Using R for scientific computing

Karline Soetaert
Centre for Estuarine and Marine Ecology
Netherlands Institute of Ecology
The Netherlands
June 2009

Abstract
R (R Development Core Team 2008) is the open-source (read: free-of-charge) version
of the language S. It is best known as a package that performs statistical analysis and
graphics. However, R is so much more: it is a high-level language in which one can perform
complex calculations, implement new methods, and make high-quality figures.
R has high-level functions to operate on matrices, perform numerical integration, advanced statistics,... which are easily triggered and which make it ideally suited for datavisualization, statistical analysis and mathematical modeling.
It is the aim of these lecture notes to make you acquainted with the R language. The
lecture notes are based on a book (Soetaert and Herman 2009) about ecological modelling
in which R is extensively used for developing, applying and visualizing simulation models.
There are many excellent sources for learning the R (or S) language. R comes with
several manuals that can be consulted from the main R program (Help/Manuals). Rintro.pdf is a good start. Many other good introductions to R are available, some freely
on the web, and accessible via the R web site (www.r-project.org). My favorite is the
R introduction by Petra Kuhnert and Bill Venables (Kuhnert and Venables 2005), but
beware: this introduction comprises more than 300 pages!

Keywords:Variables, Functions, Figures, Interpolation, Fitting, Roots, Ordinary differential

equations, R.

1. introduction
1.1. The R-software
Installing R
R is downloadable from the following web site: http://www.r-project.org/ Choose the precompiled binary distribution. On this website, you will also find useful documentation. To
use R for the examples in this course, several packages need to be downloaded.
deSolve. Performs integration. (Soetaert, Petzoldt, and Setzer 2009c)
rootSolve. Finds the root of equations (Soetaert 2009).

Using R for scientific computing

scatterplot3d. For 3-D graphics. (Ligges and Machler 2003)
seacarb. Aquatic chemistry. (Proye, Gattuso, Epitalon, Gentili, Orr, and Soetaert 2007)
marelac. Functions and constants from the marine and lacustrine sciences (Soetaert,
Petzoldt, and Meysman 2009a).
marelacTeaching. Data sets, used in this tutorial. (Soetaert, Petzoldt, and Meysman
2009b).

If you run R within windows, downloading specific packages can best be done within the R
program itself. Select menu item packages / install packages, choose a nearby site (e.g.
France (Paris)) and select the package you need. If you install package marelacTeaching then
all other packages will be automatically installed as well.

Other useful software

I prefer to run R from within the Tinn-R editor, which can be downloaded from URL
http://sourceforge.net/projects/tinn-r and http://www.sciviews.org/Tinn-R. This editor provides R-sensitive syntax and help. Download the latest Tinn-R setup file and install it. From
within the Tinn-R program, you launch R via the menu (R/start preferred Rgui).

1.2. Quick overview of R

R-code is highly readable, once you realise that:
<- is the assignment operator.
everything starting with # is considered a comment.
R is case-sensitive: a and A are two different objects.

1.3. Console versus scripts

There are two ways in which to work with R.
1. We can type commands into the R console window at the command prompt (>) and
use R as a powerful scientific calculator.
For instance, enter in the console window:
> pi*0.795^2 ; 25*6/sqrt(67) ; log(25)
[1] 1.985565
[1] 18.32542
[1] 3.218876

Karline Soetaert

Here sqrt and log are built-in functions in R; pi is a built-in constant; the semi-colon
(;) is used to separate R-commands.
In the console window, the <UP> and <DOWN> arrow keys are used to navigate
through previously typed sentences.
2. Alternatively, we can create R-scripts in an editor (e.g. Tinn-R) and save them in a file
(filename.R) for later re-use. R-scripts are sequences of R-commands and expressions.
These scripts should be submitted to R before they are executed. This can be done in
several ways:
by typing, in the R-console window:
> source ("filename.R")
by opening the file, copying the R-script to the clipboard (ctrl-C) and pasting it
(ctrl-V) into the R-console window
If you do not use the tinn-R editor, the file is opened as an R-script from within the
R console. After selecting the script, and pressing the send button the statements
are executed and the cursor moved to the next line.
you can
If you do use the Tinn-R editor,
either submit the entire file (buttons 1,2), selected parts of the text (buttons 3,4),
submit marked blocks (buttons 5,6) or line-by-line (last button).
Throughout these notes, the following convention is used:
> 3/2
denotes input to the console window (> is the prompt)
[1] 1
is R output, as written in the console window
getwd()
is an R statement in a script file (it lacks the prompt).
A screen capture of a typical Tinn-R session, with the Tinn-R editor (upper window) and the
R-console (lower window) is given below. A script file is opened in the Tinn-R editor. Note
the context-sensitive syntax (green=comments, blue= reserved words, rose = R-parameters).
Several lines of R-code have been selected (blue area) and sent to the R-console, which has
produced the graphics window that floats independently from the other windows.

Using R for scientific computing

Getting help, examples, demonstrations

R has an extensive help facility. Apart from the Help window launched from the Help menu,
or the HTML help facility, it is also available from the command line prompt. For instance,
typing
>
>
>
>
>

?log
?sin
?sqrt
?round
?Special

will explain about logarithms and exponential functions, trigonometric functions, and other
functions.
> ?Arithmetic
lists the arithmetic operations in R.
> help.search("factor")
will list occurrences of the word <factor> in R-commands.
Sometimes the best help is provided by the very active mailing list. If you have a specific problem, just type R: <problem> on your search engine. Chances are that someone encountered
the problem and it was already solved.

Karline Soetaert

Most of the help files also include examples. You can run all of them by using R-statement
example.
For instance, typing into the console window:
> example(matrix)
will run all the examples from the matrix help file.
> example(pairs)
will run all the examples from the pairs help file. (! try this ! pairs is a very powerful way
of visualizing pair-wise relationships).
Alternatively, you may select one example, copy it to the clipboard (ctrl-C for windows users)
and then paste it (ctrl-V) in the console window. In addition, the R main software and many
R-packages come with demonstration material. Typing
> demo()
will give a list of available demonstrations in the main software.
> demo(graphics)
will demonstrate some simple graphical capabilities.

Small things to remember

Pathnames in R are written with forward slashes (/) , although in windows, backslashes, (\), are used. Thus, to set a working directory in R:
setwd("C:/R code/")
If a sentence on one line is syntaxically correct, R will execute it, even if it is the intention
that it proceeds on the next line. For instance if we write:
>A <- 3 + cos(pi)
>
- sqrt(5)
[1] -2.236068
!
then A will get the value (3 + cos()) and R will print the value of ( (5)). In contrast,
in the next lines:
> 3 + cos(pi) >
sqrt(5)
[1] -0.236068

Using R for scientific computing

!
R will print the value of 3 + cos() (5): as the sentence on the first line was not
syntaxically finished, R has (correctly) assumed that it continued on the next line.

Be careful if you want to split a complex statement over several lines ! These errors are
very difficult to trace, so it is best to avoid them.

Exercises. Using R as a calculator

It is very convenient to use R as a powerful calculator. This can best be done from within
the R-console.
1. Use the console to calculate the value of:
(4/6 8 1)2/3
ln(20)

log2 (4096)
23
!

2.32 + 5.42 2 2.3 5.4 cos(/8)

tip: you may need to look at the help files for some of these functions. typing ?"+" will
open a help file with the common arithmetic operators.
2. Now write the R-statements in a script file, using the Tinn-R editor. Try the various
ways in which to submit the statements to R .

Karline Soetaert

2. R-variables
R calculates as easily with vectors, matrices and arrays as with single numbers.
R also includes more complex structures such as data frames and lists, which allow to combine
several types of data.
Learning how to create these variables, how to address them and modify them is essential if
you want the make good use of the R software.

2.1. Numbers, vectors, matrices and arrays

Assignment
When variables are used, they need to be initialised with numbers.
>A <- 1
>B <- 2
>A+B
[1] 3
R can take as arguments for its functions single numbers, vectors, matrices, or arrays.
>V<-factorial(10)
Calculates 10!(= 1 2 3 4 5 . . . 10). The operator <- assigns the result of this calculation
to variable V. V can then be used in subsequent calculations:
>V/10
[1] 362880
Note that the assignment of a value to V does not display it on the window. To display V we
simply write:
> V
[1] 3628800
Alternatively, we may assign the result of calculations to a variable AND view the results, by
embracing the statement between brackets:
> (X <- sin(3/2*pi) )
[1] -1
Apart from integers, real and complex numbers, R also recognizes infinity (Inf) and Not a
Number (NaN). Try:

Using R for scientific computing

> 1/0
> 0/0
> 1e-8 * 1000
(where the e-8 notation denotes 108 ).

