Getting started with R and RStudio

Rhondda Jones & Robin Gilliver

© 2017

Copyright vests in the authors

About this content

This content has been provided to be used in the Statistical Methods for Data Scientists subject.
Apart from being reformatted to an online delivery and some minor modifications, the content has
been created by the authors Rhondda Jones and Robin Gilliver and they retain their copyright.
However, the authors are not responsible for the delivery of the content in the Quantitative Methods
for Scientists subject.

The authors can be contacted about this material via the email: [email protected].
Preface
The first version of R was written largely for teaching purposes by two New Zealand statisticians,
Ross Ihaka and Robert Gentleman, in 1995: it evolved from the S language developed by John
Chambers at the AT&T Bell laboratories. It was an instant hit! By now, R has become the fastest
growing analytical software available, in terms of both its usage and its capacity. The ability to use it
effectively is now an essential and marketable skill for most quantitatively-oriented professions, from
bioinformatics to data science. Because it is free and open-source, R has also become the
environment where new analytical methods first appear. It has a huge and growing online user
community, and numerous websites devoted to assisting and coaching new users. Welcome to the R
community! We hope you enjoy the journey.

1 Installation and startup

R is a specialist computer language designed mainly for data analysis – although, like any other
computer language, it can do many other things too. It is best used within an integrated
development environment, or IDE, which makes writing R code much easier. Here we introduce you
both to R itself and to RStudio, the best-known IDE for R. This manual is intended to be used
actively – that is, you should try each of the procedures and commands described here for yourself,
making sure that you can generate the right outputs. For each coding section, it is also good practice
to invent and test a few similar additional commands to make sure that you really do understand the
syntax which is described. In our experience, it takes about 12 hours of active use of R and RStudio
to become reasonably fluent with them.

1.1 Installing R and RStudio

R is free and can be downloaded from the CRAN website:

https://cran.r-project.org/

This page provides links to let you download installers for the latest R versions for Windows, Mac,
and Linux machines.

RStudio can be downloaded from:

https://www.rstudio.com/products/rstudio/download3/

This website has links to download either personal (free) or commercial versions of RStudio. Most
users will use the desktop personal license. As with R, you can choose an installer appropriate to
your computer.
1
You should install R before installing RStudio. In each case, double-click on the downloaded installer
and follow the installation instructions provided. At the end of this process, you should have icons
for both R and RStudio on your desktop.

1.2 Starting R and RStudio

Once you have R installed, its desktop icon looks as shown on the right.

For RStudio, the desktop icon will be:

You can double-click on the icon in the usual way to start either program. Starting RStudio will also
start R automatically – but initially, start R on its own to see how to use it without RStudio.

1.2.1 Using R with the RGui

If you open R rather than RStudio, the program window (RGui) will look as shown below.

2
Figure 1-1: The RGui window containing the R console window.

The toolbar at the top includes a number of useful functions whose

purpose is listed to the left.

We will not dwell on them here because their functions are superseded
by the capabilities of RStudio.

The inner Console window executes R commands that you type in

manually. When you first open R, it will provide various items of
information, and then display a prompt ( > ) to type in a command.
After entering a command, you then press the Enter key on your
keyboard to request that the command be executed. When R finishes executing a command, it will
display a new prompt. If you press the Enter key after typing an incomplete command, R will display
a plus sign ( + ) to indicate that it expects additional input for the
command. Pressing the Esc key on your keyboard will cancel the Arithmetic operators :
current command and give you a > ready for new input.

To see how this works, enter a few simple commands. At the

prompt ( > ) type in any arithmetic expression, for example: + add

> (3 + 9) * 5 - subtract

Press the Enter key. R responds with the answer – in this case

[1] 60

Notice that the command appears in red and the response in blue. Depending on the command,
there can be multiple values in the response: the number at the beginning of the output line is the
index number for the first value in the line.

Now enter the same calculation again, but with an error – omit the closing bracket after the 9.

> (3 + 9 * 5

When you press the Enter key, R responds with

This time R notices the absence of a closing bracket and indicates that it expects further input. We
can give it a closing bracket and press Enter

+)
3
[1] 48

but the answer is wrong because the final bracket should not be at the end. Commands can be
retrieved using the up and down arrows on your keyboard. In this case, press the up arrow twice to
retrieve the command with the mistake, and then correct it. Then try the following, and more of
your own …

> 12/8

[1] 1.5

> 12%%8

[1] 4

> 12%/%8

[1] 1 Order of operations

> 12^2 Normal algebraic rules
[1] 144 apply— i.e. according to the
precedence list below and left
> (50 + 6) / 2^3 to right for operators with
similar
[1] 7
precedence

Inside parentheses ()
You can type more than one command on a line if you separate the commands with a semicolon.

> 12*3 ; 12-4

[1] 36

[1] 8

You can also evaluate logical expressions Logical (comparison)

For example operators:

10^2 == 99

[1] FALSE == is equal to

This expression is read as “Ten squared is equal to 99” and R correctly tells you that this statement is false.
Logical expressions can be made quite complex using “and” (&& for single values) and “or” (|| for single values).
So the first expression below is read as “10 is equal to 8 OR 5 squared is equal to 25” and evaluates as true.
!= is not equal to

> is greater than

4
< is less than

>= is greater than or

equal to
(10==8)||(5^2==25)

[1] TRUE

(10==8)&&(5^2==25) R names

[1] FALSE Names are case-sensitive, can

Try more examples of your own. be any length, and can include
letters, numbers, and the
period symbol. They must
To save the result of any expression, assign the expression to a name, start with a letter and must
for example, enter: not contain spaces.

> x = (50 + 6) / 2^3 So

> joanne1

This time, instead of displaying the result, R has stored it in a variable Joanne1
called x. To display it, use:

>x

[1] 7

R accepts three assignment operators:

<- (a ‘less-than’ symbol followed immediately by a hyphen)

-> (a hyphen followed immediately by a ‘greater-than’ symbol)

= an equals sign.

Thus the three commands

Lizzie <- "My cat"

"My cat" -> Lizzie
Lizzie = "My cat"

are all equivalent—they all say “Assign the character string "My cat" to the name Lizzie”. In this
manual we will use the last of these options. The assignment command creates an R object called
Lizzie which has the value "My cat" and stores it in your workspace. Assignment to a previously used
name will destroy the old object and create a new one.

5
An assignment command will not automatically display the value of the new object. To display the
value, either enter the name separately, or place the assignment in brackets (which will assign and
display in the one operation ... ).

height = 172

height

[1] 172

(height = 172)

[1] 172

To clear the Console window, right click on it and select Clear All.

