Creating Graphics with ggplot2

Above, we used the qplot function to build ggplot2 objects in one function call. Sometimes, you may need more flexibility than qplot provides. Alternately, you may want to write a more verbose description of your plot to make your code easier to read. To do this, you create your plot in several parts:

1. You call the ggplot function to create a new ggplot object, define the input data, and define aesthetic mappings

2. You add layers to the ggplot object

Note that you add layers (and options) to a ggplot object by using the + operator.

As an example, we could create a plot identical to the one we started with using these statements:

> plt <- ggplot(data=d, mapping=aes(x=a, y=b)) + geom_point()
> summary(plt)
data: a, b, c [10x3]
mapping:  x = a, y = b
faceting: facet_null()
-----------------------------------
geom_point: na.rm = FALSE
stat_identity:
position_identity: (width = NULL, height = NULL)

To create ggplot objects without qplot, you begin by using the ggplot function.

ggplot(data, mapping = aes(), ..., environment = globalenv())

Here is a description of the arguments to ggplot2:

Argument	Description	Default
data	The default data frame for the plot
mapping	Default list of aesthetic mappings for the plot	aes()
environment	Environment in which the aesthetics should occur	globalenv()
...

The ggplot function returns a new ggplot object with no layers. You can’t actually print a chart from this object because no layers are defined:

> ggplot(data=d, mapping=aes(x=a, y=b))
Error: No layers in plot

Typically, you specify aesthetic mappings with the aes function:

aes(x, y, ...)

The x argument specifies the x value, the y argument specifies the y value, and other arguments specify aesthetics to map as name/value pairs. See the documentation for ggplot2 for alternate ways to map aesthetics including aes_string and aes_auto. As an example, to finish specifying a plot, you need to add layers. You can create a new layer with the layer function:

layer(...)

You specify the geometric objects using short names like "point". Using our earlier example, we could define our plot object with:

> plt <- ggplot(data=d, mapping=aes(x=a, y=b)) + layer("point")

The layer function allows you to specify geometric objects as name value pairs. You do not need to specify the full function name, but simply need to part after geom_.

For reference, here is a description of the available geometric functions:

Geometric Function	Description
geom_abline	A line, specified by a slope and intercept
geom_area	Area plot (a continuous analog to a bar plot)
geom_bar	Bar plot
geom_bin2d	Heatmap of two-dimensional bins
geom_blank	Blank geometric object; doesn’t draw anything
geom_boxplot	Box plot
geom_contour	Contour plot
geom_crossbar	Crossbar plot (like a box plot, but without the whiskers and extreme values)
geom_density	Density plot
geom_density2d	Two-dimensional density plot
geom_errorbar	Error bars (typically added to other plots like bar plots, point plots, and line plots)
geom_errorbarh	Horizontal error bars
geom_freqpoly	Frequency polygon (similar to a histogram)
geom_hex	Hexagonal objects (typically used with hexagonal binning)
geom_histogram	Histogram
geom_hline	A horizontal line
geom_jitter	Points, automatically jittered
geom_line	A line
geom_linerange	An interval represented by a vertical line
geom_path	A geometric path, connecting a set of points in order
geom_point	Points
geom_pointrange	A vertical line with a point in the middle (related to crossbars, boxplots, and line-ranges)
geom_polygon	A polygon
geom_quantile	A set of quantile lines from a quantile regression
geom_rect	Two-dimensional rectangles
geom_ribbon	A ribbon (a y range with continuous x values, like Tufte’s famous Napoleon’s march plot)
geom_rug	A rug
geom_segment	Line segments
geom_smooth	A smoothed condition mean
geom_step	A stepped plot connecting points
geom_text	Text
geom_tile	Tiles
geom_vline	Vertical line

ggplot2 includes some convenience functions for applying a statistical transformation and adding a layer to a plot. Some of these functions are listed below.

Statistic Function	Description
stat_abline	Adds a line with a slope and intercept.
stat_bin	Splits data into bins then plots as a histogram.
stat_bin2d	Shows density across two dimensions using rectangles.
stat_binhex	Shows density across two dimensions using hexagons.
stat_boxplot	Creates a box-and-whiskers plot.
stat_contour	Shows contours of three-dimensional data.
stat_density	Plots density.
stat_density2d	Plots density in two dimensions.
stat_function	Superimposes a function.
stat_hline	Adds a horizontal line.
stat_identity	Plots data without a statistical transformation.
stat_qq	Calculations for a quantile-quantile plot.
stat_quantile	Continuous quantiles.
stat_smooth	Adds a smoother.
stat_spoke	Plots directional data at points (specifying location with x and y, and angle separately).
stat_sum	Plots sums of unique values (typically on a scatter plot).
stat_summary	Plots summarized data.
stat_unique	Plots only unique values (removes duplicates).
stat_vline	Plots a vertical line.

You can manually specify different scales with ggplot2; mapping data to different scales lets you control how ggplot2 shows different densities, quantities, or other values. Scales can specify ranges of colors, objects, or labels. The following table shows some of these scale functions :

Scale function	Description
scale_alpha	Alpha channel values (grayscale).
scale_brewer	Colors derived from scales shown on colorbrewer.org.
scale_continuous	Continuous scales.
scale_date	Dates.
scale_datetime	Dates and times.
scale_discrete	Discrete values.
scale_gradient	Smooth gradients between two colors.
scale_gradient2	Smooth gradients among three colors.
scale_gradientn	Smooth gradients among n colors.
scale_grey	Grayscale colors.
scale_hue	Evenly spaced hues.
scale_identity	Uses values without scaling.
scale_linetype	Shows differences as line patterns.
scale_manual	Manually created discrete scales.
scale_shape	Different shapes (“glyphs”) for different values.
scale_size	Shows different values as different size objects.

With ggplot2, you can plot data using several different coordinate systems:

Coordinate function	Description
coord_cartesian	Cartesian coordinates
coord_equal	Equal scale coordinates
coord_flip	Flipped Cartesian coordinates
coord_map	Map projections
coord_polar	Polar projections
coord_trans	Transformed Cartesian coordinates

There are two options for faceting data bundled with the ggplot2 package:

Faceting function	Description
facet_grid	Lay out panels in a grid
facet_wrap	Wraps a one-dimensional list of facets into two dimensions

When you are plotting multiple geometric objects (such as multiple bars), you can specify where different objects should be plotted.

Position function	Description
position_dodge	Positions objects by dodging overlaps to the side (lays them out in a non-overlapping way)
position_fill	Stacks overlapping objects on top of one another
postition_identity	Doesn’t adjust the position
position_jitter	Jitters objects
postion_stack	Stacks objects