Vectors
Vectors can be created in many ways:
Using R-function vector
The function c() combines numbers into a vector

The operator : creates a sequence of values, each 1 larger (or smaller) than the previous
one
A more general sequence can be generated by R-function seq
The same quantity is repeated using R-function rep
For instance, the commands:
>c(0, pi/2, pi, 3*pi/2, 2*pi)
[1] 0.000000 1.570796 3.141593 4.712389 6.283185
>seq(from=0,to=2*pi, by=pi/2 )
[1] 0.000000 1.570796 3.141593 4.712389 6.283185
>seq(0, 2*pi, pi/2 )
[1] 0.000000 1.570796 3.141593 4.712389 6.283185
will all create a vector, consisting of: 0, , . . . 2 .

Note that R-function seq takes as input (amongst others) parameters from, to and by (2nd
example). If the order is kept, they not be specified by name (3rd example).
The next command calculates the sine of this vector and outputs the result:
>sin( seq(0, 2*pi, pi/2 ))
[1] 0.000000e+00
[5] -2.449294e-16

1.000000e+00

1.224647e-16 -1.000000e+00

Function rep is used to repeat elements:

1

This is perhaps THE most important function in R

Karline Soetaert
>rep(1,times=5)
[1] 1 1 1 1 1
>rep(c(1,2),times=5)
[1] 1 2 1 2 1 2 1 2 1 2
>c(rep(1,5),rep(2,5))
[1] 1 1 1 1 1 2 2 2 2 2
The next statements:
> V <- 1:20
> sqrt(V)
[1] 1.000000 1.414214 1.732051 2.000000 2.236068 2.449490 2.645751
[8] 2.828427 3.000000 3.162278 3.316625 3.464102 3.605551 3.741657
[15] 3.872983 4.000000 4.123106 4.242641 4.358899 4.472136

create a sequence of integers between 1 and 20 and take the square root of all of them,
displaying the result to the screen. The operator <- assigns the sequence to V.
Some other examples of the : operator are:
>(V <- 0.5:10.5)
[1]

0.5

1.5

2.5

3.5

4.5

5.5

6.5

7.5

8.5

9.5 10.5

>6:1
[1] 6 5 4 3 2 1
Finally, the statements:
>V <- vector(length=10)
>FF<- vector()
generate a vector V comprising 10 elements, and a vector FF of unknown length.
Note: a peculiar feature of R is that the elements of a vector can also be given names:
>(fruit <- c(banana=1, apple=2, orange =3))
banana
1

apple orange
2
3

10

Using R for scientific computing

>names(fruit)
[1] "banana" "apple"

"orange"

Matrices
Matrices can also be created in several ways:
By means of R-function matrix
By means of R-function diag which constructs a diagonal matrix
The functions cbind and rbind add columns and rows to an existing matrix, or to
another vector
The statement:
>A <-matrix(nrow=2,data=c(1,2,3,4))
creates a matrix A, with two rows, and, as there are four elements, two columns. Note that
the data are inputted as a vector (using the c() function).
The next two statements display the matrix followed by the square root of its elements:
>A

[1,]
[2,]

[,1] [,2]
1
3
2
4

>sqrt(A)
[,1]
[,2]
[1,] 1.000000 1.732051
[2,] 1.414214 2.000000
By default, R fills a matrix column-wise (see the example above). However, this can easily be
overruled, using parameter byrow:
>(M <-matrix(nrow=4, ncol=3, byrow=TRUE, data=1:12))

[1,]
[2,]
[3,]
[4,]

[,1] [,2] [,3]

1
2
3
4
5
6
7
8
9
10
11
12

The unity matrix (I) is created using R-function diag:

Karline Soetaert

11

>diag(1,nrow=2)
[,1] [,2]
[1,]
1
0
[2,]
0
1
The names of columns and rows are set as follows:
>rownames(A) <- c("x","y")
>colnames(A) <- c("c","b")
>A
c b
x 1 3
y 2 4
note that we also use the c() function here ! Row names and column names are in fact vectors
containing strings.
Matrices can also be created by combining (binding) vectors, e.g. rowwise:
>V <- 0.5:5.5
>rbind(V,sqrt(V))
[,1]
[,2]
[,3]
[,4]
[,5]
[,6]
V 0.5000000 1.500000 2.500000 3.500000 4.500000 5.500000
0.7071068 1.224745 1.581139 1.870829 2.121320 2.345208
t(A) will transpose matrix A (interchange rows and columns).
>t(A)
x y
c 1 2
b 3 4

Arrays
Arrays are multidimensional generalizations of matrices; matrices and arrays in R are actually
vectors with a dimension attribute. A multi-dimensional array is created as follows:
>AR <-array(dim=c(2,3,2),data=1)
In this case AR is a 2*3*2 array, and its elements are all 1.

2.2. Dimensions
The commands

12
>
>
>
>

Using R for scientific computing

length(V)
dim(A)
ncol(M)
nrow(M)

Will return the length (total number of elements) of (vector or matrix) V, the dimension of
matrix or array A, and the number of columns and rows of matrix M respectively.

2.3. Selecting and extracting elements

To select subsets of vectors or matrices, we can either
specify the numbers of the elements that we want (simple indexing)
specify a vector of logical values (TRUE/FALSE) to indicate which elements to include
(TRUE) and which not to include (FALSE). This uses logical expressions

Simple indexing
The elements of vectors, matrices and arrays are indexed using the [] operator:
M[1 , 1]
M[1 , 1:2]
M[1:3 , c(2,4)]
Takes the element on the first row, first column of a matrix M (1st line), then selects the
entries in the first row and first two columns (2nd line) and then the elements on the first
three rows, and 2nd and 4th column of matrix M (3rd line).
If an index is omitted, then all the rows (1st index omitted) or columns (2nd index omitted)
are selected. In the following:
M[
,2] <-0
M[1:3, ] <- M[1:3, ] * 2
first all the elements on the 2nd column (1st line) of M are zeroed and then the elements on
the first three rows of M multiplied with 2 (2nd line). Similar selection methods apply to
vectors:
V[1:10]
V[seq(from=1,to=5,by=2)]
The statement on the 1st line takes the first 10 elements of vector V, whilst on the 2nd line,
the 1st , 3rd and 5th element of vector V are selected.

Logical expressions
Logical expressions are often used to select elements from vectors and matrices that obey
certain criteria.
R distinguishes logical variables TRUE and FALSE, represented by the integers 1 and 0.

13

Karline Soetaert
> ?Comparison
> ?Logic

will list the relational and logic operators available in R. The following will return TRUE for
values of sequence V that are positive:
>(V <- seq(-2,2,0.5))
[1] -2.0 -1.5 -1.0 -0.5

0.0

0.5

1.0

1.5

2.0

> V>0
[1] FALSE FALSE FALSE FALSE FALSE

TRUE

TRUE

TRUE

TRUE

while
>V [V > 0]
[1] 0.5 1.0 1.5 2.0
will select the positive values from V,
>V [V > 0] <- 0
will zero all positive elements in V,
> sum(V < 0)
[1] 4
will return the number of negative elements: it sums the TRUE (=1) values, and
>V [V != 0]
[1] -2.0 -1.5 -1.0 -0.5
will display all nonzero elements from V ( ! is the not operator). Logical tests can also be
combined, using | (the or operator), and & (and).
>V [V<(-1) | V>1]
[1] -2.0 -1.5
will display all values from V that are < -1 and > 1. Note that we have enclosed -1 between
brackets (can you see why this is necessary?) Finally,
>which (V == 0)

14

Using R for scientific computing

[1] 5 6 7 8 9
>which.min (V)
[1] 1
will return the element index of the 0-value, and of the minimum.

2.4. Removing elements

When the index is preceded by a -, the element is removed.
> M[ ,-1]
Will show the contents of matrix M, except the first column.
> x <- x[-1]
> M <- M[-1, ]
> V <- V[-V>=0]
will delete the 1st element of x, (1st line), the 1st row of M (2nd line), and the positive elements
of V (3rd line). For more information, type
> ?Extract

2.5. More complex data structures

R also allows creating more complex structures such as data frames and lists.

lists
A list is a combination of several objects; each object can be of different length:
> list(Array = AR, Matrix = M)
will combine the previously defined array AR and matrix M.

data.frames
These are combinations of different data types (e.g. characters, integers, logicals, reals),
arranged in tabular format:
>genus <- c("Sabatieria","Molgolaimus")
>dens <- c(1,2)
>Nematode <-data.frame(genus=genus,density=dens)
>Nematode

Karline Soetaert

15

genus density
1 Sabatieria
1
2 Molgolaimus
2
In the example above, the data.frame Nematode contains two columns, one with strings (the
genus name), one with values (the densities). Data.frames are in fact special cases of lists,
consisting of vectors with equal length. Many matrix-operations work on data.frames with a
single data type, but there exist also special operations on data.frames.

