Introduction to Matlab programming

Patrick Winistörfer
Study Center Gerzensee
Fabio Canova
ICREA-UPF, CREI, AMeN and CEPR

January 22, 2008

Contents
1 The Command Window 3
1.1 Interacting with the Matlab Command Window . . . . . . . . . . 3
1.2 Command Line Editor . . . . . . . . . . . . . . . . . . . . . . . . 3

2 Matrix Construction and Manipulation 4

2.1 Building a Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2 The Help Facility . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Exporting and Importing Data . . . . . . . . . . . . . . . . . . . 6
2.3.1 Exporting Data Files . . . . . . . . . . . . . . . . . . . . . 6
2.3.2 Importing Data Files . . . . . . . . . . . . . . . . . . . . . 7

3 Basic Manipulations 8
3.1 Using Partitions of Matrices . . . . . . . . . . . . . . . . . . . . . 8
3.2 Matrix Operations . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.1 Making Vectors from Matrices and Reverse . . . . . . . . 9
3.2.2 Transposition of a Matrix . . . . . . . . . . . . . . . . . . 9
3.2.3 Basic Operators . . . . . . . . . . . . . . . . . . . . . . . 9
3.2.4 Array Operations . . . . . . . . . . . . . . . . . . . . . . . 10
3.2.5 Relational Operations . . . . . . . . . . . . . . . . . . . . 10
3.2.6 Logical Operations . . . . . . . . . . . . . . . . . . . . . . 10
3.3 More Built-in Functions . . . . . . . . . . . . . . . . . . . . . . . 11
3.3.1 Display Text or Array . . . . . . . . . . . . . . . . . . . . 11
3.3.2 Sorting a Matrix . . . . . . . . . . . . . . . . . . . . . . . 11
3.3.3 Sizes of Each Dimension of an Array . . . . . . . . . . . . 12
3.3.4 Sum of Elements of a Matrix . . . . . . . . . . . . . . . . 12
3.3.5 Smallest (Largest) Elements of an Array . . . . . . . . . . 12
3.3.6 Inverse of a Matrix . . . . . . . . . . . . . . . . . . . . . . 13
3.3.7 Eigenvectors and Eigenvalues of a Matrix . . . . . . . . . 13
3.4 Special matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4.1 Elementary Matrices . . . . . . . . . . . . . . . . . . . . . 14
3.4.2 Special Matrices . . . . . . . . . . . . . . . . . . . . . . . 14
3.4.3 Manipulating Matrices . . . . . . . . . . . . . . . . . . . . 14

4 Mathematical Functions 15
4.1 Trigonometric Functions . . . . . . . . . . . . . . . . . . . . . . . 15
4.2 Exponential Functions . . . . . . . . . . . . . . . . . . . . . . . . 16
4.3 Complex Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 16
4.4 Numeric Functions . . . . . . . . . . . . . . . . . . . . . . . . . . 16

5 Data Analysis 17
5.1 Using Column-Oriented Analysis . . . . . . . . . . . . . . . . . . 17
5.2 Missing values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
5.3 Polynomials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
5.4 Functions functions . . . . . . . . . . . . . . . . . . . . . . . . . . 20

1
6 Graphics 23
6.1 Histogramm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
6.2 Plotting Series versus their Index . . . . . . . . . . . . . . . . . . 23
6.3 Scatterplot (2-D) . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
6.4 Adding Titles and Labels . . . . . . . . . . . . . . . . . . . . . . 24
6.5 Plotting Commands . . . . . . . . . . . . . . . . . . . . . . . . . 25
6.6 3-D Graphics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

7 Scripts and Functions 29

7.1 Script m-files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
7.2 Function m-files . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29

8 Controlling the Flow 31

8.1 The FOR Loop . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
8.2 The WHILE Loop . . . . . . . . . . . . . . . . . . . . . . . . . . 32
8.3 The IF Statement . . . . . . . . . . . . . . . . . . . . . . . . . . 32
8.4 How to Break a FOR or WHILE Loop . . . . . . . . . . . . . . . 33

9 Random Numbers 33

10 Some Examples 33
10.1 loaddata.m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
10.2 example var.m . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
10.3 example gmm.m . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
10.4 draw rn.m . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2
1 The Command Window
When we invoke Matlab, the command window is created and made the active
window. The command window is the interface through which we communicate
with the Matlab interpreter. The interpreter displays its prompt (À) indicating
that it is ready to accept instructions. We can now type in Matlab commands.
In what follows, typeset letters denote Matlab code.

