Plotting
Plotting in R
A lot of analyses are represented by figures in the final papers. Additionally, visualization of data may allow better scrutiny of data patterns. As such, creating figures in R, or ‘plotting’ (because we are making plots).
The way plots are created varies, but two large methods stand out. The first we will discuss is implemented by the {ggplot2} package. This package is based on the grammar of graphics, a landmark work on statistical graphics (ergo ‘gg’ for ‘grammr of graphics’). The second method we will discuss is the base R plotting, which is part of the R {graphics} package.
As example data, we will use the package {palmerpenguins}, which contains data on 344 penguins from three islands in the Palmer Archipelago, Antarctica (more here).
We will also load the {ggplot2} package. The {graphics} package is installed and loaded by default.
{ggplot2}
Basically, the grammar of graphics provides a foundation that describes how any statistical graphic can be build up, which is based on layers. This layered system gives rise to how the package {ggplot2} allows us to create figures (i.e. plots).
Layer 1: the data
The first layer we specify is the data layer. In this layer, we specify our data and the mapping of the aesthetics, meaning we define what our x-axis will portray, the y-axis, what colours, etc. will portray. This first layer is called using ggplot():
Here, we specified the first argument (data = penguins) to be our dataset. For the second argument (mapping = aes()), we supply a function: aes(). This function allows us to specify what will make up the aesthetics of the graph: the x-axis will be ‘body_mass_g’ and the y-axis will be ‘flipper_length_mm’. Because we did not supply any other layer, we can see that the figure remains largely empty, only having some automatically set limits and titles for the axes.
Layer 2: the geometries
Now we can add the geometries. We are offered a large number of geometries by {ggplot2}, with some basics including points, lines, bars, histograms, rectangles, segments, steps, and ribbons. Each geometry is called by a geom_...() function, such as geom_point() and geom_line().
For our figure, we will add points.
Warning: Removed 2 rows containing missing values or values outside the scale range
(`geom_point()`).
You may notice that layers in these plots are added together by using a plus (+) sign. Additionally, the warning you is just {ggplot2} telling you there were some NA’s that could not be drawn in the plot. Note that this error is also given if you have some values that are outside of the boundaries of your plot.
Within each geometry, we can edit many things, such as the colour, fill (i.e. the colour of shapes that contain an inside), size, alpha (i.e. transparency), and shape (or linetype for lines). Much more is possible, but can be learned from the help functions for each geometry.
Note that any geometry is applied in subsequent order. If we add a trendline (geom_smooth(method = "lm", formula = "y ~ x")), this will be drawn on top of the points if we specify it after geom_point() or before if we specify it before geom_point().
Let’s do this now:
Answer
Warning: Removed 2 rows containing non-finite outside the scale range
(`stat_smooth()`).Warning: Removed 2 rows containing missing values or values outside the scale range
(`geom_point()`).
We specified shape as a number, for which you can find online what different numbers mean. We specified the colour as a hex code, but you can also use the rgb() function for RGB specification, or English language for a number of colours that can be found here. Lastly, we specify the British English colour instead of the American English color. {ggplot2} will accept both, for arguments and for functions (e.g. scale_color_manual() and scale_colour_manual() both exist and do the same thing).
Layer 3: the scaling
For the third layer, we will apply scaling, which means we will edit anything that can be a scale (e.g. axes, colours, fill).
Note that we deviate from the grammar of graphics in building our plot. The order in which we specify our layers does not adhere to the grammar of graphics. Luckily, in the background, {ggplot2} ignores this, only keeping the order in which we called geometries for geometries, but for instance not drawing theming prior to geometries, even if we specify so.
In our current plot, we have two scales: the x-axis and y-axis. We can edit these using scale functions that concur with the scale type (i.e. continuous, discrete, etc.). In our case, both scales are continuous, so we will use scale_x_continuous() and scale_y_continuous(). For other types of scales, we might also use scale_x_discrete(), scale_x_date(), etc.
Answer
ggplot(penguins, aes(x = body_mass_g, y = flipper_length_mm)) +
# Geometries
geom_point(shape = 5) +
geom_smooth(method = "lm", formula = "y ~ x", colour = "#B58900", fill = "#B58900", alpha = 0.3) +
# Scaling
scale_x_continuous(name = "Body mass (grams)",
limits = c(2500, 6500),
breaks = seq(2500, 6500, 500),
labels = prettyNum(seq(2500, 6500, 500), big.mark = ",")) +
scale_y_continuous(name = "Flipper length (millimeters)",
limits = c(170, 240),
breaks = seq(170, 240, 10)) Warning: Removed 2 rows containing non-finite outside the scale range
(`stat_smooth()`).Warning: Removed 2 rows containing missing values or values outside the scale range
(`geom_point()`).
In the scale functions, we can give the name of the axis, the limits of the axis, at which points there should be ticks (breaks), and the labels given to each tick.
Layer 4: labels
In the next layer, we will apply labels to the plot, which includes the title, subtitle, axes labels, and legends.
For the axes labels, we can use xlab() and ylab() respectively. We won’t need these as we specified the name already in the scaling function, but they might be useful if you do not need the scale function. The title can be specified using the ggtitle() function, which also allows you to specify a subtitle. Alternatively, this can be done through the labs() function, which allows much more control, such as a tag (for multipanel labels where you want to tag A, B, C, etc.), a caption, and alt text (for accessibility purposes).
For the plot we’re building, we will aply a title and a subtitle.
Answer
ggplot(penguins, aes(x = body_mass_g, y = flipper_length_mm)) +
# Geometries
geom_point(shape = 5) +
geom_smooth(method = "lm", formula = "y ~ x", colour = "#B58900", fill = "#B58900", alpha = 0.3) +
# Scaling
scale_x_continuous(name = "Body mass (grams)",
limits = c(2500, 6500),
breaks = seq(2500, 6500, 500),
labels = prettyNum(seq(2500, 6500, 500), big.mark = ",")) +
scale_y_continuous(name = "Flipper length (millimeters)",
limits = c(170, 240),
breaks = seq(170, 240, 10)) +
# Labels
ggtitle("Relation between penguin body mass and flipper length", subtitle = "Data from the Palmer Archipelago penguins") Warning: Removed 2 rows containing non-finite outside the scale range
(`stat_smooth()`).Warning: Removed 2 rows containing missing values or values outside the scale range
(`geom_point()`).