Selecting data from data.frames and lists

Data.frames and lists can be accessed by their names, or by the [ ] or [[ ]] operator. The
object resulting from a selection using single brackets [ ], will be a data.frame respectively
a list itself; with double brackets [[ ]], one obtains a vector (data.frames) or a variable
data-type (lists).
For instance:
>Nematode$density/sum(Nematode$density)
[1] 0.3333333 0.6666667
will divide all density values (1,2) by the summed density.
>mean(Nematode[,2])
[1] 1.5
will calculate the mean of Nematode density (the 2nd column).
Try:
> Nematode$genus
> Nematode[1]
> Nematode[[1]]
These statements will all output the two genus names, but in a different format.
> ?Extract
also explains the various ways in which to extract elements from lists and data frames.

2.6. Data Conversion

Conversion from one type of data structures to another can easily be done, e.g. by:
> as.data.frame(M)
> as.vector(A)
If unsure about the type, you can write:

16

Using R for scientific computing

> is.data.frame(M)
> is.vector(A)
Or you can display the data type by:
> class(M)

2.7. Data import from external sources

In all previous examples, data was entered from the console (or from a script file). There
are ample facilities to import data from external sources. Most often, we will use functions
read.table, read.csv, or read.delim to read matrices or data frames written in tabular
form as text files. R also has plenty of built-in data sets. They are listed by:
> data()

2.8. Exercises
Creating and manipulating matrices and vectors is essential if we want to use R as a mathematical tool. Although this has been implemented in a consistent way in R , it is not simple
for novice users! Practice is the best teacher, so you will get plenty of exercise.
Most of the exercises can be answered with one single R -statement. However, as these
statement smay be quite complicated, it is often simpler to first break them up into smaller
parts, after which they are merged into one.

Vectors, sequences.
Use R-function mean to estimate the mean of two numbers, 9 and 17. (you may notice
that this is not as simple as you might think!).
Vector V
Create a vector, called V, with even numbers, between 16 and 56. Do not use
loops. (tip: use R-function seq )
Display this vector
What is the sum of all elements of V? Do not use loops; there exists an R-function
that does this; the name of this function is trivial.
Display the first 4 elements of V
Calculate the product of the first 4 elements of V
Display the 4th , 9th and 11th element of V . (tip: use the c() function).
Vector W
Create a new vector, W, which equals vector V, multiplied with 3; display its
content.

Karline Soetaert

17

How many elements of W are smaller than 100?

First create a new vector that contains only the elements from W<100 (call it
W100), then calculate the length of this new vector.
Now perform the same calculation, in one R statement.
Create a sequence that contains the values (1,1/2,1/3,1/4, . . ., 1/10)
Compute the square root of each element
Compute the square (2 ) of each element
Create a sequence with values (0/1,1/2,2/3,3/4, . . ., 9/10)
Vector U
Create a vector, U, with 100 random numbers, uniformly distributed between -1
and 1.
tip: R-statement runif generates uniformly distributed random numbers; use
?runif to see how it works.
Check the range of U; all values should be within -1 and +1.
tip: there exists an R-function to do that-its name is trivial.
Calculate the sum and the product of the elements of U
How many elements of U are positive?
Zero all negative values of U.
Sort U
Vectors x, y
Create two vectors: vector x, with the elements: 2,9,0,2,7,4,0 and vector y with
the elements 3,5,0,2,5,4,6 (in that order). (tip: use the c() function).
Divide all the elements of y by the elements of x.
Type in the following commands; try to understand:
x>y
x==0
Select all values of y that are larger than the corresponding values of x
Select all values of y for which the corresponding values of x are 0.
Remove all values of y for which the corresponding values of x equal 0.
Zero all elements of x that are larger or equal than 7. Show x.

Matrices
Use R -function matrix to create a matrix with the following contents:
"
#
3 9
7 4

18

Using R for scientific computing

display it to the screen
Use R -function matrix to create a matrix called A:

1 1/2 1/3

1/4 1/5 1/6

1/7 1/8 1/9

Take the transpose of A.

Create a new matrix, B, by extracting the first two rows and first two columns of
A. Display it to the screen.
Use diag to create the following matrix, called D:

1 0 0

0 2 0
0 0 3

Use cbind and rbind to augment this matrix,

1 0 0

0 2 0

0 0 3

such that you obtain:

4
5 5 5 5

It is simplest to do this in two statements (but it can be done in one!)

Remove the second row and second column of the previous matrix

Diversity of deep-sea nematodes

We will now work on a data set consisting of nematode species densities, found in Mediterranean deep-sea sediments, at depths ranging from 160 m to 1220 m. The densities are
expressed in number of individuals per 10 cm2.
Nematodes are small (<1mm) worms, and they are generally very abundant in all marine
sediments.
The data (from (Soetaert, Heip, and Vincx 1991), originally present in an ACCESS database,
have been exported as a table in a comma-separated-values (so-called csv) format. This
format has the advantage that it can easily be read by text editors (such as the TINN-R
editor) as well as in spreadsheet programs.
Open the file nemaspec.csv in Tinn-R2 .
Check its structure.
2
: if you do not have this file, it can be found in the marelacTeaching package directory. Within R, type:
browseURL(paste(system.file(package=marelacTeaching), /lecture/, sep=))

Karline Soetaert

19

You may also open the file in EXCEL, but do not forget to close it before proceeding. EXCEL
is very territorial, and will not allow another program, such as R , to access a file that is open
in EXCEL.
On the first line is the heading (the names of the stations), the first column contains the
species names. Before importing the file in R , check the working directory:
> getwd()
If the file called nemaspec.csv is not in this directory, you may need to change the working
directory:
> setwd("directory name")
(do not forget that R requires / where windows uses \) .

Make a script file in which you write the next steps; submit each line to R to check its
correctness. Read the comma-delimited file, using R-command read.csv. Type ?read.csv
if you need help.
Specify that the first row is the heading (header=TRUE) and the first column contains the
rownames (row.names=1).
Put the data in data.frame Nemaspec.
Nemaspec <- read.csv("nemaspec.csv", header=TRUE, row.names=1)
Check the contents of Nemaspec. As the dataset is quite substantial, it is best to output only
the first part of the data:
head(Nemaspec)
The rest is up to you:
Select the data from station M160b (the 2nd column of Nemaspec); put these data in a
vector called dens.
(remember: to select a complete column, you select all rows by leaving the first index
blanc).
Remove from vector dens, the densities that are 0. Display this vector on the screen.
(Answer: [1] 6.580261 5.919719 etc. . .)
Calculate N, the total nematode density of this station. The total density is simply the
sum of all species densities (i.e. the sum of values in vector dens). What is the value of
N ? (Answer :699).
Divide the values in vector dens by the total nematode density N. Put the results in
vector p, which now contains the relative proportions for all species. The sum of all
values in p should now equal 1. Check that.
Calculate S, the number of species: this is simply the length of p; call this value S.
(Answer: S=126)

20

Using R for scientific computing

Estimate the values of diversity indices N1 and N2 and Ni, given by the following
formulae:
P
N 1 = e pi loge (pi )
*+ 2 ,
N 2 = 1/
pi
N i = 1/ max(pi )

You can calculate each of these values using only one R statement ! (A: 90.15358,
66.77841, 22.56157)
The 126 nematode species per 10 cm2 were obtained by looking at all 699 individuals.
Of course, the fewer individuals are determined to species, the fewer species will be
encountered. Some researchers determine 100 individuals, other 200 individuals. To
standardize their results, the expected number of species in a sample can be recalculated
based on a common number of individuals. The expected number of species in a sample
with size n, drawn from a population which size N, which has S species is given by:
"
*N N i , #
S
ES(n) =
1 n N
(n )
i=1

where Ni is the number of individuals in the ith species in the full sample and is the
so-called binomial coefficient, the number of different sets with size n that can be
chosen from a set with total size N.
In R, binomial coefficients are estimated with statement choose(N,n).
What is the expected number of species per 100 individuals ? (n=100,N=699). (A:
ES(100) = 60.68971).
Print all diversity indices to the screen, which should look like:
N
N0
699.00000 126.00000

N1
90.15358

N2
66.77841

Ni
22.56157

ESS
60.68971

Karline Soetaert

21

3. R functions
One of the strengths of R is that one can make user-defined functions that add to R -s built-in
functions.

3.1. Function definition

Typically, complex functions are written in R -script files, which are then submitted to R .
For instance,
Circlesurface <- function (radius) pi*radius^2
defines a function (called Circlesurface) which takes as input argument a variable called
radius and which returns the value radius2 (which is the surface of a circle).

After submitting this function to R , we can use it to calculate the surfaces of circles with
given radius:
>Circlesurface(10)
[1] 314.1593
>Circlesurface(1:20)
[1]
[6]
[11]
[16]

3.141593
113.097336
380.132711
804.247719

12.566371
28.274334
50.265482
78.539816
153.938040 201.061930 254.469005 314.159265
452.389342 530.929158 615.752160 706.858347
907.920277 1017.876020 1134.114948 1256.637061

the latter statement will calculate the surface of circles with radiuses 1, 2, ... ,20.
More complicated functions may return more than one element:
Sphere <- function(radius)
{
volume <- 4/3*pi*radius^3
surface <- 4 *pi*radius^2
return(list(volume=volume,surface=surface))
}
Here we recognize
the function heading (1st line), specifying the name of the function (Sphere) and the
input parameter (radius)
the function specification. As the function comprises multiple statements, the function
specification is embraced by curly braces {. . .}.
The return values (last line). Sphere will return the volume and surface of a sphere, as
a list.

