ifelse(subject_race == "White", "White subjects", "Nonwhite subjects"))) # this says, if race_known is unknown, stop there and just mark unknown. But if we know the race, check to see if the race is coded as white. If it is, mark white and stop there. If not, mark nonwhite.
race_plot <- biopics %>%
group_by(year_release, race_plotting) %>% # we want our calculation of the total to be per racial group, per year
summarize(total = n()) # count up the observations in each group!
race_plot$race_plotting <- factor(race_plot$race_plotting, levels = c("White subjects", "Nonwhite subjects", "Unknown")) # change around the order of our factor levels for plotting purposesIn this graph, it loots like a count of total movies is the y axis and year_release is mapped to the x axis.
# first we need to create a variable for that!
ggplot(data = race_plot, aes(x= year_release)) # setting our x axis to year_release from the biopics table
Nice! Now we have the beginnings of a plot.
Adding a geom layer
Next, we need to add our geom layer. This is what the shapes on your plot are (line, point, bar, …..). We are using geom_bar!
ggplot(data = race_plot, aes(x = year_release, y = total)) +
geom_bar(stat = "identity") # stat="identity" argument allows us to map to a y aesthetic with geom_bar! 
Adding colors!
In the fivethirtyeight plot, colors correspond to whether or not the subject was white. We will modify our plot to show that as well (and try to pick the same colors in the process)!
ggplot(data = race_plot, aes(x= year_release, y = total)) +
geom_bar(aes(fill = race_plotting), stat = "identity") 
Closer! Now let’s add a title, subtitle, and axis labels! One quick note about the subtitle: fivethirtyeight filtered out some of their data. We have 761 biopics in this data so that’s the value we will use instead of 676.
Adding labels
ggplot(data = race_plot, aes(x= year_release, y = total)) +
geom_bar(aes(fill = race_plotting), stat = "identity") +
labs(title = "Biopic Subjects Are Mostly White",
subtitle = "Number of white and nonwhite subjects in 761 biopics since 1915",
x = NULL,
y = NULL)
Changing colors
Next, we will add a manual scale to change the colors! I searched around using the google color picker to find the closest matches.
ggplot(data = race_plot, aes(x= year_release, y = total)) +
geom_bar(aes(fill = race_plotting), stat = "identity") +
labs(title = "Biopic Subjects Are Mostly White",
subtitle = "Number of white and nonwhite subjects in 761 biopics since 1915",
x = NULL,
y = NULL) +
scale_fill_manual(values = c("#d847be", "#36bc6c", "#909392")) # I changed the color scale as well as the width of the bars so we can see the colors more easily
Changing the final look of the plot
Now we have pretty much all of the elements in place! The last things we will do will focus on the look of the plot.
First, we want to move the legend to the upper lefthand corner of the plot.
ggplot(data = race_plot, aes(x= year_release, y = total)) +
geom_bar(aes(fill = race_plotting), stat = "identity", width = 0.6) +
labs(title = "Biopic Subjects Are Mostly White",
subtitle = "Number of white and nonwhite subjects in 761 biopics since 1915",
x = NULL,
y = NULL) +
scale_fill_manual(values = c("#d847be", "#36bc6c", "#909392")) +
theme(legend.position = c(.2,.7),
legend.background = element_rect(fill="lightgray",
size=0.5,
linetype="solid",
colour ="darkgray"),
legend.title = element_blank())
One nice trick with ggplot is the package ggthemes. It has preloaded templates for different popular press outlets and plotting tools. Lucky for us, there’s a fivethirtyeight theme. We’ll use that to make our lives a little easier, but you could style it the same way with your own ggplot code! In this case, though, why reinvent the wheel?
Notice in the code below that theme_fivethirtyeight has to come before our theme arguments. We want to override a few defaults from ggtheme and R executes code sequentially. If you write them in the other order, you’ll be overriding our overrides.
ggplot(data = race_plot, aes(x= year_release, y = total)) +
geom_bar(aes(fill = race_plotting), stat = "identity", width = 0.6) +
labs(title = "Biopic Subjects Are Mostly White",
subtitle = "Number of white and nonwhite subjects in 761 biopics since 1915",
x = NULL,
y = NULL) +
scale_fill_manual(values = c("#d847be", "#36bc6c", "#909392")) +
theme_fivethirtyeight() +
theme(legend.position = c(.2,.7),
legend.background = element_rect(fill="#eaedeb",
size=0.5,
linetype="solid",
colour ="darkgray"),
legend.direction = "vertical",
legend.title = element_blank(),
legend.text = element_text(face = "italic"))
Ta-da! Done! That wasn’t so bad, right? On to our next viz!
Looking at gender
We will follow pretty similar steps for our next viz, so I’m going to move a little more quickly through the steps. Refer back to the previous explanations if you get stuck, or feel free to shout out questions!
First, take a look at our goal for this part of the activity!
Take a minute to think about how the variables are being mapped to different aesthetics in this plot. What is the x-axis? Y-axis? Color?
