Legends (ggplot2)

    Solution

    Start with an example graph with the default options:

    Use , replacing fill with the desired aesthetic.

    You can also remove all the legends in a graph, using theme.

    1. # Remove legend for a particular aesthetic (fill)
    2. bp + guides(fill=FALSE)
    4. # It can also be done when specifying the scale
    5. bp + scale_fill_discrete(guide=FALSE)
    7. # This removes all legends
    8. bp + theme(legend.position="none")

    plot of chunk unnamed-chunk-4

    Changing the order of items in the legend

    This changes the order of items to trt1, ctrl, trt2:

    1. bp + scale_fill_discrete(breaks=c("trt1","ctrl","trt2"))

    Depending on how the colors are specified, you may have to use a different scale, such as scale_fill_manual, scale_colour_hue, scale_colour_manual, scale_shape_discrete, scale_linetype_discrete, and so on.

    Reversing the order of items in the legend

    To reverse the legend order:

    1. # These two methods are equivalent:
    2. bp + guides(fill = guide_legend(reverse=TRUE))
    3. bp + scale_fill_discrete(guide = guide_legend(reverse=TRUE))
    5. # You can also modify the scale directly:

    plot of chunk unnamed-chunk-7

    Instead of scale_fill_discrete, you may need to use a different scale, such as scale_fill_manual, scale_colour_hue, scale_colour_manual, scale_shape_discrete, scale_linetype_discrete, and so on.

    This will hide the legend title:

    Modifying the text of legend titles and labels

    There are two ways of changing the legend title and labels. The first way is to tell the scale to use have a different title and labels. The second way is to change data frame so that the factor has the desired form.

    Using scales

    The legend can be a guide for fill, colour, linetype, shape, or other aesthetics.

    With fill and color

    Because group, the variable in the legend, is mapped to the color fill, it is necessary to use scale_fill_xxx, where is a method of mapping each factor level of group to different colors. The default is to use a different hue on the color wheel for each factor level, but it is also possible to manually specify the colors for each level.

    1. bp + scale_fill_discrete(name="Experimental\nCondition")
    3. bp + scale_fill_discrete(name="Experimental\nCondition",
    4.                          breaks=c("ctrl", "trt1", "trt2"),
    5.                          labels=c("Control", "Treatment 1", "Treatment 2"))
    7. # Using a manual scale instead of hue
    8. bp + scale_fill_manual(values=c("#999999", "#E69F00", "#56B4E9"), 
    9.                        name="Experimental\nCondition",
    10.                        breaks=c("ctrl", "trt1", "trt2"),
    11.                        labels=c("Control", "Treatment 1", "Treatment 2"))

    plot of chunk unnamed-chunk-10

    Note that this didn’t change the x axis labels. See Axes (ggplot2)) for information on how to modify the axis labels.

    If you use a line graph, you will probably need to use scalecolour_xxx and/or scale_shape_xxx instead of scale_fill_xxx. **_colour maps to the colors of lines and points, while fill maps to the color of area fills. shape** maps to the shapes of points.

    We’ll use a different data set for the line graphs here because the PlantGrowth data set does not work well with a line graph.

    1. # A different data set
    2. df1 <- data.frame(
    3.     sex = factor(c("Female","Female","Male","Male")),
    4.     time = factor(c("Lunch","Dinner","Lunch","Dinner"), levels=c("Lunch","Dinner")),
    5.     total_bill = c(13.53, 16.81, 16.24, 17.42)
    6. )
    8. # A basic graph
    9. lp <- ggplot(data=df1, aes(x=time, y=total_bill, group=sex, shape=sex)) + geom_line() + geom_point()
    10. lp
    12. # Change the legend
    13. lp + scale_shape_discrete(name  ="Payer",
    14.                           breaks=c("Female", "Male"),
    15.                           labels=c("Woman", "Man"))

    plot of chunk unnamed-chunk-11

    If you use both colour and shape, they both need to be given scale specifications. Otherwise there will be two two separate legends.

    1. # Specify colour and shape
    2. lp1 <- ggplot(data=df1, aes(x=time, y=total_bill, group=sex, shape=sex, colour=sex)) + geom_line() + geom_point()
    3. lp1
    5. lp1 + scale_colour_discrete(name  ="Payer",
    6.                             breaks=c("Female", "Male"),
    7.                             labels=c("Woman", "Man"))
    9. # Specify both colour and shape
    10. lp1 + scale_colour_discrete(name  ="Payer",
    11.                             breaks=c("Female", "Male"),
    12.                             labels=c("Woman", "Man")) +
    13.       scale_shape_discrete(name  ="Payer",
    14.                            breaks=c("Female", "Male"),
    15.                            labels=c("Woman", "Man"))

    plot of chunk unnamed-chunk-12plot of chunk unnamed-chunk-12

    Kinds of scales

    There are many kinds of scales. They take the form scale_xxx_yyy. Here are some commonly-used values of xxx and yyy:

    yyy Description
    hue Equally-spaced colors from the color wheel
    manual Manually-specified values (e.g., colors, point shapes, line types)
    gradient Color gradient
    grey Shades of grey
    discrete Discrete values (e.g., colors, point shapes, line types, point sizes)
    continuous Continuous values (e.g., alpha, colors, point sizes)

    Changing the factor in the data frame

    Another way to change the legend title and labels is to directly modify the data frame.

    The legend title “Experimental Condtion” is long and it might look better if it were broken into two lines, but this doesn’t work very well with this method, since you would have to put a newline character in the name of the column. The other method, with scales, is generally a better way to do this.

    Also note the use of backticks instead of quotes. They are necessary because of the spaces in the variable name.

    Modifying the appearance of the legend title and labels

    1. # Title appearance
    2. bp + theme(legend.title = element_text(colour="blue", size=16, face="bold"))
    4. # Label appearance
    5. bp + theme(legend.text = element_text(colour="blue", size = 16, face = "bold"))

    plot of chunk unnamed-chunk-14

    By default, the legend will not have a box around it. To add a box and modify its properties:

    1. bp + theme(legend.background = element_rect())
    2. bp + theme(legend.background = element_rect(fill="gray90", size=.5, linetype="dotted"))

    plot of chunk unnamed-chunk-15

    Changing the position of the legend

    Position legend outside the plotting area (left/right/top/bottom):

    1. bp + theme(legend.position="top")

    It is also possible to position the legend inside the plotting area. Note that the numeric position below is relative to the entire area, including titles and labels, not just the plotting area.

    plot of chunk unnamed-chunk-17plot of chunk unnamed-chunk-17

    Hiding slashes in the legend

    If you make bar graphs with an outline (by setting colour=”black”), it will draw a slash through the colors in the legend. There is not a built-in way to remove the slashes, but it is possible to cover them up.

    1. # No outline
    2. ggplot(data=PlantGrowth, aes(x=group, fill=group)) +
    3.     geom_bar()
    5. # Add outline, but slashes appear in legend
    6. ggplot(data=PlantGrowth, aes(x=group, fill=group)) +
    7.     geom_bar(colour="black")
    9. # A hack to hide the slashes: first graph the bars with no outline and add the legend,
    10. # then graph the bars again with outline, but with a blank legend.
    11. ggplot(data=PlantGrowth, aes(x=group, fill=group)) +

    plot of chunk unnamed-chunk-18

    For further information, see:https://github.com/hadley/ggplot2/wiki/Legend-Attributes