Author: David Zelený
Author: David Zelený
(The R code used for this website is here)
This is introduction to plotting in R. You will learn how to use function
plot, where to find meaning and setting of its graphical arguments, what is the difference between high-level and low-level graphical functions, and how to combine them together in order to make a fine changes in the figure setting.
But first, let's prepare the data to use. We will use the dataset
cars, which is R build-in dataset - you don't need to upload it from anywhere. Simply type
cars in the concole. Let's have a short review of this dataset:
class (cars) # data.frame - the function class checks the type of the object (vector, matrix, data.frame, list) names (cars) # "speed" "dist" - names of the variables (columns) in the data frame cars dim (cars) # 50 2 - it has 50 rows (samples) and 2 columns (variables)
The dataset is an old one (check
?cars to see the help file with the description), from 1920, containing information about the breaking distance (
dist, in feets) of the car going certain speed (
speed, miles per hour).
The data frame can be subsetted in several ways (square brackets, $ operator), but in the following we will use the fact that individual variables can be accessed by their names (=names of the columns).
cars$speed # returns vector with values of the first column (variable speed) cars$dist # returns vector with values of the second column (variable dist)
We want to plot the relationship of the breaking distance on the speed of the car (speed is independent variable and breaking distance is dependent - the faster the car goes, the longer is going the be the breaking distance) - the distance will be on the y-axis (vertical) and speed on the x-axis (horizontal).
More convenient instead of
plot (x, y) format is to use formula interface inside the
plot function, ie
plot (y ~ x, data) format:
The formula interface is used when there is dependent (y) and independent (x) variable, e.g. in the regression of y on x (which can be expresed by regression equation y = a + b*x, where coefficient a is the intercept and b is the regression slope). The
~ operator is called tilda (or tilde) and on the keyboard can be found in the upper left part (try it, you will often need it). The formula interface has two parts, separated by comma; first is the dependence of variables (y ~ x), and second is defining where (in which data.frame) these variables are stored (unless they are in Global environment).
To understand how this works, try the following:
... is not working, because variables
speed are not in the Global environment. But we can allow R to find it by
attaching it to the Global environment using
attach (cars) plot (dist ~ speed) # this works - both variables are searchable from Global environment
The opposite of
cars from Global environment, I can't call them directly anymore. But I can use the argument
data = to make sure that they can be found inside the
Now, let's modify the basic figure in the following way:
colours ()for the list of all named colours in R) and the filling (background) of the symbol into yellow,
plot (dist ~ speed, data = cars, xlab = 'Speed [mph]', ylab = 'Distance [ft]', # 1 pch = 21, # 2 col = 'tomato', bg = 'yellow', # 3 xlim = c(0, 25), # 4 main = 'Scatterplot') # 5
There is a long list of graphical parameters which can be modified when you are plotting your figures. You can search fo them in several places:
plot.defaultoffers detail description
parcontains even longer list of parameters (those which can be used also inside low-level graphical functions, see below).
Plotting functions in R are of two types: high-level and low-level. High-level functions do “all the job”, ie they open the graphical device (if it's not already open), prepare the axes so as all plotted data can display on them, and then plot what is required. Apart to
plot, other high-level functions are e.g.
pie. In contrast, low-level functions are only adding into already existing plot. If you want to add data as points in the scatter plot, use
points, or you can connect them by
lines, you can also add customized
axis one by one, draw the
box around the plot, and add
titles (e.g. x- and y-axis labels or the main title) or
Remember: low-level functions must be used after the first high-level function has been called, otherwise you will get the error message that the new plot was not plotted yet. The example above would look like this when using high-level and low-level graphical functions:
plot (dist ~ speed, data = cars, # a) data to be used, xlim = c(0, 25), # define range of the x-axis, axes = FALSE, ann = FALSE, type = 'n') # make sure nothing will be actually plotted points (dist ~ speed, data = cars, # b) add data points, pch = 21, col = 'tomato', bg = 'yellow') # with given symbols and colours title (xlab = 'Speed [mph]', ylab = 'Distance [ft]', # c) add x- and y-labels, main = 'Scatterplot') # add main title to the plot box () # d) add box around the plot axis (1) # e) add axis 1 (bottom horizontal) axis (2) # f) add axis 2 (left vertical)
The figure below shows how is the sequence of the plotting (each panel is labeled by the same letter as the appropriate line of the code above; the first panel is empty, because it is the result of plotting by the
plot function with arguments
type set to plot “nothing”):
The benefit of combining high- and low-level graphical functions is finer control of what can be changed in the plot. For example, let's modify the figure above by doing the following:
plot (dist ~ speed, data = cars, xlim = c(0, 25), axes = FALSE, ann = FALSE, type = 'n') points (dist ~ speed, data = cars, pch = 21, col = 'tomato', bg = 'yellow', lwd = 2) # lwd = line width, here related to data points title (xlab = list ('Speed [mph]', cex = 1.2), ylab = list ('Distance [ft]', cex = 1.2), main = 'Scatterplot', line = 2.5) # line = the distance of the titles from plot margin box (bty = 'l') # bty = box type, here 'l' (L shape) axis (1, at = seq (0, 25), tck = .02) # at = positions where tickmarks are plotted, # tck = tickmarks pointing inside axis (2, las = 2, tck = .02) # las = orientation of tick-mark labels, # tck = tickmarks pointing inside
Note that even if we asked R to plot each tickmark at the x-axis, the labels are plotted only for every second one - R makes sure the labels do not overlap, and should this happen, it will skip some. If you resize the figure (make it wider), you will see that you get labels at every tickmark.