1.1 Interacting with the Matlab Command Window

After typing a command you execute it by pressing the enter key. Issuing the
command clear x clears the value of variable x from the memory and clear
clears all values that have been created during a Matlab session.
In general, if you type
>> A
Matlab does the following:
1. It verifies if A is a variable in the Matlab workspace.
2. If not, it verifies if A is a built-in-function.
3. If not, it verifies if a M-file named A.m exists in the current directory.
4. If not it verifies if A.m exists anywhere on the Matlab search path, by
searching the path in the order in which it is specified.
Finally, if Matlab can not find A.m anywhere on the Matlab search path, an
error message will appear in the command window.
We can display and change the current Matlab search path: Select Set Path
from the File menu.
When Matlab displays numerical results it displays real numbers with four digits
to the right of the decimal point. If you want to have more than four digits
displayed issue the command format long. To go back to displaying four digits
after the decimal point (the default setting), type format short.
Note also that a ’;’ (semicolon) at the end of a line tells Matlab not to display
the results from the command it executes. Text following ’%’ is interpreted as
a comment (plain text) and is not executed.

1.2 Command Line Editor

To simplify the process of entering commands to the Matlab interpreter, we
can use the arrow keys on the keypad to edit mistyped commands and to recall
previous command lines. For example, suppose we execute the command line:
>> log(sqt(atan2(3,4)))
and misspell ’sqrt’. Matlab responds with the error message:

3
 ??? Undefined function or variable ’sqt’
Instead of retyping the entire line, simply press the Up-arrow key. The mis-
spelled command is redisplayed, and we can move the cursor over the appro-
priate place to insert the missing ’r’ by using the Left-arrow key or the mouse.
After pressing Return, the command returns the appropriate answer.
>> log(sqrt(atan2(3,4)))
ans =
-0.2204

2 Matrix Construction and Manipulation

In Matlab each variable is a matrix. Since a matrix contains m rows and n
columns, it is said to be of dimension m-by-n. An m-by-1 or 1 -by-n matrix is
called a vector. A scalar, finally, is a 1 -by-1 matrix.
You can give a matrix whatever name you want. Numbers may be included in
the names. Matlab is sensitive with respect to the upper or lower case. If a
variable has previously been assigned a value, the new value overwrites the old
one.

2.1 Building a Matrix

There are several ways to build a matrix:
One way to do it is to declare a matrix as if you wrote it by hand
>> A=[1 2 3
4 5 6]
This results in the output
A =
1 2 3
4 5 6
Another way is to separate rows with ’;’
>> A=[1 2 3;4 5 6]

A final way is to do it element by element

>> A(1,1)=1;
>> A(1,2)=2;
>> A(1,3)=3;
>> A(2,1)=4;
>> A(2,2)=5;
>> A(2,3)=6;

4
There are some special matrices that are extremely useful:
The zero (or Null) matrix: zero(m,n) creates a m-by-n matrix of zeros. Thus,
>> B=zeros(3,2)

results in
B =
0 0
0 0
0 0

The ones matrix: zero(m,n) creates a m-by-n matrix of zeros. Thus,

>> C=ones(2,3)
results in
C =
1 1 1
1 1 1
The identity matrix: eye(n) creates a n-by-n matrix with ones along the diag-
onal and zeros everywhere else.
>> D=eye(3)
results in
D =
1 0 0
0 1 0
0 0 1

You can generate a m-by-n matrix of random elements using the command
rand(m,n), for uniformly distributed elements, randn(m,n), for normally dis-
tributed elements. That is, rand will draw numbers in [0; 1] while randn will
draw numbers from a N (0, 1) distribution.
The colon ’:’ is an important symbol in Matlab. The statement
>> x=1:5
generates a row vector containing the numbers from 1 to 5 with unit increments.
We can also use increments different from one:
>> y=0:pi/4:pi

results in
y =
0 0.7854 1.5708 2.3562 3.1416

5
2.2 The Help Facility
The help facility provides online information about Matlab functions, commands
and symbols.
The command
>> help

with no arguments displays a list of directories that contains Matlab related

files.
Typing help topic displays help about that topic if it exists. The command
help elmat provides information on ’Elementary matrices and matrix manipu-
lation’, help elfun for information on ’Elementary math functions’, and help
matfun for information on ’Matrix functions - numerical linear algebra’.