22

Using R for scientific computing

The earth has approximate radius 6371 km, so its volume (km3) and surface (km2) are:
>Sphere(6371)
$volume
[1] 1.083207e+12
$surface
[1] 510064472
The next statement will only display the volume of spheres with radius 1, 2, . . . 5
>Sphere(1:5)$volume
[1]

4.18879

33.51032 113.09734 268.08257 523.59878

Sometimes it is convenient to provide default values for the input parameters.

For instance, the next function estimates the density of standard mean ocean water (in
kgm3 ), as a function of temperature, T, (and for salinity=0, pressure = 1 atm) (Millero
and Poisson 1981); the input parameter T is by default equal to 20 C:
Rho_W <- function(T=20)
{
999.842594 + 0.06793952 * T - 0.00909529 * T^2 +
0.0001001685 * T^3 - 1.120083e-06 * T^4 + 6.536332e-09 * T^5
}
By ending the first sentence with a + we made clear that the statement is not finished and
continues on the next line. It would have been wrong to put the + on the next line.
Calling the function without specifying temperature, uses the default value:
>Rho_W()
[1] 998.2063
>Rho_W(20)
[1] 998.2063
>Rho_W(0)
[1] 999.8426

3.2. Programming
R has all the features of a high-level programming language:

If, else, ifelse constructs

Try to understand the following:

Karline Soetaert
Dummy <- function (x)
{
if ( x<0 ) string <- "x<0"
if ( x<2 ) string <- "0>=x<2"
string <- "x>=2"
print(string)
}

23

else
else

>Dummy(-1)
[1] "x<0"
>Dummy(1)
[1] "0>=x<2"
>Dummy(2)
[1] "x>=2"
Note that we have specified the else clause on the same line as the if part so that R knows
that the statement is continued on the next line!
If and else constructs involving only one statement can be combined:
>x<-2
>ifelse (x>0, "positive", "negative,0")
[1] "positive"

Loops
Loops allow a set of statements to be executed multiple times:
The for loop iterates over a specified set of values. In the example below, the variable i
takes on the values (1,2,3):
>for (i in 1:3) print(c(i,2*i,3*i))
[1] 1 2 3
[1] 2 4 6
[1] 3 6 9
while and repeat will execute until a specified condition is met.
>i<-1 ; while(i<3) {print(i); i<-i+1}

24

Using R for scientific computing

[1] 1
[1] 2
break exits the loop
next stops the current iteration and advances to the next iteration.
>i<-1
>repeat
+ {
+
print(i)
+
i <-i+1
+
if(i>2) break
+ }
[1] 1
[1] 2
The curly braces {. . .} embrace multiple statements that are executed in each iteration.

Note: loops are implemented very inefficiently in R and should be avoided as often as possible.
Fortunately, R offers many high-level commands that operate on vectors and matrices. These
should be used as much as possible!
For more information about if constructs and loops, type
> ?Control

3.3. R- packages
A package in R is a file containing many functions that perform certain related tasks. Packages
can be downloaded from the R website.
Once installed, we generate a list of all available packages, we load a package and we obtain
a list with its contents by the following commands:
>library()
>library(deSolve)
>library(help=deSolve)
>help(package=deSolve)

3.4. Exercises
R-function sphere
Extend the Sphere function with the circumference of the sphere at the place of maximal
radius. The formula for estimating the circumference of a circle with radius r is: 2 r.
What is the circumference of the earth near the equator?

Karline Soetaert

25

An R-function to estimate saturated oxygen concentrations

The saturated oxygen concentration in water (tmol kg-1), as function of temperature (T),
and salinity (S) can be calculated by: SatOx = eA where :
A= -173.9894 + 25559.07/T + 146.4813* loge(T/100) -22.204*T/100 + S *
(-0.037362+0.016504*T/100-0.0020564 *T/100*T/100)
and T is temperature in Kelvin (Tkelvin = Tcelsius+273.15).
Tasks:
Make a function that implements this formula; the default values for temperature and
salinity are 20 C and 35 respectively.
What is the saturated oxygen concentration at the default conditions? (A: 225.2346)
Estimate the saturated oxygen concentration for a range of temperatures from 0 to
30rC, and salinity 35. (Tip: no need to use loops).

Loops
The Fibonacci numbers are calculated by the following relation: Fn = Fn1 + Fn2
with F1 = F2 = 1
Tasks:
Compute the first 50 Fibonacci numbers; store the results in a vector (use R-command
vector to create it). You have to use a loop here

For large n, the ratio Fn /Fn1 approaches the golden mean: (1 + 5)/2
What is the value of F50 /F49 ; is it equal to the golden mean?
When is n large enough? (i.e. sufficiently close (<1e6 ) to the golden mean)

Diversity of deep-sea nematodes for all stations

Starting from your code to estimate diversity indices for deepsea station M160b, now
write a loop that does so for all the stations in Nemaspec.
First create a matrix called div, with the number of rows equal to the number of deepsea
stations, and with 6 columns, one for each diversity index. This matrix will contain the
diversity values.
The column names of div are: N, N0, N1, N2, Ninf, ESS
The row names of matrix div are the station names (= the column names of Nemaspec).
Tip: Use R-command colnames(), rownames()
Now loop over all columns of data frame Nemaspec, estimate the diversity indices and
put the results in the correct row of matrix div:

26

Using R for scientific computing

for (i in 1:ncol(Nemaspec))
{
# you have to write this part of the code
}
Show matrix div to the screen

Diversity indices as a function

Based on the results obtained in previous section, make a function that will calculate the
diversity indices for any data matrix.

Rarefaction diversity
If you still have time and the courage: try an alternative way of estimating the number of
species per 100 individuals by taking random subsamples of 100 individuals and estimating
the number of species from this subsample.
3

If the procedure is repeated often enough, the mean value should converge to the expected
number of species, ESS(100); this is the rarefaction method of (Sanders 1968).
You may need the following R -functions:
round (converting reals to integers),
cumsum (take a cumulative sum),
sample (take random selection of elements),
table (to make a table of counts),
as well as length, mean.
(Hurlbert 1971) showed that rarefaction generally overestimates the true estimated number
of species ; can you corroborate this finding?

This question requires significant thought and imagination; there are several ways to do this.

Karline Soetaert

27

4. Statistics
R originated as a statistical package, and it is still predominantly used for this purpose.
You can do virtually any statistical analysis in R .
As there exist many documents that may help you with statistical analyses in R, we will not
deal with the subject here.
Statistics is used just to show you how to use efficiently use R , in cases where you have no
clue where to begin!

4.1. Using R in four steps

Suppose you want to perform a hierarchic clustering and plot the dendrogram of a multivariate
data set.
If you have never done that in R here are the steps:
1. Find a function that performs the requested task.
for instance, use help.search ("cluster") to help you.
Depending on the number of packages that you have installed, R will list a number of
possible functions whose help file contains the word cluster. Use the function from the
package stats (part of the core of R).
2. Open the help file (?<name-of-the-function>) and look up the syntax for this function.
If you have no time to read it completely, at least read (part of) the section Description,
Usage, and Examples.
3. Try the examples in the help file. You may:
Try them all at once (example(<name-of-the-function>).
Alternatively, you may select the statements in the Examples section that look
applicable to your problem, copy-paste them into your script file (Ctrl-C / Ctrl-V)
and execute them; e.g. line by line. Chances are real that, if they are suited for
your case, you will transform them anyway.
4. Transform a promising example so that it suits your problem.

4.2. Exercise: multivariate statistics on the nematode species data.

Use R to perform a multivariate statistical analysis of the nematode data.
Beware: the nematode data have stations as columns and species as rows.
Perform a hierarchic clustering and plot the dendrogram
Perform a principal component analysis (PCA) and plot the results; you may also repeat
the PCA analysis, with the first two stations removed!