At this point we will move on to using RStudio as an environment for R, so close the RGui window.

6
1.2.2 Using R with RStudio
Start RStudio by double-clicking its startup icon

When you start RStudio for the first time, it should resemble the screenshot below, with a large
window to the left, and two smaller windows to the right.

Figure 1-2: The RStudio windows before creating a script file.

The left-hand window is the R console for whichever version

of R you have previously installed. Notice that it is helpfully The layout described here is for
labelled Console at the top, and displays the directory RStudio version 1.0.136. Earlier
where your R workspace is located. The console window versions label some windows
displays commands and the results of their execution, just differently .
as the ordinary R console does.

The top right-hand window has two tabs – Environment and History. The Environment tab gives you
the contents of your current R workspace – which will be empty at this point. The History tab keeps
track of all the commands you have used – and again, is empty when you first open RStudio.

7
The bottom right-hand window has five tabs. The first one, Files, gives you the files within the folder
where your workspace is located. The second, Plots, will display plots you create in your R session.
The third, Packages, gives you a list of the packages you currently have access to, and the fourth,
Help, lets you search the available help files. The fifth (Viewer) can display local web content.

You will find the right-hand windows very helpful. But the most useful part of RStudio is not visible
yet: to obtain it, we need a new window which appears when you create a script, and gives you
access to the source editor of RStudio.

1.3 Creating, editing and running scripts

The key advantage of using an IDE like RStudio is the ability to edit and test both individual
commands and sets of commands within a specialized script editor.

To create a script file to hold your commands, press

File > New File > R Script

A new window is created above the Console window, so that RStudio now looks as shown below:

8
Figure 1-3: The RStudio windows including a script window.

Notice that the script window tab is labelled “Untitled1”, This is currently the name of the script. To
change the name to something more descriptive and save it, press

File > Save as…

enter Introduction for the File name: and press Save . Notice that it is saved as
Introduction.R – RStudio expects any file containing an R script to end in .R.

To illustrate how the script window works, we will enter, edit, and execute a few simple commands.

In the script window, type

x = (50 + 6) / 2^3

As you type the line, notice that RStudio helpfully matches brackets as you go. It will do the same for
quote marks.
9
With the cursor still on the line, press the Run button at the top of the script window.

Notice that the command is echoed in the Console window and has been successfully executed.

Also, in the Environment window on the top right, x is now listed under Values, together with its
value (7).

Now select 2^3 within the typed line, and press Run again. The selected expression is again echoed,
and this time (because the x = part of the command was not selected), the expression is evaluated
and the result is shown in the Console window. Both commands are echoed under the History tab.

The general rule is that when you press Run, if nothing is selected, RStudio runs the line in the script
where the cursor is located. If the line is an incomplete command, it will keep reading lines until the
command is completed. If something is selected – part of a line, or several lines – it runs whatever is
selected.

Now close RStudio by clicking the red Close button on the top right. RStudio will ask you if you want
to save the data and script you have created in the session.

10
Figure 1-4 Quit dialogue for RStudio

If you press the Save Selected button, both the script and any saved data will appear the next time
you open RStudio – press it and reopen RStudio to check.

11
2 Introduction to R functions
A function is a set of R operations which are given a name and executed as a group for a particular
purpose. Most statistical operations can be executed with a single
command which calls a ready-made function. Parentheses must always be
included when you use a
Functions have the form : function in a command
somename() A function may have no
usually with one or more values inside the brackets. The quantities arguments if:
inside the brackets are called the arguments of the function— they
are the data on which the function operates. For example, the function • it has a value which is not
to calculate square roots is affected by user input—

sqrt(x) e.g. , which gives the

current date and time, or
where x is either a number or the name of a variable which contains
one or more numbers. • there is a default
argument which the
sqrt(25)
function can use if an
[1] 5 argument is not

supplied - e.g.

Any arithmetic expression you type in for R to evaluate can include any of the R functions with
appropriate arguments in parentheses ( ). If the expression cannot be evaluated, R will respond with
a missing value (NaN - for "not a number") and a warning message. Division by zero gives Inf (for
"infinity").

Type the expressions below on the left into the RStudio script window and run each one. The
Console window should then appear as shown below on the right.

5 + sqrt(16)

round(25.1872,
1)

log (-1)

12
exp(2) / log(1)

For the rest of this manual, to demonstrate a command we will generally provide the command as
entered in the script window in black, followed by the output produced on the console in blue – eg

5 + sqrt(16)

[1] 9

Where a command is too long to fit on one line of this manual, we will indent the second and
subsequent lines slightly.

Each of the arguments used by a function has a name, and for some of them, R will provide a default
value. For example, the full definition of the round() function - found by running the command
?round is

round(x, digits = 0)

x has no default value, because that is data that you must provide - the number(s) to be rounded.
But digits has a default value of 0: this means that if you do not specify a value, R will round x to
the nearest whole number, with no digits after the decimal place. If you provide a value - as we did
in the earlier example, it replaces the default.

You can specify the argument name within the command: if you provide argument names,
arguments can be in any order. If you do not provide the name, then arguments must be in the same
order as in the function definition.

So in the round() example we could have entered:

round(digits=1, x=25.1872)

[1] 25.2

There is an enormous number of R functions – to create and modify data, or to analyze it, search for
it, model it, or plot it in a variety of ways. If you can’t find a function to do exactly what you want, it
is possible to write your own. Most people have a set of functions that they use frequently and
remember, and use Google to search for those they don’t use very often.

2.1 Functions for creating and labelling vectors

13
So far we have applied functions to single values, as in the round() example above. But most R
functions are designed to operate on groups of values, which R calls vectors. A vector is a set of
values of the same data type (eg all numeric, or all character) which has a single name for the whole
set. This means that you can do calculations which apply to all elements of the vector with a single
command.