2.3 Exporting and Importing Data

First, we have to indicate the correct current directory in the command toolbar,
choose ”d:\”.

2.3.1 Exporting Data Files

In the following exercise we create four different data files:

1. Clear the workspace by issuing the command clear.

2. Define the following two matrices:

>> x=[0:0.5:50]’
>> y=randn(6,12)
3. Create the Matlab data file result1.mat:

>> save result1

The two matrices x and y are saved in the file result1.mat.
4. Create another Matlab data file result2.mat:

>> save result2 y

Only the matrix y is saved in the file result2.mat.
5. Create the data file result3.txt:

>> save result3.txt y -ascii

6
 Only the matrix y is saved in 8-digit ASCII text format in the file re-
sult3.txt.
6. Create the data file result4.dat:

>> save result4.dat x

Only the matrix x is saved in the file result4.dat.

2.3.2 Importing Data Files

In the next exercise we import the four results we previously saved:
1. Clear the workspace by issuing the command clear.
2. Import the data file result1.mat:

>> load result1

The two matrices x and y are imported in the workspace.
3.
4. Clear again the workspace by issuing the command clear.
5. Import the data file result2.mat:

>> load result2

The matrix y is imported in the workspace.
6. Clear again the workspace by issuing the command clear.
7. Import the data file result3.txt:

>> load result3.txt -ascii

A matrix called result3 is imported in the workspace. This matrix is
exactly the y matrix we created previously.
8. Clear again the workspace by issuing the command clear.
9. Import the data file result4.dat:

>> a=importdata(’result4.dat’)
A matrix called a is imported in the workspace. This matrix is exactly
the x matrix we created previously.

7
3 Basic Manipulations
3.1 Using Partitions of Matrices
You can build matrices out of several submatrices. Suppose you have subma-
trices A to D.

>> A=[1 2; 3 4];

>> B=[5 6 7; 8 9 10];
>> C=[3 4; 5 6];
>> D=[1 2 3; 4 5 6];

µ ¶
A B
In order to build E, which is given by E = , you type:
C D
>> E=[A B; C D]
E =
1 2 5 6 7
3 4 8 9 10
3 4 1 2 3
5 6 4 5 6
One of the most basic operation is to extract some elements of a matrix (called
a partition of matrices). Consider
 matrix  E we defined previously. Let’s now
2 5
isolate the central matrix F =  4 8 . In order to do this we type
4 1
>> F=E(1:3,2:3)
F =
2 5
4 8
4 1
Suppose we just want to select columns 1 and 3, but take all the lines.
>> F=E(:,[1 3])
F =
1 5
3 8
3 1
5 4
where the colon (:) means select all.

8
3.2 Matrix Operations
3.2.1 Making Vectors from Matrices and Reverse
Suppose we want to obtain the vectorialisation of a matrix, that is we want to
obtain vector B from matrix A.
A =
1 2
3 4
5 6
>> B=A(:)
B =
1
2
3
4
5
6
Suppose we have vector B. We want to obtain matrix A from B.
>> A=reshape(B,3,2)
A =
1 2
3 4
5 6

3.2.2 Transposition of a Matrix

A =
1 2
3 4
5 6
>> A’
ans =
1 3 5
2 4 6
Note: The ans variable is created automatically when no output argument is
specified. It can be used in subsequent operations.

3.2.3 Basic Operators

+ Addition
− Subtraction
∗ Multiplication
/ Right division B/A = BA−1
\ Left division A\B = A−1 B
ˆ Exponentiation Aˆ(−1) = A−1

9
3.2.4 Array Operations
To indicate an array (element-by-element) operation, precede a standard oper-
ator with a period (dot). Matlab array operations include multiplication(.*),
division (./) and exponentiation (.ˆ).
Thus, the ”dot product” of x and y is
>> x=[1 2 3];
>> y=[4 5 6];
>> x.*y
ans =
4 10 18

3.2.5 Relational Operations

< > Less (greater) than
<= >= Less (greater) than or equal to
== Equal to
∼= Not equal to

Matlab compares the pairs of corresponding elements; the result is a matrix of

ones and zeros, with one representing ’true’ and zero representing ’false’. E.g.
let’s consider the following matrix G. the command G<=2 finds elements of G
that are less or equal 2. The resulting matrix is T.
>> G=[1 2; 3 4];
>> T=G<=2;
T =
1 1
0 0

3.2.6 Logical Operations

& AND
| OR
∼ NOT

>> C=A&B;

C is a matrix whose elements are ones where A and B have nonzero elements,
and zeros where either has a zero element. A and B must have the same dimen-
sion unless one is a scalar. A scalar can operate with everything.

