Dynamic Plotting

• Overview
• The Data
• Problem Sets
  • 0. Make a Plot
  • 1. Adding Labels to a ggplot
  • 2. Isolating Elements
  • 3. Make it Interactive
  • 4. Make it Accessible
  • 5. Try Something New

Overview

We have now learned how to make some nicer looking plots, but there is still room for improvement. All plots should tell a story, and you can use specific elements to really draw a viewers attention to that story. Today I’ll be showing a few tools to do that. You won’t need everything you see today for every plot, but they are important tools to have in your toolbox. Note that while I will be using a scatter plot throughout the worksheet today, the same tricks can be applied to any kind of plot.

The Data

Today we will be using data from the Environmental Protection Agency (EPA) Fuel Economy data. It comes with the ggplot2 package, and we can load it into the environment using the following (only after you have loaded ggplot2):

library(ggplot2)

mpg = data.frame(mpg)

Here is a rundown of what the variables are:

manufacturer

manufacturer name

model

model name

displ

engine displacement, in liters

year

year of manufacture

cyl

number of cylinders

trans

type of transmission

drv

the type of drive train, where f = front-wheel drive, r = rear wheel drive, 4 = 4wd

cty

city miles per gallon

hwy

highway miles per gallon

fl

fuel type

class

“type” of car

Problem Sets

0. Make a Plot

Before we start making a fancy plot, we need a basic plot. I will provide the code for one here so everyone is starting from the same place.

ggplot(mpg) +
  aes(x = displ, y = hwy) +
  geom_point(size = 1.5) +
  labs(
    x = "Engine Displacement",
    y = "Highway MPG",
    title = "Highway MPG by Engine Size",
    caption = "Source: https://fueleconomy.gov/"
  ) +
  theme_minimal() +
  theme(plot.title = element_text(face = "bold"))

1. Adding Labels to a ggplot

In plot terminology, labels are text we add to a plot to explain some element; we add labels to the axes for example. We can also add labels to the data itself if it will help us tell our story. It’s easy to go too far though, so this is often used sparingly.

There are several methods to add labels, but I’m going to show you the most general one. We start by adding a new column to our mpg dataframe, which we’ll call labels.

Let’s see how we can incorporate these labels.

If all went according to plan, we should have labels! Not the nicest to look at though. We need to be selective about where we add our labels. Let’s cut back a bit.

2. Isolating Elements

We’re getting closer. Adding text to a plot can he helpful, but only for a small handful of points. If we want to highlight a whole group of points, using something like color is often more helpful. Let’s try using color to highlight our midsize cars. We can do this by adding another layer on top of our current points, but only for those we want to highlight.

First we need to remove the label argument and label geom we added in the last section. Next, I add a new geom_point() below the one we already have. If we are thinking in terms of layers, adding a new geom below the other will place this layer on top of the regular points. I’m going to highlight the cars in the data where the model is a “corvette” (a famous sports car). I also added a subtitle to explain the highlights. You can see the results below:

ggplot(mpg) +
  aes(x = displ, y = hwy) +
  geom_point(size = 1.5) +
  # This geom is new, and let's us highlight specific points
  geom_point(data = mpg[mpg$model == "corvette",], aes(x = displ, y = hwy), color = "red", size = 3) +
  labs(
    x = "Engine Displacement",
    y = "Highway MPG",
    title = "Highway MPG by Engine Size",
    # add subtitle to explain highlights
    subtitle = "Corvettes Highlighted in Red",
    caption = "Source: https://fueleconomy.gov/"
  ) +
  theme_minimal() +
  theme(plot.title = element_text(face = "bold"))

Now our plot highlights the corvettes in our data with a specific color! Let’s go one step further and add a call-out for those highlights. Previously we added labels to a set of points, but we can also arbitrarily add text as a new layer in ggplot using the annotate() function like the following:

ggplot(mpg) +
  aes(x = displ, y = hwy) +
  geom_point(size = 1.5) +
  # This geom is new, and let's us highlight specific points
  geom_point(data = mpg[mpg$model == "corvette",], aes(x = displ, y = hwy), color = "red", size = 3) +
  labs(
    x = "Engine Displacement",
    y = "Highway MPG",
    title = "Highway MPG by Engine Size",
    # add subtitle to explain highlights
    subtitle = "Corvettes Highlighted in Red",
    caption = "Source: https://fueleconomy.gov/"
  ) +
  # add annotation
  annotate(geom = "label", x = 6.25, y = 28, label = "Corvettes", color = "red") +
  theme_minimal() +
  theme(plot.title = element_text(face = "bold"))

Such call-outs can be a great way to help tell the story in your data. In this case we are showing how the high-powered sports cars (corvettes) seem to stand as outliers from the trend we see for most other cars.