2.1.1 Creating vectors with the c() function.

This simply combines a set of values into a vector, and is the most common way to create vectors
manually.

myData = c(25, 30, 12, 15, 8, 9, 32)

myData

[1] 25 30 12 15 8 9 32

weights = c(125, 147)

weights

[1] 125 147

You can also use it to add values to an existing vector

myData = c(myData,16, 22)

myData

[1] 25 30 12 15 8 9 32 16 22

The c() function can also create a vector of character strings.

myNames = c("Mary", "Fred")

myNames

[1] "Mary" "Fred"

2.1.2 Naming vector elements

14
Individual elements of a vector can be given names with the names()function. So to attach
“Mary” and “Fred” to the values in the weights vector we created earlier:

names(weights) = myNames

weights

Mary Fred

125 147

You can also give names to each value when a vector is created using the c()function.

roles = c(Jane="boss", Bill="foreman", Tom="analyst")

roles

Jane Bill Tom

"boss" "foreman" "analyst"

2.1.3 Accessing, displaying and changing vector elements

Each element of a vector has an index nunber, starting with 1. Use square brackets and the index
number to display, change, or refer to an individual element of the vector.

myData

[1] 25 30 12 15 8 9 32 16 22

myData[3]

[1] 12

myData[3] = 16

myData

[1] 25 30 16 15 8 9 32 16 22

You can select more than one element using a vector of index numbers inside the square brackets:

myData[c(1,3,5,7,9)]

[1] 25 16 8 32 22

Another way to select elements is by using a logical expression:

15
myData[myData >20]

[1] 25 30 32 22

2.1.4 Creating vectors with sequential or repeating data: seq() and

rep()
Vectors in which the values are a sequence or a repeating pattern are very important in setting up
analyses. Three useful functions for creating these kinds of vectors are the sequence operator ‘:’, and
the seq() and rep() functions. These are illustrated below.

my.series = 1:10

my.series

[1] 1 2 3 4 5 6 7 8 9 10

my.second.series = 4:-4

my.second.series

[1] 4 3 2 1 0 -1 -2 -3 -4

my.third.series = seq(10, -10, by = -2)

my.third.series

[1] 10 8 6 4 2 0 -2 -4 -6 -8 -10

my.fourth.series <- rep(c("A", "B"), times=4)

my.fourth.series

[1] "A" "B" "A" "B" "A" "B" "A" "B"

my.fifth.series <- rep(c("A", "B"), each = 4)

my.fifth.series

[1] "A" "A" "A" "A" "B" "B" "B" "B"

2.2 Some basic calculations and plots using vectors

Vectors are the most important data structure which R provides, and using them expands the
capacity of R immensely.
16
2.2.1 Create a set of four vectors to illustrate vector calculations
To illustrate a range of simple calculations and plots, create the five vectors shown below:

• A vector called bodyWeights containing 500 individuals from a hypothetical population of

adult males whose weights are Normally distributed with a mean of 80 and a standard
deviation of 10.

bodyWeights = rnorm(n=500, mean=80, sd=10)

• A labelled vector containing the numbers of described species in different vertebrate groups

vertebrateSpecies = c(Fish=32900, Amphibia=7302,

Reptiles=10038, Birds=10425, Mammals=5513)

• A vector called Year which holds a sequence of 10 years from 2006 to 2015

Year = 2006:2015

• A vector called Density which holds the 10 values: 5, 3, 7, 8, 6, 9, 8, 10, 14, 11, representing
hypothetical densities of seedlings in each of the years listed in the Year vector.

Density = c(5, 3, 7, 8, 6, 9, 8, 10, 14, 11)

Notice that these four vectors and some or all of the values they hold are now listed under Values in
RStudio’s Environment window.

2.2.2 Missing values

The standard way to indicate that a value is missing is with an NA (for “not available”). For example,
suppose that we had a second set of densities for the years 2006 to 2015, but were missing data for
the years 2008 and 2010 (that is, for the third and fifth values in the vector).

Density2 = c(6, 6, NA, 7, NA, 8, 10, 11, 10, 15)

2.2.3 Basic vector calculations, pie charts, and bar plots

17
Once a vector has been created, functions or arithmetic operations applied to a vector act on all
elements of the vector.

For example, to express the numbers of vertebrate species in thousands

vertebrateSpecies / 1000

Fish Amphibia Reptiles Birds Mammals

32.900 7.302 10.038 10.425 5.513

To calculate the total number of described vertebrate species

sum(vertebrateSpecies)

[1] 66178

And then, to estimate the percentage of species in each group, rounded to 1 decimal place:

percentageSpecies = round(100 * vertebrateSpecies/

sum(vertebrateSpecies), 1)

percentageSpecies

Fish Amphibia Reptiles Birds Mammals

49.7 11.0 15.2 15.8 8.3

With a labelled vector, it is also straightforward to draw pie plots or bar plots to display the data.
When you run each plot command, the plot should appear in the Plots tab on the bottom right
window of RStudio .

pie(vertebrateSpecies)

Fish

Amphibia Mammals

Reptiles Birds

18
For the bar plot, you will probably need to make the plot window wider to allow all the bar labels to
be displayed – this can be done by dragging the corner of the window after the plot has been drawn.

barplot(percentageSpecies)
40
30
20
10
0

Fish Amphibia Reptiles Birds Mammals

This plot could do with a bit of improvement – below we modify the command extensively to
produce the final product. You should try adding each new argument one at a time to see what each
does. Note the use of \n in the main argument to force part of the title to be on a new line.

barplot(percentageSpecies, las=1, col="orange",

ylab="% of total vertebrate species",

main="Distribution of species among\n vertebrate groups",

ylim=c(0,55))

box()

Distribution of species among

vertebrate groups
% of total vertebrate species

50
Colors
40

30 To have a look at the range of

20 colors you can use, run the
10
command
0
Fish Amphibia Reptiles Birds Mammals

19
There are several useful buttons above the RStudio plot window. Try them out. The Zoom button
gives you the plot in a separate window whose shape and size it is easier to change. The Export
button allows you to either save the plot or copy it to the Clipboard to transfer to another document
– eg a Word document or a PowerPoint slide. If you have drawn several plots, the arrows allow you
to move between them. The remaining two buttons delete either the current plot or all the plots you
have drawn.

2.2.4 Calculating descriptive statistics and plotting distributions

To demonstrate the calculation of descriptive statistics we use the bodyWeights vector, where the
mean and standard deviation of the population from which the sample was drawn are specified as 80
and 10 respectively. Because you will have created a different random sample, the results you get
when you run these commands will be slightly different from those shown here.

mean(bodyWeights) # calculates the mean

Comments
[1] 79.8727
The symbol # signals a
median(bodyWeights) comment to R. Everything
[1] 80.28216
between the # and the end of a
line is ignored, e.g. in the
sd(bodyWeights) # the standard deviation command

[1] 10.42242

range(bodyWeights) what is executed is just

[1] 47.45433 114.05530

var(bodyWeights) # the variance

[1] 108.6269

length(bodyWeights) # the number of elements

[1] 500

summary(bodyWeights) # gives min, max, median, mean, quartiles

Min. 1st Qu. Median Mean 3rd Qu. Max.