28

Using R for scientific computing

5. Graphics
R has extensive graphical capabilities, and allows making simple (1-D, x-y), image-like (2-D)
and perspective (3-D) figures.
Try:
> demo(graphics)
> demo(image)
> demo(persp)
simple (1-D, x-y), image-like (2-D) and perspective (3-D) capabilto obtain a display of RSs
ities.
Graphics are plotted in the figure window which floats independently from the other windows.
If not already present, it is launched by writing (in windows):
> windows()
or
> x11()
A figure consists of a plot region surrounded by 4 margins, which are numbered clockwise,
from 1 to 4, starting from the bottom. R distinguishes between:
1. high-level commands. By default, these create a new figure, e.g.
hist, barplot, pie, boxplot, ... (1-D plot)
plot, curve, matplot, pairs,... ((x-y)plots)
image, contour, filled.contour,... (2-D surface plots)
persp, scatterplot3d,... (3-D plots) 4 .
2. low-level commands that add new objects to an existing figure, e.g.
lines, points, segments, polygon, rect, text, arrows, legend, abline, locator,
rug, ... These add objects within the plot region
box, axis, mtext (text in margin), title, ... which add objects in the plot margin
3. graphical parameters that control the appearance of.
plotting objects:
cex (size of text and symbols), col (colors), font, las (axis label orientation), lty

(line type), lwd (line width), pch (the type of points),E

graphic window:
mar (the size of the margins), mfrow (the number of figures on a row), mfcol

(number figures on a column),E

4

scatterplot3d is in R -package scatterplot3d which has to be loaded first

29

0.0

1.0

0.5

sin(a)

0.5

1.0

Karline Soetaert

1.0

0.5

0.0

0.5

1.0

cos(a)

Figure 1: Simple figure with plot - see text for R-code

>
>
>
>

?plot.default
?par
?plot.window
?points

will open the help files, while

> example(plot.default)
> example(points)
will run the examples, displaying each new graph, after you have pressed <ENTER> (try
it!)

5.1. X-Y plots

A circle can be plotted by (x,y) points, where x = r cos(a) and y = r sin(a), with a the
angle, from 0 to 2, and r the radius. In the following script, we first generate a sequence of
angle values, a, from 0 to 2, comprising 100 values (length.out) and then plot a circle with
unit radius:
>a <- seq(0,2*pi, length.out=100)
>plot(x=cos(a),y=sin(a))
As plot is a high-level command, it starts a new figure.
By default, R adds axes, and labels, and represents the (x,y) data as small dots (points). Note
that the graph is not symmetrical.
We will now make a more complex figure that resembles a target face, e.g. for practicing
archery or to throw darts. We first use the same command (plot) as above, but we add a
number of graphical parameters that specify that:
Rather than dots, the points should be connected by lines (type).

30

Using R for scientific computing

The line should be twice as wide as the default (lwd)
The x- and y-axes labels (xlab,ylab) have to be empty
The axes and axes annotations (axes) are removed
The graph has to be symmetrical, i.e. the x/y aspect ratio = 1 (asp).

plot(cos(a),sin(a),type="l",lwd=2,xlab="",ylab="",axes=FALSE,
asp=1)
To this figure, we can now add several low-level objects:
a series of lines, representing smaller and smaller circles (lines).
for (i in seq( 0.1,0.9,by=0.1)) lines(i*sin(a), i*cos(a))
an innermost red polygon (polygon).
polygon(sin(a)*0.1,cos(a)*0.1,col="red")
point marks as text labels, ranging from from 10 to 1 (text). The closer to the centre,
the higher the score
for (i in 1:10) text(x=0,y=i/10-0.025,labels=11-i,font=2)
Now two archers take 10 shots at the target face.
We mimic their arrows by generating normally distributed (x,y) numbers, with mean=0
(the centre!) and where the experience of the archer is mimicked by the standard
deviation. The more experienced, the closer the arrows will be to the centre, i.e. the
lower the standard deviation.
R-statement rnorm generates normally distributed numbers; we need 20 of them, arranged as a matrix with 2 columns.
shots1 <- matrix(ncol=2, data=rnorm(n=20,sd=0.2))
shots2 <- matrix(ncol=2, data=rnorm(n=20,sd=0.5))
The shots are added to the plot as points, colored darkblue (experienced archer) and
darkgreen (beginners level). Note that we choose a 50% enlarged point size (cex), and
we choose a circular shaped point (pch=16)
points(shots1,col="darkblue",pch=16,cex=1.5)
points(shots2,col="darkgreen",pch=16,cex=1.5)
Finally, we add a legend, explaining who has done the shooting:
legend("topright",legend=c("A","B"),pch=16,
col=c("darkblue","darkgreen"),pt.cex=1.5)

31

Karline Soetaert

A
B

1
2
3
4
5
6
7

8
9
10

Figure 2: Figure with several low-level objects - see text for R-code
Note that the legend text and the colors are inputted as a vector of strings, using the
c() function (e.g. c("A", "B")).

5.2. X-Y plots; conditional plotting

As a more sophisticated demonstration of the use of symbols in R-graphs, we work on a biological example, from the R data set called Orange. This data set contains the circumference
(in mm, at breast height) measured at different ages for five orange trees. We start by looking
at the data (only part is displayed).
>head(Orange)

1
2
3
4
5
6

Tree age circumference

1 118
30
1 484
58
1 664
87
1 1004
115
1 1231
120
1 1372
142

>tail(Orange)

30
31
32
33
34
35

Tree
5
5
5
5
5
5

age circumference
484
49
664
81
1004
125
1231
142
1372
174
1582
177

32

Using R for scientific computing

Orange tree growth

200

150

100

50

circumference, mm

500

1000

1500

age, days

Figure 3: Simple plot of the orange dataset - see text for R-code
and make a rough plot of circumference versus age:
>plot(Orange$age, Orange$circumference,xlab="age, days",
+
ylab="circumference, mm", main= "Orange tree growth")
(as Orange is a dataframe, columns can be addressed by their names, Orange$age and
Orange$circumference).
The output (figure) shows that there is a lot of scatter, which is due to the fact that the five
trees did not grow at the same rate.
It is instructive to plot the relationship between circumference and age differently for each
tree. In R, this is simple: we can make some graphical parameters (symbol types, colors,
size,...) conditional to certain factors.
Factors play a very important part in the statistical applications of R; for our application, it
suffices to know that the factors are integers, starting from 1.
In the R-statement below, we simply use different symbols (pch) and colors (col) for each
tree: pch=(15:20)[Orange$Tree] means that, depending on the value of Orange$Tree (i.e.
the tree number), the symbol (pch) will take on the value 15 (tree=1), 16 (tree=2),... 20
(tree=5). col=(1:5) [Orange$Tree] does the same for the point color. The final statement
adds a legend, positioned at the bottom, right.
>plot(Orange$age, Orange$circumference,xlab="age,
+
days",ylab="circumference, mm", main= "Orange tree growth",
+
pch=(15:20)[Orange$Tree],col=(1:5) [Orange$Tree],cex=1.3)
>legend("bottomright",pch=15:20,col=1:5,legend=1:5)
The output shows that tree number 5 grows fastest, tree number 1 is slowest growing. (note:
it is also instructive to run the examples in the Orange help file. )

5.3. Zooplankton growth rates

Zoogrowth from package marelacTeaching is a literature data set, compiled by Hansen et al.

33

Karline Soetaert

Orange tree growth

200

150

100

circumference, mm

50

500

1000

1
2
3
4
5

1500

age,
days

(1997) with measurements of maximal growth rates of zooplankton organisms as a function

of body volume.
Run the example for this data set (you will need to load package marelacTeaching first):
> require(marelacTeaching)
> example(Zoogrowth)

5.4. Images and contour plots

R has some very powerful functions to create images and add contours. For example, the
data set Bathymetry from the marelac package can be used to generate the bathymetry (and
hypsometry) of the world oceans (and land):
>require(marelac)
>image(Bathymetry$x,Bathymetry$y,Bathymetry$z,col=femmecol(100),
+
asp=TRUE,xlab="",ylab="")
>contour(Bathymetry$x,Bathymetry$y,Bathymetry$z,add=TRUE)
Note the use of asp=TRUE, which maintains the aspect ratio.

5.5. Plotting a mathematical function

Plot curves for mathematical functions are quickly generated with R-command curve:
>curve(sin(3*pi*x))
>curve(sin(3*pi*x),from=0,to=2,col="blue",
+
xlab="x",ylab="f(x)",main="curve")
>curve(cos(3*pi*x),add=TRUE,col="red",lty=2)
>abline(h=0,lty=2)
>legend("bottomleft",c("sin","cos"),text.col=c("blue","red"),lty=1:2)
The first command will plot the curve for y = sin(3 x), using the default settings (Fig.
left), while the next commands afirst draw a graph of y = sin(3 x), in blue (col), and for

34

3000

3000

50

5000

4000

2000

0
100

1000

00

00

00

1000

4000

50

00

50

1000

30

5000

2000

00

4000 3000

3000

150

100

4000
4000

1000

3000

0
4000

3000

4000

0 2000

500

00
0

0
200

00 4000
50

1000

0
10

4000

0
400

1000

1000
500

500

2000

0
00
4

50

1000

100

150

Using R for scientific computing

150

100

50

50

100

150

0.0
1.0

0.5

sin(3 * pi * x)

0.5

1.0

Figure 4: Image plot of ocean bathymetry - see text for R-code

0.0

0.2

0.4

0.6

0.8

1.0

Figure 5: Plotting a mathematical function - see text for R-code

Karline Soetaert

35

x values ranging between 0 and 2 (from, to), adding a main title (main) and x- and y-axis
labels (xlab, ylab) (1st sentence).
The 2nd R-sentence adds the function y = cos(3 x), as a red (col) dashed line (lty). Note
the use of parameter add=TRUE, as by default curve creates a new plot.
The final statements adds the x-axis, i.e. a horizontal, dashed (lty=2), line (abline) at y=0
and a legend.