>> C=A|B;

10
C is a matrix whose elements are ones where A or B have a nonzero element,
and zeros where both have a zero element.
>> B=~A;

B is a matrix whose elements are ones where A has a zero element, and zeros
where A has a nonzero element.

3.3 More Built-in Functions

The type of commands used to build special matrices are called built-in func-
tions. There are a large number of built-in functions in Matlab. Apart from
those mentioned above, the following are particularly useful:

3.3.1 Display Text or Array

disp(A) displays an array, without printing the array name. If A contains a
text string, the string is displayed.
>> disp(’ Corn Oats Hay’)
Corn Oats Hay

3.3.2 Sorting a Matrix

The function sort(A) sorts the elements in ascending order. If A is a matrix,
sort(A) treats the columns of A as vectors, returning sorted columns.

>> A=[1 2; 3 5; 4 3]
A =
1 2
3 5
4 3
>> sort(A)
ans =
1 2
3 3
4 5

11
3.3.3 Sizes of Each Dimension of an Array
The function size(A) returns the sizes of each dimension of matrix A in a
vector. [m,n]=size(A) return the size of matrix A in variables m and n (recall:
in Matlab arrays are defined as m-by-n matrices). length(A) returns the size
of the longest dimension of A.
>> A=[1 2; 3 5; 4 3];
>> [m,n]=size(A)
m =
3
n =
2
>> length(A)
ans =
3

3.3.4 Sum of Elements of a Matrix

If A is a vector, sum(A) returns the sum of the elements. If A is a matrix,
sum(A) treats the columns of A as vectors, returning a row vector of the sums
of each column.

>> A=[1 2; 3 5; 4 3];

>> B=sum(A)
B =
8 10

If A is a vector, cumsum(A) returns a vector containing the cumulative sum of

the elements of A. If A is a matrix, cumsum(A) returns a matrix in the same
size as A containing sums for each column of A.
>> B=cumsum(A)
B =
1 2
4 7
8 10

3.3.5 Smallest (Largest) Elements of an Array

If A is a matrix, min(A) treats the columns of A as vectors, returning a row vec-
tor containing the minimum element from each column. If A is vector, min(A)
returns the smallest element in A. max(A) returns the maximum elements.
>> A=[1 2; 3 5; 4 3];
>> B=min(A)
ans =
1 2

12
3.3.6 Inverse of a Matrix
The function inv(A) returns the inverse of the square matrix A.
>> A=[4 2; 1 3];
>> B=inv(A)
B =
0.3000 -0.2000
-0.1000 0.4000

A frequent misuse of inv arises when solving the system of linear equations
Ax = b. One way to solve this is with x=inv(A)*b. A better way, from both an
execution time and numerical accuracy standpoint, is to use the matrix division
operator x=A\b.

3.3.7 Eigenvectors and Eigenvalues of a Matrix

The n-by-n (quadratic) matrix A can often be decomposed into the form A =
P DP −1 , where D is the matrix of eigenvalues and P is the matrix of eigenvec-
tors. Consider
>> A =
0.9500 0.0500
0.2500 0.7000
>> [P,D]=eig(A)
>> P =
0.7604 -0.1684
0.6495 0.9857
>> D =
0.9927 0
0 0.6573

If you are just interested in the eigenvalues:

>> eig(A)
ans =
0.9927
0.6573

3.4 Special matrices

A collection of functions generates special matrices that arise in linear algebra
and signal processing is next.