47.45 72.83 80.28 79.87 86.66 114.10

There are two standard ways to plot a distribution of numbers like the bodyWeights data: a
boxplot and a histogram.

20
boxplot(bodyWeights)

Using the default aguments for the boxplot function:

• The heavy horizontal line represents the median
value
• The box includes the inter-quartile range (that is, it
includes 50% of the data).
• The whiskers extend a further 1.5 times the
interquartile range
• Any remaining values are shown as individual
points (to identify them as possible outliers)
To examine arguments which modify these conventions
(e.g. the range argument which defines the extent of the
whiskers) use the command
?boxplot

hist(bodyWeights)

Histogram of bodyWeights
100
80
Frequency

60
40
20
0

50 60 70 80 90 100 110

bodyWeights

Now edit the hist(bodyWeights) command in the script window to change some of the
attributes of the plot, and run the edited command.

hist(bodyWeights, las=1, col="lightblue",xlab="Body weight (kg)",

21
 ylab="Number of individuals",

main="Distribution of male body weights",

labels=T, ylim=c(0,120))

Then put a box around the plot to obtain the final result below.

box()

Distribution of male body weights

120
108

100
85
Number of individuals

80
70
66
60
46
41
40
30 30

20
7 8
2 4 2 1
0

50 60 70 80 90 100 110

Body weight (kg)

2.2.5 Descriptive statistics when there are missing values

When a vector contains missing values, commands to calculate summary statistics like means and
standard deviations will generally need to be modified. For example, to calculate the mean of the
Density2 vector, which has two missing values, we might use:

mean(Density2)

[1] NA

R returns a missing value – because two values are missing, the mean of all 10 values in Density2
cannot be calculated. If you want the mean of the remaining values, the command needs to specify
that missing values should be removed before the calculation is made.

mean(Density2, na.rm=TRUE)
22
[1] 9.125

The function is.na() returns TRUE for values which are missing and FALSE otherwise:

is.na(Density2)
Key arguments in plot()
[1] FALSE FALSE TRUE FALSE TRUE FALSE FALSE FALSE FALSE
FALSE The initial part of this plot
command:
To find the indices of missing values:
Density~Year
which(is.na(Density2))
is a formula which can be read as
[1] 3 5 “Density (on the y-axis) is a
function of Year (on the x-axis)”.
To find the indices of values which are NOT missing:
The type argument specifies
which(!is.na(Density2))
what kind of plot – “b” means
[1] 1 2 4 6 7 8 9 10 “both points and lines”. “l” would
plot only lines, and “p” would plot
only points. There are other
options too = see ?plot

2.2.6 Calculations and plots with two pch specifies the plotting
numeric vectors character – in this case a solid
circle – there are a lot of options,
Here we use the two associated vectors Year and Density,
see ?points
which each has 10 observations. Initially draw a line plot to show how the the density changes with
year.

plot(Density~Year, type="b", pch=16,

col="red", cex=1.5, las=1)

23
 14

12

10
Density

2006 2008 2010 2012 2014

Year

The plot suggests that density increases with year – that is, there is a positive correlation between
Year and Density. To calculate the correlation coefficient (a measure of the strength of the
correlation) use

cor(Year, Density)

[1] 0.8699182

Since the maximum possible value of the correlation coefficient is +1, the correlation appears quite
strong.

To calculate the intercept and slope of the straight line which best describes the relationship
between Density and Year:

lm(Density~Year)

Call:

lm(formula = Density ~ Year)

Coefficients:

(Intercept) Year

-1807.442 0.903

24
 Key arguments in abline()
To draw the regression line superimposed on an existing scatterplot
(ie a plot which uses points instead of both points and lines): lwd changes the thickness of
the line.

lty specifies what sort of

plot(Density~Year, type="p", pch=16,
line (in this case a dashed line)
col="red", cex=1.5, las=1)

abline(-1807.442, 0.903, lwd=2, lty=2, Key arguments in text()

col="red") x and y specify the

coordinates of the middle of
the string relative to the x & y
We can then create a character string to write the equation, and use axes.
the text() command to superimpose that on the plot too, giving labels specifies what is to
the final result below.

text(x=2008, y=12,

labels="Density = -1807.442 + 0.903 Year")

25
26
3 Data in R
3.1 Data types
Data in R come in a variety of different types, some of them (like dates) quite complex. The data
types you will probably use most often are:

• Integer
• Numeric (Double)
• Character
• Factor
• Logical

The properties of each are outlined below: – we have used the first three in earlier sections.

Integer: integer data are whole numbers – eg 12, -103, 0 The Year vector created earlier
was an integer variable.

Numeric (Double): numeric data with a decimal part - eg 12.15, -103.0, 0.007 “Double” is
short for “double precision”, which means that each value is accurate up to 16 places – so 0.007 is
actually stored with another 13 zeroes after the 7. The bodyWeights variable created earlier
stored its data as doubles.

Character: character data are character strings, used most often to label or name individual
values or data rows. When you surround a value with quotation marks, it is created as character
data even if what is in quotes is a number, for example

textStrings = c("alpha", "Catch the ball!", "27.6")

textStrings

[1] "alpha" "Catch the ball!" "27.6"

Factor: when a variable is a category which is used to group data into subsets, it should be stored
as a factor variable. Factor data has two parts – an integer which defines the level of the factor, and
a label which is a text string used to describe each level. For example, consider the bodyWeights
vector holding 500 hypothetical male body weights we created in chapter 2 via the command

bodyWeights = rnorm(n=500, mean=80, sd=10)

Add 500 hypothetical female body weights with a mean of 65 and a standard deviation of 10 to the
original bodyWeights vector via the command below:

27
bodyWeights = c(bodyWeights, rnorm(n=500, mean=65, sd=10))

Notice that in the Environment window, the bodyWeights vector now has 1000 values: the first
500 are the original 500 males, and the second 500 are the females. We now create a vector called
gender which can be used to specify which individuals are male and which are female:

gender = rep(c("Male", "Female"), each = 500)

Check the vector in the Environment window: notice that it has been created as a character
variable. It can be converted to a factor with the command:

gender = factor(gender)

Notice that the Environment window now tells us that gender is a Factor with 2 levels, and that
“Female” is listed before “Male”. This means that Female is level 1 for the factor , and Male is
level 2 (note that the early values of gender are all 2, as they should be). In general, unless you
specify otherwise, the factor levels will be defined in alphabetical order: in this case “female”is
defined as level 1 because F comes before M in the alphabet. The order can be changed: the
command below creates a new gender1 vector with “Male” as the first level.

gender1 = factor(gender, levels =c("Male", "Female"))

The order of levels determines how values are plotted and analyzed, as illustrated in the two plots
generated below.

plot(bodyWeights ~ gender, col="lightgreen", las=1)

plot(bodyWeights ~ gender1, col="red", las=1)

28
Logical: logical variables can have only two possible values – TRUE and FALSE. Like factor
variables, there is also a number associated with them (FALSE is 0 and TRUE is 1). Logical variables
are often created by comparisons of some kind.