5.6. Multiple figures

There are several ways in which to arrange multiple figures on a plot.
1. The simplest is by specifying the number of figures on a row (mfrow) and on a column
(mfcol):
> par(mfrow=c(3,2))
will arrange the next plots in 3 rows, 2 columns. Graphs will be plotted row-wise.
> par(mfcol=c(3,2))
will arrange the plots in 3 columns, 2 rows, in a columnwise sequence. Note that both
mfrow and mfcol must be inputted as a vector. Try:
> par(mfrow=c(2,2))
> for ( i in 1:4) curve(sin(i*pi*x),0,1,main=i)
2. R-function layout allows much more complex plot arrangements.

5.7. Exercises
Simple curves
Create a script file which draws a curve of the function y = x3 sin2 (3x) in the interval
[-2, 2].
Make a curve of the function y = 1/cos(1 + x2 ) in the interval [-5,5].

Human population growth

The human population (N, millions of people) at a certain time t, can be described as a
function of time (t), the initial population density at t=t0 (Nt0), the carrying capacity K
and the rate of increase a by the following equation 5 :
N (t) =
5

K
1+

t0
a(tt0)
[ KN
Nt0 ]e

This is the solution of a so-called logistic differential equation (Verhulst 1838)

36

Using R for scientific computing

For the US, the population density in 1900 (N0) was 76.1 million; the population growth can
be described with parameter values: a=0.02 yr1 , K = 500 million of people.
Actual population values are:
1900 1910 1920 1930 1940 1950 1960 1970 1980
76.1 92.4 106.5 123.1 132.6 152.3 180.7 204.9 226.5
Tasks:
1. Plot the population density curve as a thick line, using the US parameter values.
2. Add the measured population values as points. Finish the graph with titles, labels etc...

Toxic ammonia
Ammonia nitrogen is present in two forms: the ammonium ion (N H4+ ) and unionized ammonia (N H3 ). As ammonia can be toxic at sufficiently high levels, it is often desirable to know
its concentration.
The relative importance of ammonia, (the contribution of ammonia to total ammonia nitrogen,
N H3 /(N H3 + N H4+ )) is a function of the proton concentration [H + ] and a parameter KN,
the so-called stoichiometric equilibrium constant:
p[N H3 ] =

KN
KN + [H + ]

Tasks:
Plot the relative fraction of toxic ammonia to the total ammonia concentration as a
function of pH, where pH = log10 ([H + ]) and for a temperature of 30 C. Use a range
of pH from 4 to 9.
The value of KN is 81010 at a temperature of 30 C.
Add to this plot the relative fraction of ammonia at 0 C; the value of KN at that
temperature is 8 1011 mol kg 1 .

The iris data set

A famous data set that is part of R is the iris data set (Fisher, 1936), which we will explore
next.
It gives measurements, in centimeters for sepal length and width and petal length and width,
respectively, for 50 flowers of the species Iris setosa, Iris versicolor and Iris virginica.
Tasks:
Have a look at the data:
What is the class of the data set? why?
What are the dimensions of the data set? (number of rows, columns)

Karline Soetaert

37

Produce a scatter plot of petal length against petal width; produce an informative title
and labels of the two axes.
Repeat the same graph, using different symbol colours for the three species.
Add a legend to the graph. Copy-paste the result to a WORD document. If you do not
have WORD, make a PDF file of the graph.
Create a box-and whisker plot for sepal length where the data values are split into
species groups; use as template the first example in the boxplot help file.
Now produce a similar box-and whisker plot for all four morphological measurements,
arranged in two rows and two columns. First specify the graphical parameter that
arranges the plots two by two.

38

Using R for scientific computing

6. Matrix algebra
Matrix algebra is very simple in R. Practically everything is possible! Here are the most
important R-functions that operate on matrices:
%*% Matrix multiplication
t(A) transpose of A
diag(A) diagonal of A
solve(A) inverse of A
solve(A,B) solving Ax=B for x
eigen(A) eigenvalues and eigenvectors for A
det(A) determinant of A
For instance the following first inverts matrix A (solve(A)), and then multiplies the inverse
with A , giving the unity matrix:
>(A <-matrix(nrow=2,data=c(1,2,3,4)))
[,1] [,2]
[1,]
1
3
[2,]
2
4
>solve(A) %*% A

[1,]
[2,]

[,1] [,2]
1
0
0
1

whilst t(A) will transpose matrix A (interchange rows and columns).

>t(A)

[1,]
[2,]

[,1] [,2]
1
2
3
4

The next set of statements will solve the linear system Ax=B for the unknown vector x:
>B <- c(5,6)
>solve(A,B)
[1] -1

39

Karline Soetaert

Finally, the eigenvalues and eigenvectors of A are estimated using R-function eigen. This
function returns a list that contains both the eigenvalues ($values) and the eigenvectors
($vectors), (the columns).
>eigen(A)
$values
[1] 5.3722813 -0.3722813
$vectors
[,1]
[,2]
[1,] -0.5657675 -0.9093767
[2,] -0.8245648 0.4159736

6.1. Exercises
Matrix algebra exercise 1
Use R-function matrix to create the matrices called A and B:

1 2 3
1 4 7

A = 6 4 1 , B = 2 5 8
2 1 1

3 6 9

Take the inverse of A and the transpose of A.

Multiply A with B.

Estimate the eigenvalues and eigenvectors of A.

For a matrix A, x is an eigenvector, and the eigenvalue of a matrix A, if A x = x.
Test it!

Matrix algebra exercise 2

Create a matrix, called P:

0.0043 0.1132

0.9775 0.9111
0
0

0
0.0736 0.9534
0

0
0
0.0452 0.9804

What is the value of the largest eigenvalue (the so-called dominant eigenvalue) and the
corresponding eigenvector?.

40

Using R for scientific computing

Create a new matrix, T, which equals P, except for the first row, where the elements
are 0.
Now estimate N= (I-T)-1, where I is the identity matrix 6 .

System of linear equations

Solve the following system of linear equations for the unknown xi:
3x1 + 4x2 + 5x3 = 0
6x1 + 2x2 + 7x3 = 5
7x1 + x2
= 6
You will first need to rewrite this problem in the form: Ax=B, where A contains the
coefficients, x the unknowns, and B the right-hand side values. Then you solve the
system, using R-function solve
Check the results (i.e. is Ax = B ?)

6
Note: this is a stage-model of a killer whale (Caswell 2001). The eigenvalue-eigenvectors estimate the rate
of increase and stable age distribution, the matrix N contains the mean time spent in each stage.

Karline Soetaert

41

7. Roots of functions
7.1. Roots of a simple function
Suppose we want to solve the following problem: cos(x) = 2 x for x.

Mathematically, we seek the root of the function y = cos(x) 2 x, this is the value of x for
which y = 0.
As the function is quite complex, it is not possible to find an exact solution (an explicit
expression) for this root.
It is always a good idea to plot the equation (1st line), and add the x-axis (2nd line).
curve(cos(x)-2*x,-10,10)
abline(h=0,lty=2)
This figure shows that there indeed exists a value x, for which y = 0.
Now R-function uniroot can be used to locate this value.
Functions that seek a root from a nonlinear equation generally work iteratively, i.e. they
move closer and closer to the root in successive steps (iterations).
It is usually not feasible to find this root exactly, so it is approximated, i.e. up to a certain
accuracy (tol, a very small number) 7 .
For the method to work, there should be at least one root in the interval.
The statement below solves for the root; it returns several values, as a list.
>(rr<-uniroot(f = function(x) cos(x)-2*x, interval=c(-10,10)))
$root
[1] 0.4501686
$f.root
[1] 3.655945e-05
$iter
[1] 5
$estim.prec
[1] 6.103516e-05
The most important value is the root itself ($root), which is 0.45103686;
the function value at the root was 3.66e-5, the function performed 5 iterations.
In this example, the function was simple enough to include it in the call to uniroot.
The next chapter gives a more complex example from aquatic chemistry, where the equation
to solve is significantly more complex.
Finally, we add the root to the figure:

7
More specifically: the root of y = cos(x) 2 x is the value x for which |cos(x)-2*x | < tol or for which
successive changes of x are < tol

42

20

10

cos(x) 2 * x

10

20

Using R for scientific computing

10

10

Figure 6: Function drawn with curve and the root of the function plotted - see text for R-code
points(rr$root,0,pch=16,cex=2)