13
3.4.1 Elementary Matrices
zeros Matrix of zeros
ones Matrix of ones
eye Identity matrix
rand Matrix of uniformly distributed random numbers
randn Matrix of normally distributed (N (0, 1)) random numbers
meshgrid X and Y arrays for 3-D plots

3.4.2 Special Matrices

compan Companion matrix
hadamard Hadamard matrix
hankel Hankel matrix
kron Kronecker tensor product
toeplitz Toeplitz matrix

3.4.3 Manipulating Matrices

Several functions rotate, flip, reshape, or extract portions of a matrix.

diag Create or extract diagonals

fliplr Flip matrix in the left/right direction
flipud Flip matrix in the up/down direction
reshape Change size
tril Extract lower triangular part
triu Extract upper triangular part
size Size of a matrix (two element vector with row and column dimension)
length Size of a vector

For example to reshape a 3-by-4 matrix A into a 2-by-6 matrix B use the
command:

>> B = reshape(A,2,6)

14
4 Mathematical Functions
A set of elementary functions are applied on an element by element basis to
arrays. For example:

>> A = [1 2 3; 4 5 6];
>> B = fix(pi*A)

(fix(X) rounds the elements of X to the nearest integers towards zero; pi stands
for π) produces:

B =
3 6 9
12 15 18

and

>> C = cos(pi*B)

gives

C =
-1 1 -1
1 -1 1

The following subsections list Matlab’s principal mathematical functions.

4.1 Trigonometric Functions

sin Sine
sinh Hyperbolic sine
cos Cosine
cosh Hyperbolic cosine
tan Tangent
tanh Hyperbolic tangent
sec Secant
sech Hyperbolic secant

15
4.2 Exponential Functions

exp Exponential (e to the power)

log Natural logarithm
log10 Common logarithm
sqrt Square root

4.3 Complex Functions

abs Absolute value

angle Phase angle
conj Complex conjugate
imag Complex imaginary part
real Complex real part

4.4 Numeric Functions

fix Round towards zero
floor Round towards minus infinity
ceil Round towards plus infinity
round Round towards nearest integer
rem Remainder after division
sign Signum function

16
5 Data Analysis
This section presents an introduction to data analysis using Matlab and de-
scribes some elementary statistical tools.

5.1 Using Column-Oriented Analysis

Matrices are, of course, used to hold all data, but this leaves a choice of orien-
tation for multivariate data. By convention, the different variables in a set of
data are put in columns, allowing observations to vary down through the rows.
Therefore, a data set of 50 samples of 13 variables is stored in a matrix size
50-by-13.
We can generate a sample data set (let’s call it B) of 13 variables with 50
observations each by using the command:

>> B = rand(50,13)

and analyze the main properties of the generated variables using the data anal-
ysis capabilities provided by the following functions:

max Largest component

min Smallest component
mean Average or mean value
median Median value
std Standard deviation
sort Sort in ascending order
sum Sum of elements
prod Product of elements
cumsum Cumulative sum of elements
cumprod Cumulative product of elements
trapz Numerical integration using trapezoidal method

For vector arguments, it does not matter whether the vectors are
oriented in a row or column direction. For array arguments, the
functions described above operate in a column-oriented fashion. This
means that if we apply the operator max on an array, the result will be a row
vector containing the maximum values over each column.

5.2 Missing values

The correct handling of NAs (Not Available) is a difficult problem and often
varies in different situations. Matlab, however, is uniform and rigorous in its
treatment of NAs; they propagate naturally to the final result in any calculation.
In Matlab NAs are treated as NaNs (Not a Number). We should remove NaNs

17
from the data before performing statistical computations. The NaNs in a vector
x are located at:

>> i = find(isnan(x));

so

>> x = x(find(~isnan(x));

returns the data in x with the NaNs removed. Two other ways of doing the
same thing are:

>> x = x(~isnan(x));

or

>> x(isnan(x)) = [];

If instead of a vector, the data are in the columns of a matrix, and we are
removing any rows of the matrix with NaNs, use:

>> X(any(isnan(X)’),:) = [];

5.3 Polynomials
Matlab represents polynomials as row vectors containing the coefficients or-
dered by descending powers. For example, the coefficients of the characteristic
equation of the matrix:
 
1 2 3
A= 4 5 6 
7 8 0
can be computed with the command:

>> p = poly(A)

18
which returns

p =
1 -6 -72 -27

This is the Matlab representation of the polynomial:

x3 − 6x2 − 72x − 27
The roots of this equation can be obtained by:

>> r = roots(p);