For example, consider the relationship between Density and Year which we evaluated at the end
of Chapter 2, producing the regression equation and plot shown again below.

Year = 2006:2015

Density = c(5, 3, 7, 8, 6, 9, 8, 10, 14, 11)

lm(Density~Year)

plot(Density~Year, type="p", pch=16, col="red", cex=1.5, las=1)

abline(-1807.442, 0.903, lwd=2, lty=2, col="red")

text(x=2008, y=12,

labels="Density = -1807.442 + 0.903 Year")

We can use the equation to obtain the fitted values for each year:

fitted = -1807.442 + 0.903*Year

The fitted points should of course sit exactly on the regression line, as we can demonstrate by
plotting them on the same graph:

points(fitted ~ Year, cex=1.5, col="black")

So which years had densities higher than expected (ie above the fitted line)? We can identify them
by creating a logical variable which compares Density with fitted:

isHigh = Density > fitted

isHigh
29
[1] TRUE FALSE TRUE TRUE FALSE TRUE FALSE FALSE TRUE FALSE

If isHigh = TRUE, the relevant year had a density above the line. We can show those years
using the which() function to find the indices of the TRUE values:

which(isHigh)

[1] 1 3 4 6 9

Year[which(isHigh)]

[1] 2006 2008 2009 2011 2014

3.2 Data structures: data frames

A data structure is something which can hold a set of data – so vectors are one type of data
structure. However, an experiment or survey normally involves several variables which are
associated in some way, so we want to keep them together in something like a spreadsheet – that is,
a two-dimensional structure where each variable is a column in the spreadsheet, and each “case” is a
row. The R data structure to do this is called a data frame.

3.2.1 Data frames

A data frame combines a set of vectors which are all the same length. Its vectors can be of different
data types, so a single data frame might include vectors of categorical data (the factor data type), of
logical data, and of numerical data. While each vector can only contain data of one type, the data
frame as a whole can include data of several different types.

Usually, you will start with the data in a spreadsheet of some kind, and then read the whole data set
into R – we will demonstrate how to do that shortly.

However, data frames can also be created manually by creating the individual vectors and then
copying them into a data frame with the data.frame() function, So, for example, we might want
to copy the Year, Density, fitted and isHigh vectors created in Section 3.1 into a data
frame called seedlings:

seedlings = data.frame(Year, Density, fitted, isHigh)

Similarly we can combine the bodyWeights and gender vectors into a data frame called
weights:

weights = data.frame(gender, bodyWeights)

30
In the RStudio Environment window, notice that there are still “loose” Year, Density
bodyWeights and gender vectors under the Values heading, but there is now also a Data
heading which lists the seedlings and weights data frames: click the little blue arrows to the
left of seedlings and weights to display their vectors. Then click the grid symbol to the right of
seedlings to show the whole data frame in a new tab to the left, as in the screenshot below.

You can refer to (and change) any value in the data frame by specifying its row index and column
index in square brackets using dataframeName[row index, column index]. So to display
the value in the first row and first column:

seedlings[1, 1]

[1] 2006

Similarly you can obtain blocks of values by specifying vectors of row and column numbers: for
example

seedlings[1:3, c(1,3)]

Year fitted

31
1 2006 3.976

2 2007 4.879

3 2008 5.782

Leaving out the row (or the column) number requests a whole column (or row) – notice that the
comma still has to be provided.

seedlings[, 1]

[1] 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015

seedlings[1, ]

Year Density fitted isHigh

1 2006 5 3.976 TRUE

Once vectors have been copied into a data frame, it is good practice to go back to the script tab and
remove the originals from the workspace to avoid confusion:

rm(Year, Density, fitted, isHigh)

Notice that the vectors have disappeared from the Values part of the environment window, but they
are still in the seedlings data frame.

To use vectors stored within a data frame, the name of the data frame must be specified within the
command. There are four common ways to do this:

1. Specify the name of the data frame whenever you refer to one of its vectors, separating the
two names with a $ sign.

plot(seedlings$Density ~ seedlings$Year, pch=16)

2. Specify the vectors via their column numbers (leaving out the row numbers requests the
whole column):

plot(seedlings[ ,2] ~ seedlings[ ,1], pch=16)

32
 3. Use the with() function. This instructs R to look inside the specified data frame for the
vector names. This is often the most economical approach.

with(seedlings, plot(Density ~Year, pch=16) )

4. Commands which include a formula will usually allow the data frame to be specified in the
command itself.

plot(Density ~ Year, pch=16, data=seedlings)

3.2.2 Importing an existing data set into a data frame

A sample data set is shown in the green box to the right,
variety treatment height.cm biomass.gm
which contains the results of a hormone treatment on the
height and biomass of two varieties of a plant species. A None 14 73
There are eighteen rows, each representing an individual
plant. The columns specify two categorical vectors A None 17 65
(variety and treatment) and two numeric vectors
A None 20 84
(height.cm and biomass.gm)
A Low 24 119
All columns in the data set must be the same length, and
data in a single column must all be of the same type (eg A Low 25 105
either all numeric, or all categorical).
A Low 21 121
This data set is provided as a csv file (that is, as comma-
separated variables in a text file) called A High 36 190
plantGrowthExpt.csv – notice that this file
A High 28 123
contains only the variable names and the data itself – no
other headings, and no empty rows. Files can be saved as A High 38 173
csv files from Excel and from most other statistical
B None 16 78
packages. To import it into R, use the command:
B None 22 115
plantGrowthExpt = read.csv(file.choose())

When you run the command, you will get a Select file window which B letsNone 24to the file,180
you navigate
select it, and press Open to load it into R. Then inspect the fileBin the Environment
Low window.
27 By 158
clicking the little grid button to the right of the file name, display it in a script window tab, as shown
below.You can also obtain this display with the command B Low 34 196

view(plantGrowthExpt) B Low 21 161

B High 32 213

33 B High 30 174

B High 39 262
Notice that RStudio is telling you that PlantGrowthExpt has 18 observations of 4 variables, and
that the first two variables – variety and treatment – are factor variables – that is, they define
categories. Variety has two factor levels, A and B, and treatment has three – High, Low, and None. In
general, when you use read.csv() to import a dataset into a data frame, it will assume that any
non-numeric variables should be imported as factor variables rather than as character variables.