7.2. Root of a complex function: solving for the pH

In aquatic systems, the buffering capacity of dissolved inorganic carbon (DIC) species: carbon
dioxide (CO2 ), bicarbonate ions (HCO3 ) and carbonate ions (CO32 ) keep pH in a narrow
range. The buffering capacity is measured by total alkalinity, TA, (defined below).
If alkalinity and inorganic carbon concentration (DIC) are known, it is possible to calculate pH
(Zeebe and Wolf-Gladrow 2003), by solving the following equations for the unknown proton
concentration [H + ] 8 .
[HCO
3]=
[CO2
3 ]=

KC1 [H + ]
DIC
[H + ] [H + ] + KC1 [H + ] + KC1 KC2
[H + ]

[H + ]

KC1 KC2
DIC
+ KC1 [H + ] + KC1 KC2

+
T A = 2[CO2
3 ] + [HCO3 ] [H ]

Here is how to solve for the proton concentration [H + ] (or the pH value) in R .
The trick is to estimate alkalinity based on a guess of proton concentration, using equation (3)
and compare that with the measured alkalinity value. If both are equal within the tolerance
level, the proton concentration has been found.
In the implementation below, the dissociation constants for carbonate (kc1, kc2) and at
salinity 0, temperature 20, and pressure 0 are calculated in R s package seacarb, which has
to be loaded first (require).
We then define a function whose root has to be solved (pHfunction). In this function we
estimate total alkalinity, based on the guess of pH, the dissociation constants (kc1,kc2) and
8
in practice, it is possible to merge these 3 equations such that only one equation is obtained, but this is
neither didactically clearer nor computationally more efficient

Karline Soetaert

43

the DIC concentration. The difference of this calculated alkalinity (EstimatedAlk) with the
true alkalinity is then returned; if pH is correctly estimated, then true and estimated alkalinity
will be equal, and the difference will be zero. So, to find the pH, we need to find the root of
this function.
Note that the conversion from pH to [H + ] gives the proton concentration in molkg 1 . As
the concentrations of the other substances are in molkg 1 , we convert using a factor 106 .
We restrict the region of the pH root in between 0 and 12 (which is more than large enough),
and we set the tolerance (tol) to a very small number to increase precision.
require(seacarb)
kc1 <- K1(S=0,T=20,P=0)
kc2 <- K2(S=0,T=20,P=0)
pHfunction
{
H
<HCO3 <CO3 <-

# Carbonate k1
# Carbonate k2

<- function(pH, kc1,kc2, DIC, Alkalinity )

10^(-pH)
H*kc1 /(H*kc1 + H*H + kc1*kc2)*DIC
kc1*kc2 /(H*kc1 + H*H + kc1*kc2)*DIC

EstimatedAlk <- -H *1.e6 + HCO3 + 2*CO3

return ( EstimatedAlk - Alkalinity)
}
>uniroot(pHfunction, interval=c(0, 12), tol=1.e-20, kc1=kc1, kc2=kc2,
+
DIC=2100, Alkalinity=2200)
$root
[1] 8.295545
$f.root
[1] 0
attr(,"unit")
[1] "mol/kg-soln"
attr(,"pH scale")
[1] "total scale"
$iter
[1] 10
$estim.prec
[1] 1.047918e-07

7.3. Exercises
Simple functions

44

Using R for scientific computing

Find the root of the equation ex = 4x2 in the interval [0,1].
First draw the function curve.
Solve the equations 1000 = y (3 + x) (1 + y)4 for y and with x varying over the range
from 1 to 100. Plot the root as a function of x.
Tip: first make a sequence of x-values, then loop over each x value, each time estimating the root and putting it in a vector.

Chemistry: pCO2 rises increase acidity

pH can also be estimated based on the measured alkalinity and pCO2 , the partial pressure
of CO2 . To solve this equation, it is simplest to use another (equivalent) way to write the
relationships between the DIC species:
[HCO3 ] = KC1
[CO32 ] = KC2

[CO2 ]
[H + ]

[HCO3 ]
[H + ]

pCO2 relates to [CO2 ] through Henrys constant, Kh, which can also be estimated as a function
of salinity, temperature and pressure, using R-package seacarb:
pCO2 =

[CO2 ]
Kh

Estimate the pH at equilibrium with alkalinity 2300 molkg 1 and the current pCO2
of 360 ppm.
Use package seacarb to estimate the dissociation constants and Henrys constants at
temperature 20 C, salinity 0, and pressure 0. (A: pH=8.19)
The Intergovernmental Panel on Climate Change predicts for 2100 an atmospheric CO2
concentration ranging between 490 and 1250 ppmv, depending on the socio-economic
scenario (IPCC, 2007). These increases of pCO2 make the water more acid. Make a
plot of pH as a function of these increased atmospheric pCO2 levels. (Assume that the
pCO2 of the ocean is at equilibrium with the atmospheric pCO2 ). What is the maximal
drop of pH ? (A: at pCO2 of 1250 ppmv, pH=7.68).

45

Karline Soetaert

interpolation,smoothing

spline
approx
smooth.spline
2

10

Figure 7: smoothing and interpolation - see text for R-code

8. Interpolation, smoothing
Interpolating and smoothing in R can be done in several ways:
approx linearly interpolates through points
spline uses spline interpolation, which is smoother
smooth.spline smoothens data sets; this means that it does not connect the original
points.
The use of these functions is exemplified in the following script and corresponding output:
>x <- 1:10
>y <- c(9,8,6,7,5,8,9,6,3,5)
>plot(x,y,pch=16,cex=2,main="interpolation,smoothing")
>lines (spline(x,y, n=100),lty=1)
>points(approx(x,y, xout=seq(1,10,0.1)),pch=1)
>lines (smooth.spline(x,y),lty=2)
>legend("bottomleft",lty=c(1,NA,2),pch=c(NA,1,NA),
+
legend=c("spline","approx","smooth.spline"))

8.1. Curve fitting

R also has several curve fitting procedures. Depending on whether the function to be fitted
is linear, or non-linear, you may use:
lm and glm for fitting linear models and generalised linear models
nls, nlm, optim, constrOptim for nonlinear models.

46

Using R for scientific computing

As an example, we now fit the US population density values, at 10-year intervals, with the
K
logistic growth model (see previous chapter). The model was: N (t) =
,
KNt0
a(tt )
1+[

and the data:

Nt0

]e

1900 1910 1920 1930 1940 1950 1960 1970 1980

76.1 92.4 106.5 123.1 132.6 152.3 180.7 204.9 226.5
We start by inputting the data.
>year<- seq(1900,1980,by=10)
>pop <- c(76.1,92.4,106.5,123.1,132.6,152.3,180.7,204.9,226.5)
The simplest method for non-linear curve fitting is by using R-function nls.
This functions requires as input the formula (y ~ f (x,parameters)) and starting values of
the parameters.
In the example, y are the population values, f is the logistic growth formulation.
As starting conditions, we use: K=500,N0=76.1,a=0.02.
>fit <- nls(pop~K/(1+(K-N0)/N0*exp(-a*(year-1900))),
+
start=list(K=500,N0=76.1,a=0.02))
We end the fitting by printing a summary of the fitting parameters, which shows the estimates
of the parameters and their standard errors. Clearly, it is not possible to obtain reliable
estimates of the value of K based on the data.
>summary(fit)
Formula: pop ~ K/(1 + (K - N0)/N0 * exp(-a * (year - 1900)))
Parameters:
Estimate Std. Error t value Pr(>|t|)
K 1.008e+03 8.932e+02
1.129 0.30210
N0 7.866e+01 2.531e+00 31.084 7.36e-08 ***
a 1.550e-02 2.505e-03
6.188 0.00082 ***
--Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Residual standard error: 3.685 on 6 degrees of freedom
Number of iterations to convergence: 6
Achieved convergence tolerance: 4.301e-06
The values of the coefficients themselves are retrieved using R-function coef.
>(cc<-coef(fit))
K
N0
a
1.008227e+03 7.866365e+01 1.550343e-02

Karline Soetaert

47

8.2. Exercises
Smoothing
An anemometer measures wind-velocity at three hourly intervals. On a certain day, these
velocities are: 5,6,7,9,4,6,3,7,9 at time 0, 3, ... 24 oclock respectively. In order to estimate
air-sea exchange, we need hourly measures.
Tasks:
Interpolate the three-hourly measurements to hourly measurements.
Make a plot of the interpolated values

Fitting
Primary production is measured by 14 C incubations from phytoplankton samples, at different
light intensities.
The data are:
>ll <- c(0.,1,10,20,40,80,120,160,300,480,700)
>pp <- c(0.,1,3,4,6,8,10,11,10,9,8)
Fit the resulting production estimates (pp), as a function of light intensity (ll) with the
3-parameter Eilers-Peeters equation. The primary production is calculated as:
pp = p max

2 (1 + ) I/Iopt
(I/Iopt)2 + 2 I/Iopt + 1

where I is light and pmax, and Iopt are parameters.