These roots are, of course, the same as the eigenvalues of matrix A. We can
reassemble them back into the original polynomial with poly:

>> p2 = poly(r);

Now consider the polynomials a(x) = x2 + 2x + 3 and b(x) = 4x2 + 5x + 6. The

product of the polynomials is the convolution (conv) of the coefficients:

>> a = [1 2 3];
>> b = [4 5 6];
>> c = conv(a,b);

We can also use deconvolution (deconv) to divide a(x) back out:

>> [q,r] = deconv(c,a)

gives

q =
4 5 6

r =
0 0 0 0 0

19
Other polynomial functions:

roots Find polynomial roots

poly Construct polynomial with specified roots
polyval Evaluate polynomial
polyvalm Evaluate polynomial with matrix argument
residue Partial-fraction expansion (residues)
polyfit Fit polynomial to data
polyder Differentiate polynomial
conv Multiply polynomials
deconv Divide polynomials

5.4 Functions functions

There is a class of functions in Matlab that doesn’t work with numerical matri-
ces, but with mathematical functions. These function functions include:
(a) Numerical integration.
(b) Nonlinear equations and optimization.
(c) Differential equation solution.

Functions
Matlab represents mathematical functions by function m-files. For example, the
function:
1 1
f (x) = 2 + 2 −6
(x − 0.3) + 0.01 (x − 0.9) + 0.04
is made available to Matlab by creating a m-file called humps.m:

function y = humps(x)
y = 1 ./ ((x-.3).^2 + .01) + 1 ./((x-.9).^2 + .04) - 6;

A graph of the function can be obtained by typing:

>> x = -1:0.01:2;
>> plot(x,humps(x))

(a) Numerical Integration

The area beneath a function, f (x) can be determined by numerically inte-
grating f (x), a process referred to as quadrature. To integrate the function
defined by humps.m from 0 to 1 type:

20
 >> q = quad(’humps’,0,1)

this command returns

q =
29.8583

The two Matlab functions for quadrature are

quad Adaptive Simpson’s rule

quad8 Adaptive Newton Cotes 8 panel rule

Notice that the first argument to quad is a quoted string containing the
name of a function. This is why quad is called a function function - it is
a function that operates on other functions.
(b) Nonlinear Equations and Optimization Functions
The function functions for nonlinear equations and optimization include:
(check, if your matlab version is 6.5 or above the names have
changed!)

fmin Minimum of a function of one variable

fmins Minimum of a multivariate function (unconstrained nonlinear optimization
fzero Zero of a function of one variable

Using again the function defined in humps.m, the location of the minimum
in the region from 0.5 to 1 is computed with fmin:

>> xm = fmin(’humps’,0.5,1)

It returns:

xm =
0.6370

Its value at the minimum is:

>> y = humps(xm)
y =
11.2528

21
Looking at the graph, it is apparent that humps has two zeros. The
location of the zero near x = 0 is:

>> xz1 = fzero(’humps’,0)

xz1 =
-0.1316

The Optimization Toolbox contains several additional functions:

attgoal Multi-objective goal attainment

constr Constrained minimization
fminu Unconstrained minimization
fsolve Nonlinear equation solution
leastsq Nonlinear least squares
minimax Minimax solution
seminf Semi-infinite minimization

22
6 Graphics
Matlab provides a variety of functions for displaying data as planar plot or 3-D
graphs and for annotating these graphs. This section describes some of these
functions and provides examples of some typical applications.
The first important instruction is the clf instruction that clears the graphic
screen.

6.1 Histogramm
The instruction hist(x) draws a 10-bin histogram for the data in vector x. The
command hist(x,c), where c is a vector, draws a histogram using the bins
specified in c. Here is an example:
>> c=-2.9:0.2:2.9;
>> x=randn(5000,1);
>> hist(x,c);

6.2 Plotting Series versus their Index

2-D graphs essentially rely on the instruction plot(x). It plots the data in vector
x versus its index. The command plot([x y]) plots the two column vectors
of the matrix [x y] versus their index (within the same graph). Consider the
following example:

>> x=rand(100,1);
>> y=rand(100,1);
>> clf;
>> plot([x y]);

6.3 Scatterplot (2-D)

The general form of the plot instruction is
plot(x,y,S)
where x and y are vectors or matrices and S is a one, two or three charter string
specifying color, marker symbol, or line style. The command plot(x,y,S) plots
y against x, if x and y are vectors. If x and y are matrices, the columns of y
are plotted versus the columns of x in the same graph. If only x or y is a
matrix, the vector is plotted versus the rows or columns of the matrix. Note
that plot(x,[a b],S) and plot(x,a,S,x,b,S), where a and b are vectors, are
alternative ways to create exactly the same scatterplot.