For importing csv files like this one, using the read.csv() function is probably the simplest option,
and we will continue to use it in this manual, However, RStudio provides support for importing files
saved in several different formats. To look at the options available, click the Import Dataset button at
the top of the Environment window.

(The first time you use any of these import options, RStudio will require you to load additional
libraries from the internet which underpin the import process.)

34
The other two data structures commonly used in R – arrays and lists - are less often needed by
beginning R users, and we only deal with them briefly here.

35
36
4 Manipulating and describing
data
This chapter uses the plantGrowthExpt, weights, and seedlings data frames created
earlier to illustrate some options for manipulating data and summarizing it numerically. The next two
chapters will review a range of ways to display data graphically.

The simplest option for describing the data held in a data frame is the summary() function, which
provides summary statistics for each vector in the data frame:

summary(plantGrowthExpt)

variety treatment height.cm biomass.gm

A:9 High:6 Min. :14.0 Min. : 65.0

B:9 Low :6 1st Qu.:21.0 1st Qu.:107.5

None:6 Median :24.5 Median :140.5

Mean :26.0 Mean :143.9

3rd Qu.:31.5 3rd Qu.:178.5

Max. :39.0 Max. :262.0

summary(weights)

gender bodyWeights

Female:500 Min. : 39.44

Male :500 1st Qu.: 64.07

Median : 72.50

Mean : 73.04

3rd Qu.: 82.10

Max. :117.98

summary(seedlings)

Year Density fitted isHigh

37
Min. :2006 Min. : 3.00 Min. : 3.976 Mode :logical

1st Qu.:2008 1st Qu.: 6.25 1st Qu.: 6.008 FALSE:5

Median :2010 Median : 8.00 Median : 8.040 TRUE :5

Mean :2010 Mean : 8.10 Mean : 8.040 NA's :0

3rd Qu.:2013 3rd Qu.: 9.75 3rd Qu.:10.071

Max. :2015 Max. :14.00 Max. :12.103

Notice that for factor and logical variables, using summary() provides the number of observations
for each level. For numeric variables (both integer and double), it provides the minimum, maximum,
first and third quartiles, and the median and mean.

This is straightforward, but doesn’t begin to cover all the ways you might want to manipulate and
summarize the data. For example, you might want to apply a log transform to the biomass data, or
calculate mean body weights for males and females separately. In fact, most data analysts spend
more time getting their data into the right form than they do on formal analysis.

Below we look at a number of techniques for manipulating, transforming and summarizing data.
Initially (section 4.1) we look at some techniques available in the R base package. Then, in section
4.2 we look at the options provided by a specialist R package, dplyr, written specifically to make
data manipulation and description easier, faster, and more flexible.

4.1 Simple data transformations using base R

4.1.1 Numeric transforms
The plantGrowthExpt data frame includes both the height and the biomass of each plant.
Suppose we want to look at how robust (or spindly) each plant is, and decide that the biomass/height
ratio is an appropriate measure. To create a new variable (bh.ratio) for this measure within the
plantGrowthExpt data frame:

plantGrowthExpt$bh.ratio = with(plantGrowthExpt,

biomass.gm / height.cm)

Similarly, to create a variable which is the square root of the biomass:

plantGrowthExpt$sqrt.biom = with(plantGrowthExpt,

sqrt(biomass.gm))

38
R includes a wide range of functions which may be used to transform data. The table below gives a
brief list of some of the most commonly used. In each case x is a number or vector of numbers.

Transform What it does When it is used

log(x) log10(x) Calculate natural and x must be positive and greater than 0
base 10 logarithms Often used to make right-skewed data more
symetric and closer to a Normal distribution,
sqrt(x) Calculate the square root x must be positive
Sometimes used to normalize somewhat right-
skewed data.
asin(x) Calculate the arcsine x must lie between 0 and 1
Sometimes used to normalize proportional data
where values are not too close to 0 or 1.
1/x Calculate the reciprocal Sometimes used to normalize very right-skewed
data.
x^2 Calculate the square Sometimes used to normalize somewhat left-
skewed data where all values have the same sign
exp(x) Calculate the exponential May be used to normalize left-skewed data or to
convert log-transformed values back to their
original form
scale(x, With the default Rescales the data so that it has a mean of 0 and a
center=TRUE, arguments, calculate the standard deviation of 1.
scale=TRUE) z-score

39
4.1.2 Logical transforms
It is often useful to create a logical variable which simply takes the value TRUE or FALSE. For
example in the plantGrowthExpt data, we might want to create a variable isLarge which has
the value TRUE if the plant's biomass is greater than the mean biomass for the whole group, and
FALSE otherwise. This is straightforward using a logical expression:

plantGrowthExpt$isLarge = with(plantGrowthExpt,

biomass.gm > mean(biomass.gm))

Then we might check whether large plants are similarly distributed between varieties: the result
suggests that larger plants may be more likely for variety B than for variety A.

varietyTable = with(plantGrowthExpt, table(variety, isLarge))

varietyTable

isLarge

variety FALSE TRUE

A 7 2

B 2 7

Use a mosaic plot to display the table.

mosaicplot(varietyTable, col=c("red","blue"))

40
4.1.3 Creating categories from numeric data
To demonstrate this we use the TullySugar.ONI csv file. It contains three vectors:

• Year (from 1945 to 1999),

• Cane.yield (measuring sugarcane yield in the Tully region for each year)
• ONI.SON. This third variable represents the Oceanic Niño Index (ONI) during September-
October-November each year. The index represents the deviation from the long-term
average of the equatorial mid-Pacific sea surface temperature at this time of year. So positive
values mean the water is warmer than average, and negative values mean that it is cooler.
The ONI is used to define the climatic phase – values of 0.5 or higher are defined as El Niño,
values of -0.5 or lower are defined as La Niña, and values between -0.5 and +0.5 are defined
as Neutral.

The command below reads in the file:

TullySugar.ONI = read.csv(file.choose())

The command below creates a variable called phase which specifies the climatic phase for each
year. It uses the cut() function to match the right phase label with each ONI.SON value.

TullySugar.ONI$phase = with(TullySugar.ONI,

cut(ONI.SON, breaks=c(-Inf,-0.49, 0.49, Inf),

labels=c("La.Nina", "Neutral", "El.Nino")))

Notice the use of –Inf and Inf to represent minus and plus infinity. The breaks define three
categories – all values up to -0.49, values between -0.49 and 0.49, and all values greater than 0.49.
Because the ONI.SON values are reported with only one decimal place, setting the break points
with an additional decimal place means that we don’t have to worry about whether values which
exactly equal a break point are inside or outside a particular category (because breaks are defined so
that no value will ever exactly equal a break point). Notice also that the categories are created as
factor data, and that the order of the levels is not alphabetic – they are in order of increasing size of
the cut variable.

