Legends on graphs and charts

Exercise 6.3.1a.

These notes about legends in graphs/charts supplement the text in Chapter 6.

• Introduction
• Legend essentials
• Legend from barplot()
  • Alter the y-axis to make room
  • Pass legend parameters via barplot()
  • Alter the x-axis to make room
• Place the legend in the margin
  • Marginal legend with a bar plot
  • Place a legend in the bottom margin
  • Legends in the margin of scatter/line plots

A legend is a tool to help explain a graph. You are most commonly going to want to add one to a bar chart where you have several data series. You’ll also want to add one to a line or scatter plot when you have more than one series. Essentially you use a legend to help make a complicated plot more understandable.

In R you can add a legend to any plot using the legend() command. You can also use the legend = TRUE parameter in the barplot() command. The barplot() command is the only general plot type that has a legend parameter (the others need a separate legend).

The legend() command has a host of parameters, which can be tweaked to produce the finished article. Generally, the most difficult part is making room on the chart for the legend itself!

The legend() command has a wealth of parameters at its disposal. This gives it a great deal of flexibility and customizability <sic> but this also makes it daunting and hard to get to grips with.

The legend() command has the following general form:

legend(x, y = NULL, legend, col, pch, lty, lwd, fill, border, bty, ncol, y.intersp)

There are a number of other parameters that I’ve not listed. The ones here are the most essential.

When adding a legend you need to make sure that the items in the legend() command match the parameters you set in the plotting command. It helps to specify pch, col, lty and so on explicitly in the plotting command, as you can match the parameters more easily than if you relied on the defaults.

The barplot() command is the only general plotting command that has a legend parameter. You can pass additional parameters to the legend using the args.legend parameter, as you’ll see shortly.

The barplot() command allows use of a legend parameter, which calls legend() with its basic settings. You can pass parameters to the legend() command by adding args.legend and giving the details as a list().

The biggest problem is usually how to make appropriate space for the legend to fit in the plot window. There are two main options:

• Alter the axis size to give extra room (vertically or horizontally).
• Place the legend into the plot margin.

The following examples use a matrix dataset that gives the abundance of some butterfly species at a site over several sample years:

> bf
        1996 1997 1998 1999 2000
M.bro     88   47   13   33   86
Or.tip    90   14   36   24   47
Paint.l   50    0    0    0    4
Pea       48  110   85   54   65
Red.ad     6    3    8   10   15
Ring     190   80   96  179  145

> class(bf) # Check you have a matrix
[1] "matrix"

You can get the sample data here: butterfly.RData.

Note that the xlim parameter in this example set the axis from 0 to 45. Each bar in the barplot() takes up a space, so you need to allow about 1 unit per bar plus a bit extra for the legend itself.

Note also that the colparameter set the colors for the plot, which were passed automatically to legend() without requiring the args.legend parameter. Of course if you wanted other parameters (such as supressing the legend box) you would require the args.legend parameter.

Altering the x-axis or y-axis size to accommodate the legend is a fairly simple matter. Sometimes however you want a legend to be at the bottom or even the left, and however you alter the axes you will not make space!

What you need is to be able to place the legend in the margin of the plot, so that it does not overlap the plotting zone at all. To do this you need to tweak the graphical parameters via the par() command.

The general running order is as follows:

1. Set the plot margins as you need to give the space in the required margin.
2. Make your plot (any plot).
3. Reset the graphical parameters back to defaults.
4. Now set the plot margins to 0 and at the same time set plotting to allow “overplot”.
5. Use legend() to place the legend where your extra-large margin is.
6. Reset the graphical parameters back to defaults.

Steps 3 and 6 are not absolutely essential but preferable, as you can get into an awful mess if you forget the current settings.

To make the plot margin larger use par(oma = c(b, l, t, r)) where b, l, t, r are values for the margin sizes at the bottom, left, top and right. For example:

> opar = par(oma = c(0,0,0,4)) # Large right margin for plot
> mycols = c("tan", "orange1", "magenta", "cyan", "red", "sandybrown")
> barplot(bf, beside = TRUE, col = mycols)
> par(opar) # Reset par

The bottom margin of a plot can present some slight difficulties because the axis labels get in the way. The easiest solution is to use the inset parameter and shift the legend outwards (use a small negative value).

Start by making space in the bottom outer margin, then make the basic plot:

> opar = par(oma = c(2,0,0,0))
> barplot(bf, beside = TRUE, col = cm.colors(6))
> par(opar) # Reset par

If you use a line plot or a scatter plot you’ll have to contend with different plotting characters and line types. It is easier to set the parameters explicitly in the plot() command so that they are more easily matched in the legend().

In the following example a matplot() is used to create a multiple-series line plot. The legend is placed in the right margin.

> t(bf) # Rotate the data as matplot() takes columns as series
     M.bro Or.tip Paint.l Pea Red.ad Ring
1996    88     90      50  48      6  190
1997    47     14       0 110      3   80
1998    13     36       0  85      8   96
1999    33     24       0  54     10  179
2000    86     47       4  65     15  145

> mycols = c("tan", "orange1", "magenta", "cyan", "red", "sandybrown")
> opar = par(oma = c(0,0,0,5.5)) # Set right margin

## Plot without axes (because axis labels are numbers)
> matplot(t(bf), ylab = "", type = "b", pch = 1:6, lty = 1:6,
                axes = FALSE, lwd = 2, col = mycols)

> axis(2) # Use default y-axis

# Custom x-axis using years as labels
> axis(1, at = 1:5, labels = colnames(bf))

> box() # Bounding box around plot

> title(ylab = "Abundance", xlab = "Year")
> par(opar) # Reset par

> opar = par(oma = c(0,0,0,0), mar = c(0,0,0,0), new = TRUE)

# Legend pars to match matplot
> legend(x = "right", legend = rownames(bf), col = mycols,
         pch = 1:6, lty = 1:6, bty = "n",
         ncol = 1, text.col = "blue", y.intersp = 2)
> par(opar) # Reset par