Setting up our data
gender_plotting <- biopics %>%
group_by(subject_sex,year_release) %>%
summarize(total = n()) # this is going to calculate the number of biopics about men and women separately for each yearSetting up our axes
ggplot(data = gender_plotting, aes(x = year_release, y = total))
Lets add our geom!
ggplot(data = gender_plotting, aes(x = year_release, y = total)) +
geom_bar(stat = "identity") # don't forget stat = "identity"
Adding colors and a legend
# changing the stacking order
gender_plotting$subject_sex <- factor(gender_plotting$subject_sex, levels = c("Male", "Female"))
ggplot(data = gender_plotting, aes(x = year_release, y = total, fill = subject_sex)) +
geom_bar(stat = "identity") 
We can already see a pretty clear pattern in the data and we’ve done the hard part of plotting! Now let’s make it a little prettier. We will use a lot of the same code from our last plot for the design aspects!
Design
In this section, I’ll introduce saving a plot as a variable so that we don’t have to type the same thing over and over.
# I again changed the width of the bars here
g <- ggplot(data = gender_plotting, aes(x = year_release, y = total, fill = subject_sex)) +
geom_bar(stat = "identity", width = 0.6) +
labs(title = "Biopic Subjects Are Mostly Male",
subtitle = "Number of male and female subjects in 761 biopics since 1915",
x = NULL,
y = NULL)
g
# add the right colors now!
g <- g +
scale_fill_manual(values = c("#f9cd52", "#4304a0"))
g
# format the legend in a better way and add the fivethirtyeight theme!
g <- g +
theme_fivethirtyeight() +
theme(legend.position = c(.2,.7),
legend.background = element_rect(fill="#eaedeb",
size=0.5,
linetype="solid",
colour ="black"),
legend.direction = "vertical",
legend.title = element_blank(),
legend.text = element_text(face = "italic"))
g
Nice! Now we’ve finished recreating the two graphics from the article, but there are so many others. We’ll walk through creating one more. To combine the previous two graphs (and learn a new feature of ggplot!), we’ll make a plot of gender and race in biopics. The article claims that women of color are by far the most underrepresented, so this plot will allow us to see that pattern.
Race and gender, together
For this plot, we don’t have an image of the goal we are working towards. In the end we want to have two side by side plots of race in biopics. To do this we will learn about facet, a feature in ggplot that allows you to create side by side plots separated by a categorical variable (such as sex).
But, we have a slight numerical problem. As we saw in the gender plot, men are way more represented. So it will be hard to compare racial patterns between men and women when we are comparing on different scales. Instead we want to use percentages, so we will fix that problem with dplyr!
race_and_gender <- biopics %>% # start with our original data
group_by(subject_sex, race_plotting) %>% # group it by race and gender
summarise(count = n()) %>% # calculate the number in each group
mutate(prop = count/sum(count)) # calculate relative frequencyWe are going to make a cooler (in my opinion!) version of the bar plot. We’ll take advantage of ggplot to combine geom elements to create something new! We are essentially going to create a lollipop chart.
In this section I’m not going to write the code for the plots as methodically, but I’ll be sure to leave plenty of comments!
ggplot(race_and_gender) + # we are going to set our aesthetics in the geoms this time!
geom_linerange(aes(x = race_plotting, ymin = 0, ymax = prop-0.01), size = 1) + # add the lines that lead up to the points
geom_point(aes(x = race_plotting, y = prop, color = race_plotting), size = 5) + # add a point at the proportion value on the x axis
geom_text(aes(x = race_plotting, y = prop + 0.07, label = round(prop, 2))) + # write the proportion 0.07 points above the point (so you can see it better! no overlap)
facet_wrap(~subject_sex) + # separate by gender
scale_color_manual(values = c("#d847be", "#36bc6c", "#909392")) + # change colors to match the earlier plot
theme_fivethirtyeight() +
theme(legend.background = element_rect(fill="#eaedeb",
size=0.5,
linetype="solid",
colour ="black"),
legend.direction = "vertical",
legend.title = element_blank(),
legend.text = element_text(face = "italic")) +
labs(title = "Racial representation doesn't vary by gender", subtitle = "Based on data from 761 biopics since 1915", y = "Proportion of biopics")
To get an idea of magnitude, though, we will also plot the raw values in a similar way! But now we can read those values more critically than we might have before. These two plots together help to tell a very compelling story!
# the code for this plot is almost the same as the previous one. the only difference is the y variable! and the labels.
ggplot(race_and_gender) + # we are going to set our aesthetics in the geoms this time!
geom_linerange(aes(x = race_plotting, ymin = 0, ymax = count-0.1), size = 1) +
geom_point(aes(x = race_plotting, y = count, color = race_plotting), size = 5) +
facet_wrap(~subject_sex) +
scale_color_manual(values = c("#d847be", "#36bc6c", "#909392")) +
theme_fivethirtyeight() +
theme(legend.background = element_rect(fill="#eaedeb",
size=0.5,
linetype="solid",
colour ="black"),
legend.direction = "vertical",
legend.title = element_blank(),
legend.text = element_text(face = "italic")) +
labs(title = "Biopics heavily favor white, male subjects", subtitle = "Based on data from 761 biopics since 1915", y = "Number of biopics")
Your turn!
Try to create your own data story using the biopics data. Some ideas are:
How does the distribution of box office earnings vary by demographic group?
How does occupation of subject vary by demographic group?
Does country of production have an impact on demographics of the subject?
How about movies with multiple subjects? (some movies are in the table twice for that reason!)
Happy coding!