Pretty plots with ggplot2

How to create ggplot2 objects

You can add as much data in the inital function call as you want. All of these work, but the final version is the only “complete” object that fully specifies the data used for the plot.

ref <- ggplot()
ref <- ggplot(data)
ref <- ggplot(data, aes(x = bodywt, y = sleep_total)) 
# This is the same thing as the line above:
ref <- ggplot(data, aes(x = data$bodywt, y = data$sleep_total))

To store an object (to add to it later/plot it on demand), just give it a reference. Simply typing the reference will display the plot (if you’ve provided enough information to make it.)

ref

As you can see, we haven’t specified everything we need yet. There are 3 components to making a plot with a ggplot object: your data, the aesthetic mappings of your data, and the geometry. If you are missing one, you won’t get a functional plot.

Your data should be a dataframe with everything you want to plot. Note that it is possible to put data from multiple sources (ie. different dataframes) in the same plot, but it’s easier if everything is in the same 2-dimensional dataframe.

ref <- ggplot(data)

The aesthetic mappings tell ggplot2 how to interpret your data. Which values in your dataframe are the y-values, x-values, what should be used for colors, etc.

ref <- ggplot(data, aes(x = bodywt, y = sleep_total))

The geometry is the actual stuff that goes on the plot. You can specify any geometry as long as you have supplied the values it needs. If you’ve specified the required aesthetic mappings (which data corresponds to x, y, etc.), all you need to do is tell ggplot2 to create a certain geometry- for instance a scatterplot.

Just add the geometry you want to your object. In this case, we are making a scatterplot.

ref <- ggplot(data, aes(x = bodywt, y = sleep_total)) + geom_point()
ref

Adding additional elements

All you need to do to add more information to your plot/change things is add on more elements. Lets add a logarithmic scale on the x axis.

ref <- ggplot(data, aes(x = bodywt, y = sleep_total)) + geom_point() + scale_x_log10()
ref

Lets add a smoothed mean.

ref + geom_smooth()

## `geom_smooth()` using method = 'loess'

You can also specify aesthetics inside the call to create geomtery.

ggplot(data) + geom_point(aes(x = bodywt, y = sleep_total)) + scale_x_log10() + geom_smooth()

Error: stat_smooth requires the following missing aesthetics: x, y

Why didn’t that work? This is because when we specfy aesthetics inside a call to geomtery it only applies for that layer (only geom_point got the x and y values). The only information that gets passed to all geometery calls is aethetics specified in the initial creation of the ggplot object.

So if we wanted that to work, we’d have to do this:

ggplot(data) + scale_x_log10() + 
  geom_point(aes(x = bodywt, y = sleep_total)) +
  geom_smooth(aes(x = bodywt, y = sleep_total))

## `geom_smooth()` using method = 'loess'

It’s important to note that geometry will automatically use any aesthetic mappings that it understands, and ignore ones it doesn’t. So if you specify as much stuff as you can in the inital call that can be used, it’ll save you work.

Like this:

ggplot(data, aes(x = bodywt, y = sleep_total)) + scale_x_log10() + geom_point() + geom_smooth()

## `geom_smooth()` using method = 'loess'

Adding additional mappings

Let’s follow up with a few very common plot/geometry types and mappings you might be interested in:

These x and y mappings (and the log scale on the x axis will be used for all later plots).

plot <- ggplot(data, aes(x = bodywt, y = sleep_total)) + scale_x_log10()
plot + geom_point()

plot + geom_point(aes(color = vore))

plot + geom_point(aes(color = log(brainwt)))

plot + geom_point(aes(color = log(brainwt))) + scale_color_gradient2()

plot + geom_point() + scale_y_continuous(limits = c(5, 15))

## Warning: Removed 23 rows containing missing values (geom_point).

plot + geom_point(aes(size = sleep_rem, alpha = sleep_rem)) + 
  xlab("this is our x axis") + ylab("this is our y axis") + ggtitle("title") + scale_alpha("our legend")

## Warning: Removed 22 rows containing missing values (geom_point).

plot + geom_point(aes(shape = vore), size = 6, color = "orange")

## Warning: Removed 7 rows containing missing values (geom_point).

plot + geom_point() + facet_wrap(~vore)

Other types of plots

All other types of plots work identically to the scatterplot - let’s see a few examples…

ggplot(data, aes(x = vore, y = sleep_total, fill = vore)) + geom_boxplot()

ggplot(data, aes(x = bodywt, y = brainwt, group = vore, color = vore)) +
  geom_line() + scale_x_log10() + scale_y_log10()

## Warning: Removed 8 rows containing missing values (geom_path).

ggplot(data, aes(x = sleep_total, fill = vore)) + geom_density(alpha = 0.5)

ggplot(data, aes(x = vore, y = sleep_total)) + geom_violin()

ggplot(data, aes(x = vore)) + geom_bar()

Changing how plots look

Chances are, you might not be instantly “in love” with ggplot2’s default theme. Fear not - there are several other themes you can use to choose from. Using the gapminder dataset as an example, let’s check out several different themes.

library(gapminder)
ggplot(gapminder, aes(x=year, y=lifeExp, group=year)) + geom_boxplot()

library(gapminder)
ggplot(gapminder, aes(x=year, y=lifeExp, group=year)) + geom_boxplot() +
    theme_classic()

library(gapminder)
ggplot(gapminder, aes(x=year, y=lifeExp, group=year)) + geom_boxplot() +
    theme_light()

library(gapminder)
ggplot(gapminder, aes(x=year, y=lifeExp, group=year)) + geom_boxplot() +
    theme_void()