3. Make it Interactive

Making an interactive plot can be a complex task and often requires a mix of data science and web development skills. We are going to be taking the easier route and relying on our new-found ggplot capabilities. Using the plotly package in R, we can quickly turn ggplots into interactive versions of themselves.

First we will need to install plotly:

install.packages("plotly")

After that we can load in in using library():

library(plotly)

plotly relies on the same layer-based logic as ggplot, but is slightly different. It is completely possible to build a plotly plot from scratch like a ggplot. For example:

plot_ly(mpg, type = "scatter", mode = "markers") |>
  add_markers(x = mpg$displ, y = mpg$hwy) |>
  layout(title = "Highway MPG by Engine Size",
         xaxis = list(title = "Engine Displacement"),
         yaxis = list(title = "Highway MPG"))

However, it also has a very helpful function called ggplotly() which we can wrap around our ggplot to convert it automatically:

mpg_ggplot = ggplot(mpg) +
  aes(x = displ, y = hwy) +
  geom_point(size = 1.5) +
  labs(
    x = "Engine Displacement",
    y = "Highway MPG",
    title = "Highway MPG by Engine Size",
    caption = "Source: https://fueleconomy.gov/"
  ) +
  theme_minimal() +
  theme(plot.title = element_text(face = "bold"))

ggplotly(mpg_ggplot)

Ta-Da! Quick and easy interactive plot. By default we can hover over the data points to get a readout of where the points sit on our X and Y axis. We can also add custom information using a little trickery. We can add a fake aesthetic in our aes() called “text” and write a custom message for our pop-ups. This will rely on some information about working with text and HTML that we won’t cover until later, but I want you to know it is possible.

mpg_ggplot = ggplot(mpg) +
  aes(x = displ, y = hwy,
      text = paste0("CAR INFO:</br></br>",
                    "Engine Displacement: ", displ,
                    "</br>Highway MPG: ", hwy,
                    "</br>Model: ", model)
      ) +
  geom_point(size = 1.5) +
  labs(
    x = "Engine Displacement",
    y = "Highway MPG",
    title = "Highway MPG by Engine Size",
    caption = "Source: https://fueleconomy.gov/"
  ) +
  theme_minimal() +
  theme(plot.title = element_text(face = "bold"))

ggplotly(mpg_ggplot, tooltip = "text")

That’s handy. The same process can work for most plot types we have covered so far in class.

4. Make it Accessible

The last thing I want us to think about regarding out plots is accessibility. Accessibility broadly is an area of study dedicated to making sure people of all different abilities can equally access your content. Many people make it their whole career to study good accessibility practices (like our own Dr. Cao in SDS!). Today I will give you a few things to keep in mind while making your plots which move you toward best practices.

Color Blindness

First is to be aware of your color choices. There are many different ways in which the colors you choose may be difficult for some people to interpret. We want to make sure our color palette is as clear as possible for everyone. A tool I like to help with this is Viz Palette. It will help you build a color palette, and then see what it would look like if you had various difficulties seeing color.

Resolution

Not only do high resolution plots look better, they are also easier to understand for people with vision difficulties. There are some tricks in R for getting nice crisp plots. The first is to export them in a “vector” rather than “raster” format. A vector format essentially converts your plot into a long formula which a computer can use to draw the plot no matter what size it is. This is compared to a raster format which has a certain number of pixels, or dots, and it can never be any larger than that.

There aren’t many tricks for this, aside from this: when possible, export your static (non-interactive) plots as an SVG or PDF file. These are both options when you click the “Export” button in the plots pane to the right. Just know that sometimes whatever service you are using won’t accept them, and you’ll need to use something else.

Interactive plots are already math based, and should always be able to scale. They are far less compatible with many systems, however.

Alt Text

Whenever you are working online, it is possible to include certain metadata along with your plots. One of the most important you can include is called “Alt Text.” This text is typically a description of whatever the thing is it is attached to. This alt text is used by software called screen readers, which will use the text to describe the thing to someone who cannot see the screen.

Whenever you are authoring a R Markdown or quarto document, you can include alt text for the images you include. For example, you can include an image using the following syntax:

![Cpation for image](./path/to/imag.svg "Alt text that describes the image")

For example, try hovering your mouse over this plot for a few seconds:

![Alt Text Example](img/alt_text.png "This plot shows engine displacement on the X axis compared against the highway miles per gallon on the Y axis. Corvette cars are outliers wither higer miles per gallon despire their larg engines")

5. Try Something New

For the last part of this worksheet, take a look at our mpg dataframe and pick some aspect of it you would like to visualize. I encourage you to look at From Data to Vis and use the decicion tree there to help pick what plot type fits the data. Once you have decided on a plot type, click on its icon, read through its desciprtion, and then scroll down to the “R Graph Gallery” to see some ways you can make it in R. If you feel it is a good fit for the data, try to make one! If not, go back and pick a different type.