2.3 Visualizing data using ggplot2

Now that you’ve grown familiar with the basics of data manipulation using dplyr and reshape2, the next logical step in our workflow is to visualize the data that we’ve exhaustively prepared. Be aware that there are plenty of marvellous graphics packages available in R depending on which tasks you’d like to perform. Within the scope of our short course, ggplot2 will do just fine not least because it is closely linked to the previously performed manipulation tasks (maybe because of the same author).

In ggplot2, every plot we want to draw consists of incremental calls on how to represent the individual plot components (basically x and y). This means a much more consistent way of building visualizations as compared, for instance, to the much more conventional style of lattice, but it also means that things are rather different from what you might have learned about syntax and structure of (plotting) objects in R so far.

Note also that the brief examples shown below are mainly taken from another MARA course about Creating publication quality graphs using R. Feel free to browse the code tutorial or the accompanying course slides in case you’re not entirely satisfied with the rather shallow introduction given here or if you wish to learn more about lattice-style figures.

Scatter plots

Scatter plots are produced by invoking geom_points(). If we wanted to plot variables ‘price’ and ‘carat’ from diamonds against each other, the ggplot() call would be as follows.

library(ggplot2)

ggplot(aes(x = carat, y = price), data = diamonds) + 
  geom_point()

Figure 2.1: A basic scatter plot created with ggplot2.

Let’s look at the above code in a little more detail. The first line is the fundamental definition of what we want to plot. We provide the ‘aesthetics’ for the plot (aes()) and state that we want the values on the x-axis (y-axis) to represent ‘carat’ (‘price’). The data to take these variables from is our diamonds dataset. That’s basically it, and this will not change a hell of a lot in the subsequent plotting routines.

What will change in the code chunks that follow is how we want the relationship between these variables to be represented in our plot. This is done by defining so-called geometries (geom_*()). In this first case we stated that we want the relationship between x and y to be represented as points, hence we used geom_point().

For the sake of practicing, let’s add another layer to our plot. In order to provide the regression line, for instance, we need a function called stat_smooth(). Note that we also specify the smoothing method and the line colour, size and linetype. se = FALSE tells the function not to display confidence intervals for the moment. See ?stat_smooth for a more detailed overview of costumization possibilities.

ggplot(aes(x = carat, y = price), data = diamonds) + 
  geom_point() +
  stat_smooth(method = "lm", se = FALSE, 
              colour = "red", size = 1.5, linetype = "dashed")

Figure 2.2: A ggplot2 scatter plot incl. regression line.

If we wanted to provide a plot showing the relationship between ‘price’ and ‘carat’ in panels representing the quality of the diamonds, we need what in ggplot2 is called facetting (similar to panels in lattice). To achive this, we simply repeat our plotting call from earlier and add another layer to the call which does the facetting. Note that this time, confidence intervals are included via se = TRUE and fill-ed grey.

ggplot(aes(x = carat, y = price), data = diamonds) + 
  geom_point() +
  stat_smooth(method = "lm", se = TRUE, 
              colour = "red", fill = "grey35") + 
  facet_wrap(~ cut)

Figure 2.3: A facetted scatter plot created with ggplot2.

One thing that some people dislike about the default settings in ggplot2 is the grey background of the plots. This grey background is, in my opinion, a good idea when colors are involved as it tends to increase the contrast of the colors. If, however, the plot is a simple black-and-white scatter plot, a white facet background seems more reasonable. We can easily change this using a pre-defined theme called theme_bw(). Note that in the following, we also change the axis titles (labs()) and the number of rows (nrow) and columns (ncol) into which the facets should be arranged – simple and straightforward, isnt’t it?

ggplot(aes(x = carat, y = price), data = diamonds) + 
  geom_point(colour = "grey65") +
  stat_smooth(method = "lm", se = TRUE, 
              colour = "black", fill = "grey35") +
  facet_wrap(~ cut, nrow = 2, ncol = 3) +
  labs(x = "Carat", y = "Price ($)") +
  theme_bw()

Figure 2.4: A facetted scatter plot created with ggplot2 and using a modified theme.

Box and whisker plots

Box and whisker plots (in the following ‘boxplots’) drawn with ggplot2 look quite nice and, in contrast to lattice, do not require the user to exhaustively modify graphical parameter settings in order to get an acceptable result. But see for yourselves (using the default settings).

## basic frame
p <- ggplot(aes(x = color, y = price), data = diamonds)

## add boxplot
p_bw <- p + geom_boxplot()

## print
print(p_bw)

Figure 2.5: A basic boxplot created with ggplot2.

Note that this time, we stored the basic frame of our plot in a variable p and added the desired geom_boxplot() layer afterwards. In order to keep a clear structure, this is possibly the ideal way to create graphics using ggplot2 - so stick to it!

As we’ve already seen, facetting and modifying the default theme require also just one more line of code each.

## add faceted boxplots
p_bw <- p + 
  geom_boxplot(fill = "grey90") +
  facet_wrap(~ cut) +
  theme_bw()

print(p_bw)

Figure 2.6: A facetted boxplot created with ggplot2 and using a modified theme.

Histograms and density plots

Much like with boxplots, ggplot2 produces quite nice histograms and density plots when using the default settings.

## new basic frame
p <- ggplot(aes(x = price), data = diamonds)

## add histogram
p_hist <- p + geom_histogram()

print(p_hist)

Figure 2.7: A basic histogram produced with ggplot2.

When working with density plots in ggplot2, it is quite easy to fill the area under the curve which really contributes to the visual representation of the data.

## add densityplot
p_dens <- p +  geom_density(fill = "black", alpha = 0.5)

print(p_dens)

Figure 2.8: A basic density plot produced with ggplot2

In addition to theme_bw(), there’s a lot of different theme() settings that let you customize your plots at will. For instance, let us slightly rotate the tick labels on the x-axis and move them horizontally a little bit. element_text() is an essential function to achieve such things as it allows you to modify, among others,

• font family and face (e.g. ‘bold’),
• text colour and size,
• horizontal (hjust) and vertical (vjust) justification as well as
• angle of the tick labels.

(In case you prefer American English, element_text() – or for that matter any ggplot2 function – also understands color instead of colour ☺)

p_dens <- p +
  geom_density(fill = "black", alpha = 0.5) +
  facet_grid(color ~ cut) +
  theme_bw() +
  theme(axis.text.x = element_text(angle = 45, hjust = 1))

print(p_dens)

Figure 2.9: A facetted density plot conditioned according to 2 variables created with ggplot2.