Then the commands below find how many years were in each phase and display the results in a bar
plot

with(TullySugar.ONI, table(phase))

La.Nina Neutral El.Nino

41
 19 17 14

with(TullySugar.ONI, barplot(table(phase),col="slateblue",

ylab="# years", las=1, ylim=c(0,20)))

box()

20

15
# years

10

0
La.Nina Neutral El.Nino

42
5 Plotting data with base R

R has a very extensive plotting and plot-editing capability, and is particularly well suited to
exploratory data analysis. Here we expand on the introduction to plots given in chapter 2 to cover
some of the more common plotting functions which you are likely to need.

For the first few plots demonstrated below, we show you both the plot with default options and the
code changes needed to put it in its final form.

5.1 Displaying frequency distributions

5.1.1 Bar graphs for categorical data: barplot()
We demonstrated a basic barplot command in section 2.3: check its arguments using ?barplot.

Gross Domestic Product (GDP) is the per capita

value of goods and services produced Country GDP
by a country in a year.
This table shows the 2014 per-capita Australia 47600
GDP (in $US) of countries in Oceania,
as estimated by the International Federated States of
Monetary Fund. Micronesia 3000
We want to put them in ascending
order (using the arrange() Fiji 9400
function in dplyr), and then display
the table as a bar plot, with the bar Kiribati 1800
lengths proportional to the per-capita
GDP and each bar labeled with the Marshall Islands 3200
country.
New Zealand 37100
Because some country names are
long, they will be easier to read if the Palau 15300
bar plot is horizontal.
Papua New Guinea 5400

43
 Samoa 2000

Solomon Islands 1653

Tonga 5300

Tuvalu 3600

Vanuatu 2600

The data shown in the table above can be found in the file OceaniaGDP.csv. Read the file into a
data frame, check the structure of the data in the environment window and examine the first few
rows, and then use arrange() to sort the countries in order of per capita GDP:

OceaniaGDP = read.csv(file.choose())

head(OceaniaGDP)

Country per.capita.GDP

1 Australia 47600

2 Federated States of Micronesia 3000

3 Fiji 9400

4 Kiribati 1800

5 Marshall Islands 3200

6 New Zealand 37100

library(dplyr)

OceaniaGDP = arrange(OceaniaGDP, per.capita.GDP)

The names argument in barplot() can be used to specify a vector holding the bar labels – it
needs a vector of character strings, but the Country variable has been read as factor data, so the
code below creates a character vector called labels.

44
labels = as.character(OceaniaGDP$Country)

Below is a first try at the bar plot, using the las=1 argument to force the y-axis labels to be
horizontal.

barplot(OceaniaGDP$ per.capita.GDP, horiz=TRUE, names=labels, las=1)

The result is less than satisfactory! Things to change include:

1. The left margin of the plot is much too narrow for the labels.
2. The x-axis should extend to 50000 so that it extends a little past the largest bar.
3. The colour is a little uninspired.
4. The x-axis should have a label to say what the numbers mean.
5. It might be good to have a title for the plot.

The first of these is the trickiest since margins have to be set as a global graphical parameter using
the par() function. (Use ?par to check the function’s numerous arguments.) The required
argument here is mar (for margins), which takes a vector of four values, one for each margin,
starting at the bottom and moving clockwise. The default settings are 5,4,4,2, so we need to increase
the second one – to 14 should be enough. When the par() settings are changed, they stay changed
until you change them back, so it is useful to store the old settings in order to restore them later.

opar = par() ## store the default graphics settings in opar

par(mar=c(5,14,4,2)) ## increase the left margin

The remaining changes can be handled via arguments within the barplot() command.

45
barplot(OceaniaGDP$per.capita.GDP, horiz=TRUE, names=labels, las=1,

xlim=c(0, 50000), col="turquoise", xlab="Per capita GDP, 2014",

main = "Per Capita GDP in countries of Oceania")

Finally, add a frame for the plot and restore the default graphics settings

box()

par(opar)

Per Capita GDP in countries of Oceania

Australia
New Zealand
Palau
Fiji
Papua New Guinea
Tonga
Tuvalu
Marshall Islands
Federated States of Micronesia
Vanuatu
Samoa
Kiribati
Solomon Islands

0 10000 20000 30000 40000 50000

Per capita GDP, 2014

5.1.2 Stacked bar plots for proportions: barplot()

The most common way to display proportional data is a stacked bar plot. The original data will
usually be counts: although you can stack the raw counts, it is usually better to convert them to
proportions or percentages (so that they add up to 1 or to 100%) before creating the plot.

Below is a table of counts resulting from a poll about Australian voter intentions in three towns. We
will construct a stacked-bar plot which compares the percentages of each voting intention in each
town.

Location LNP Labour Green Other

Town A 2246 1877 482 205
Town B 1125 1318 404 130
46
Town C 978 525 215 101

In R, a stacked bar plot requires a matrix for data entry: the values in each column of the matrix are
stacked. Since we want each bar to represent a town, the rows in the table above will need to
become columns in the matrix The code below creates a vector of percentages for each town, uses
cbind() to combine them into a matrix called VoterIntention, and then uses the party
affiliations as row names for the matrix.

Town.A = c(2246,1877,482,205) ## create vector of voter counts

Town.A = 100*Town.A/sum(Town.A) ## convert counts to percentages

Town.B = c(1125,1318,404,130)

Town.B = 100*Town.B/sum(Town.B)

Town.C = c(978,525,215,101)

Town.C = 100*Town.C/sum(Town.C)

VoterIntention = cbind(Town.A, Town.B, Town.C)

rownames(VoterIntention)= c("LNP","Labour","Green","Other")

VoterIntention

Town.A Town.B Town.C

LNP 46.684681 37.789721 53.765805

Labour 39.014758 44.272758 28.862012

Green 10.018707 13.570709 11.819681

Other 4.281854 4.366812 5.552501

Then we can try the default plot. Setting the legend.text argument to TRUE requests the plot to
include a legend and to use the row names as labels for the legend.

barplot(VoterIntention, legend.text=TRUE)

47
 100

Other
Green
80

Labour
LNP
60
40
20
0

Town.A Town.B Town.C

Not quite right yet. We need to:

1. Make space for the legend. Since the legend is within the plot area, not in the margin, changing
the margins won’t help this time. However we can extend the x-axis to make space within the
plot area (bars are 1 unit apart: trial and error suggests x-axis limits of about 0.2 and 4.4 are OK).
2. Make the y-axis values horizontal.
3. Provide a y-axis label.
4. Change the colour scheme.
5. Add a title for the plot,
6. Add a horizontal line at y=0 to serve as an x-axis.

barplot(VoterIntention, legend.text=TRUE, xlim=c(0.2,4.4),

las=1, ylab="Percentage of respondents",

col=c("blue","red","green","grey"),

main= "Voter intentions in three towns")

abline(h=0)

48
 Voter intentions in three towns

100
Other
Percentage of respondents

80 Green
Labour
LNP
60

40

20

0
Town.A Town.B Town.C

5.1.3 Box plots for numerical data: boxplot()

Boxplots and histograms are the most common ways to display and compare the distributions of sets
of numerical values.