First plot primary production (pp) versus light (ll). Use large symbols.
Then use R-function nls to fit the model to the data
Add the best-fit line to the graph. (note: use coef to retrieve the best parameter
values).

48

Using R for scientific computing

9. Differential equations
Differential equations express the rate of change of a constituent (C) along one or more
dimensions, usually time and/or space.
Consider the following set of two differential equations:
dA
dt
dB
dt

= r (x A) k A B

= r (y B) + k A B

A and B are called differential variables (or state variables),

dA
dt

is the derivative (or the rate of change),

r,x,y and k are parameters (constants).

To solve sets of differential equations in R we define a function (here called model), and which
has as input the time (t), the values of the state variables (state) and the values of the
parameters (pars). This function simply calculates the rate of change of the state variables
(dA and dB) and returns those as a list.
The R-statement with (as.list (c(state,pars)), ensures that the state variables and
parameters can be addressed by their names.
model <- function(t,state,pars)
{
with (as.list(c(state,pars)),
{
dA <- r*(x-A)-k*A*B
dB <- r*(y-B)+k*A*B
return (list(c(dA,dB)))
}
)
}
Before we can solve this model, we
generate a sequence of time values at which we want output (times),
assign initial conditions to the state variables (state) and
give values to the parameters (parms):
times <- seq(0,300,1)
state <- c(A=1,B=1)
parms <- c(x =1, y = 0.1, k = 0.05, r = 0.05)
integration routine ode; which can be
The model can now be solved. To do so, we use R Ss
found in R -package deSolve. This package is loaded first.
require(deSolve)

Karline Soetaert

49

At each time t, ode will call function model, with the current values of the state variables
and the parameter values.
The output is stored in a data.frame, called out.
out

<- as.data.frame(ode(state,times,model,parms))

All we need to do now is to plot the model output. Before we do so, we have a look at
data.frame out:
>head(out)

1
2
3
4
5
6

time
0
1
2
3
4
5

A
1.0000000
0.9523189
0.9090687
0.8699226
0.8345728
0.8027203

B
1.000000
1.003787
1.005285
1.004715
1.002285
0.998201

The data are arranged in three columns: first the time, then the concentrations of A and B.
As out is a data frame we can extract the data using their names (out$time, out$A, out$B).
Before plotting the model output, the range of concentrations of substances A and B is
estimated; this is used to set the limits of the y-axis (ylim).
R -function plot creates a new plot; lines adds a line to this plot; lty selects a line type;
lwd=2 makes the lines twice as thick as the default. Finally a legend is added.
ylim
<- range(c(out$A,out$B))
plot(out$time,out$A,xlab="time",ylab="concentration",
lwd=2,type="l",ylim=ylim,main="model")
lines(out$time,out$B,lwd=2,lty=2)
legend("topright",legend=c("A","B"),lwd=2,lty=c(1,2))

9.1. Exercises
Lotka-volterra model
Write a script file that solves the following system of ODEs 9 :
dx
dt
dy
dt

= a x (1

x
K)

bxy

=gbxyey

for initial values x=300,y=10 and parameter values: a=0.05, K=500, b=0.0002, g=0.8,
e=0.03
9

The Lotka-Volterra models are a famous type of models that either describe predator-prey interactions
or competitive interactions between two species. A.J. Lotka and V. Volterra formulated the original model in
the 1920s almost simultaneously ((Lotka 1925), (Volterra 1926)).

50

Using R for scientific computing

1.0

model

0.7
0.6
0.4

0.5

concentration

0.8

0.9

A
B

50

100

150

200

250

300

time

Figure 8: ode model - see text for R-code

Make three plots, one for x and one for y as a function of time, and one plot expressing
y as a function of x (this is called a phase-plane plot). Arrange these plots in 2 rows
and 2 columns.
Now run the model with other initial values (x=200, y=50); add the (x,y) trajectories
to the phase-plane plot

Butterfly
The Lorenz equations (Lorenz 1963) were the first chaotic system of differential equations to
be discovered. They are three differential equations that were derived to represent idealized
atmosphere.
behavior of the earthSs
dx
dt
dy
dt
dz
dt

= 83 x + y z

= 10 (y z)

= x y + 28y z

It takes about 10 lines of R-code to generate the solutions and plot them.
Function scatterplot3d from the package scatterplot3d generates 3-D scatterplots.
Can you recreate the following butterfly ? Use as initial conditions x=y=z=1; create
output for a time sequence ranging from 0 to 100, and with a time step of 0.005.

51

Karline Soetaert

0
10

out$z

10

20

30

Lorenz butterfly

20

20

10
0

30

10
20
0

10

20

30

40

50

out$x

10. Finally
10.1. The questions
These lecture notes have been generated with LaTeX and making use of R-package Sweave
(Leisch 2002), which allows to merge LaTeX with R-code.
If you do not like the layout a PDF version (ScientificComputing.pdf) made with WORD
(Microsoft) can be found in the /inst/lecture subdirectory of package marelacTeaching.

10.2. The answers

The answers to the questions in this course are present as an R-vignette in package marelacTeaching. From within R, type:
> vignette("Answers")
Or, you can find the file Answers.pdfin the /inst/doc subdirectory of package marelacTeaching.

References
Caswell H (2001). Matrix population models: construction, analysis, and interpretation. Sinauer, Sunderland, second edition edition.
Hurlbert SH (1971). The nonconcept of species diversity: critique and alternative parameters. Ecology, 52, 577586.
Kuhnert P, Venables W (2005). An introduction to R: software for statistical modelling &
computing. URL www.r-project.org.

52

Using R for scientific computing

Leisch F (2002). Sweave: Dynamic Generation of Statistical Reports Using Literate Data
Analysis. In WH
ardle, BR
onz (eds.), Compstat 2002 - Proceedings in Computational
Statistics, pp. 575580. Physica Verlag, Heidelberg. ISBN 3-7908-1517-9, URL http:
//www.stat.uni-muenchen.de/~leisch/Sweave.
Ligges U, Machler M (2003). Scatterplot3d - an R Package for Visualizing Multivariate
Data. Journal of Statistical Software, 8(11), 120.
Lorenz E (1963). Deterministic non-periodic flows. J. Atmos. Sci, 20, 130141.
Lotka AJ (1925). Elements of Physical Biology. Williams & Wilkins Co., Baltimore.
Millero F, Poisson A (1981). International one-atmosphere equation of state for seawater.
Deep-Sea Research, 28(6), 625629.
Proye A, Gattuso JP, Epitalon JM, Gentili B, Orr J, Soetaert K (2007). seacarb: Calculates
parameters of the seawater carbonate system. R package version 1.2.3, URL http://www.
obs-vlfr.fr/~gattuso/seacarb.php.
R Development Core Team (2008). R: A Language and Environment for Statistical Computing.
R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, URL http:
//www.R-project.org.
Sanders H (1968). Marine benthic diversity: a comparative study. Amercian Naturalist,
102, 243282.
Soetaert K (2009). rootSolve: Nonlinear root finding, equilibrium and steady-state analysis of
ordinary differential equations. R package version 1.4.
Soetaert K, Heip C, Vincx M (1991). Diversity of nematode assemblages along a mediterranean deep-sea transect. Marine Ecology Progress Series, 75, 275282.
Soetaert K, Herman PMJ (2009). A Practical Guide to Ecological Modelling. Using R as a
Simulation Platform. Springer. ISBN 978-1-4020-8623-6.
Soetaert K, Petzoldt T, Meysman F (2009a). marelac: Constants, conversion factors, utilities
for the MArine, Riverine, Estuarine, LAcustrine and Coastal sciences. R package version
1.4.
Soetaert K, Petzoldt T, Meysman F (2009b). marelacTeaching: Datasets and tutorials for use
in the MArine, Riverine, Estuarine, LAcustrine and Coastal sciences. R package version
1.0.
Soetaert K, Petzoldt T, Setzer RW (2009c). deSolve: General solvers for initial value problems of ordinary differential equations (ODE), partial differential equations (PDE) and
differential algebraic equations (DAE). R package version 1.3.
Verhulst PF (1838). Notice sur la loi que la population pursuit dans son accroissement.
Correspondance mathematique et physique, 10, 113121.
Volterra V (1926). Variazioni e fluttuazioni del numero dindividui in specie animali conviventi. Mem. R. Accad. Naz. dei Lincei. Ser. VI, 2, 31113.

Karline Soetaert

53

Zeebe R, Wolf-Gladrow D (2003). CO2 in Seawater: Equilibrium, kinetics, isotopes. Elsevier.

Affiliation:
Karline Soetaert
Centre for Estuarine and Marine Ecology (CEME)
Netherlands Institute of Ecology (NIOO)
4401 NT Yerseke, Netherlands E-mail: [email protected]
URL: http://www.nioo.knaw.nl/users/ksoetaert