• Boxplots allow a simple comparison of the median, interquartile range, and total range of a
numerical variable;
• Histograms split the data values into consecutive intervals (bins) in order to plot frequencies
in each interval.

To demonstrate how these are drawn and used to compare distributions, we use a sample of blood
haemoglobin concentrations for human males living in different locations. In this section we use
boxplots, in the next, we will use histograms.

Three of the locations are at high altitude (Tibet, Ethiopia, and the Andes) and one is at sea-level
(USA). The data are available in the file haemoglobin.csv. Import the data into a new R data
frame called haemoglobin, check its structure in the Environment window, and examine its first
few rows .

haemoglobin = read.csv(file.choose())

head(haemoglobin)

group Hb

1 Ethiopia 12.47
49
2 Ethiopia 12.68

3 Ethiopia 12.54

4 Ethiopia 12.55

5 Ethiopia 12.77

6 Ethiopia 12.57

The Environment window tells us that group is a factor variable with four levels. Generating the
default boxplot is straightforward.

boxplot(Hb ~ group, data = haemoglobin, las=1)

The upper and lower boundaries of the boxes show the interquartile range. The whiskers extend an
additional 1.5 times the interquartile range. More distant points are plotted as individual outliers.
This is the most conventional form of boxplot, but other forms are possible. Run ?boxplot to
investigate alternative options.

Modifications are also straightforward. The command below adds a y-axis label and a plot title, and
changes the colour scheme.”

boxplot(Hb ~ group, data = haemoglobin, las=1,

ylab="Haemoglobin concentration (g/dl)",

main="Haemoglobin concentrations in four populations",

50
 col="orange", border="darkred")

The plot makes it very obvious that the sample from the Andes tends to have higher values than the
other three locations.

It is also possible to examine the effects of two factors simultaneously. Below we load and examine
the plantGrowthExpt data set and then plot the effect of variety and hormone treatment on plant
height. via two different layouts. Each layout draws six groups in two colours. In the first, we want
two colours to alternate, so we need only specify only two, and the pair will repeat. In the second,
we want two sets of three, so we need to specify values for all six.

plantGrowthExpt = read.csv(file.choose())

head(plantGrowthExpt, 4)

variety treatment height.cm biomass.gm

1 A None 14 73

2 A None 17 65

3 A None 20 84

4 A Low 24 119

boxplot(height.cm~variety*treatment, data=plantGrowthExpt,

col=c("gold","lightgreen"),las=1,

xlab="Variety and Treatment", ylab="Height (cm)",

51
 main="Heights of varieties A & B\n with different hormone levels")

boxplot(height.cm~treatment*variety, data=plantGrowthExpt,

col=rep(c("gold","lightgreen"), each=3), las=1,

xlab="Variety and Treatment", ylab="Height (cm)",

main="Heights of varieties A & B\n with different hormone levels")

5.1.5 Diagnostic plots

R provides a variety of plots designed to evaluate whether the assumptions of particular tests are
met. We will look at one common example here. Many commonly-used statistical tests assume that
data are Normally distributed. The qqnorm() and qqline() functions can provide a quick visual

52
check of this assumption. A quantile-quantile (Q-Q) plot plots the quamtiles of one data set against
the quantiles of another. If they come from the same distribution, the points will lie close to a
straight line with a slope of 1, through the origin. So to check if a set of data are Normally distributed
we plot the quantiles of the data against what the quantiles should be if the data really are Normally
distributed. Substantial curvature of the points away from the line would indicate a failure of the
normality assumption

## Generate a set of 100 randomly chosen data points from a Normal

## distribution, and then check their normality with a QQ plot

trialdata = rnorm(100)

qqnorm(trialdata, las=1)

qqline(trialdata, col = "red", lwd=2)

5.2 Showing relationships between two variables

5.2.1 Line plots for showing data sequences through time
We will plot the per capita CO2 emissions for Australia for the years from 2002 to 2008. The data
(from the World Bank) are provided as a csv data file called perCapitaCO2.Australia.csv:
load it into a data frame called CO2Australia. Note that it has two vectors, Year and CO2.

CO2Australia = read.csv(file.choose())

53
We provided code to produce a line plot in section 2.3. Equivalent code for this data set is shown
below (note the use of \n to spread the title over two lines:

plot(CO2 ~ Year, data=CO2Australia, type="b",

lwd=2, las=1, col="darkgreen", cex=1.2,

ylab ="Per capita CO2 emissions",

main = "Australian per capita CO2 emissions\n 2002 to 2008")

Australian per capita CO2 emissions

2002 to 2008

0.565
Per capita CO2 emissions

0.560

0.555

0.550

0.545

0.540

0.535

2002 2003 2004 2005 2006 2007 2008

Year

The y-axis label is a little cramped, the code below expands the y-axis margin slightly to
accommodate it, and uses the title()function to add the axis label (title() allows the distance
between the axis label and the axis line to be modified). (The problem could also be fixed by
reducing the size of all tick labels using cex.axis=0.8 within the plot() command – try it!)

opar = par()

par(mar=c(5,6,4,2))

plot(CO2 ~ Year, data=CO2Australia, type="b",

lwd=2, las=1, col="darkgreen",

ylab ="",

main = "Australian per capita CO2 emissions\n 2002 to 2008")

title(ylab = "Per capita CO2 emissions", line=4)

par(opar) ## warning messages can be ignored

54
 Australian per capita CO2 emissions
2002 to 2008

0.565
Per capita CO2 emissions

0.560
0.555
0.550
0.545
0.540
0.535

2002 2003 2004 2005 2006 2007 2008

Year

55