23
The S string is optional and is made of the following (and more) characters:

Symbol Colour Symbol Linestyle

y yellow . point
m magenta o circle
c cyan x x-mark
r red + plus
g green * asterisk
b blue − solid
w white : dotted
k black −. dashdot
−− dashed

Thus,
>> plot(x,y,’b*’);

plots a blue asterisk at each point of the data set.

Here is an example:
>> x=pi*(-1:.01:1);
>> y=sin(x);
>> clf;
>> plot(x,y,’r-’);

6.4 Adding Titles and Labels

You can add titles and labels to your plot, using the following instructions:
>> title(’Graph’);
>> xlabel(’x-axis’);
>> ylabel(’y-axis’);

Consider a list of further graphical commands:

>> help graph2d

24
6.5 Plotting Commands
Elementary X-Y graphs commands
plot Linear plot
loglog Log-log scale plot
semilogx Semi-log scale plot (log scale for the x-axis)
semilogy Semi-log scale plot

Specialized X-Y graphs

polar Polar coordinate plot
bar Bar graph
stem Discrete sequence or ”stem” plot
stairs Stairstep plot
hist Histogram plot
fplot Plot function
comet Comet-like trajectory
title Graph title
xlabel X-axis label
ylabel Y-axis label
text Text annotation
gtext Mouse placement of text
grid Grid lines

6.6 3-D Graphics

Matlab provides a variety of functions to display 3-D data. Some plot lines in
three dimensions, while others draw surfaces and wire frames using pseudocolors
to represent a fourth dimension. The following list summarizes these functions.

plot3 Plot lines and points in 3-D space

fill3 Draw filled 3-D polygons in 3-D space
contour Contour plot
countour3 3-D Contour plot
mesh 3-D mesh surface
meshc Combination mesh/contour plot
meshz 3-D Mesh with zero plane
surf 3-D shaded surface
surfc Combination of surf/contour plot
surfl 3-D shaded surface with lighting

In addition to the graph annotation functions described in the section on 2-D

plotting, Matlab supports the following functions for annotating the graphs dis-
cussed in this section:

zlabel Creates a label for the z-axis

clabel Adds contour labels to a contour plot

25
Matlab allows us to specify the point from which we view the plot. In general,
views are defined by a 4-by-4 transformation matrix that in Matlab uses to
transform the 3-D plot to the 2-D screen. However, two functions allow us to
specify the view point in a simplified manner:

view 3-D graph viewpoint specification

viewmtx View transformation matrices

Line Plots
The three dimensional analog of the plot function is plot3. If x, y and z are
three vectors of the same length,

>> plot3(x,y,z)

generates a line in 3-D space through the points whose coordinates are the
elements of x, y and z.

Meshgrid
Matlab defines a mesh surface by the z coordinates of points above a rectangular
grid in the x-y plane. It forms a plot by joining adjacent points with straight
lines. The first step in displaying a function of two variables, (z = f (x, y)) , is
to generate X and Y matrices consisting of repeated rows and columns, respec-
tively, over the domain of the function. Then use these matrices to evaluate and
graph the function. The meshgrid function transforms the domain specified by
two vectors , x and y, into row matrices X and Y . We then use these matrices
to evaluate functions of two variables. The rows of X are copies of the vector x
and the columns of Y are copies of the vector y. Example:

>> x = -8:0.5:8;
>> y = x;
>> [X,Y] = meshgrid(x,y);
>> R = sqrt(X.^2 + Y.^2) + eps;
>> Z = sin(R)./R;
>> mesh(Z)

Note: Adding eps (floating point relative accuracy) prevents the divide by zero
which produces NaNs in the data.

26
 1

0.8

0.6

0.4

0.2

−0.2

−0.4
40

30 35
30
20 25
20
15
10 10
5
0 0

Contour Plots
Matlab supports both 2-D and 3-D functions for generating contour plots. The
contour and contour3 functions generate plots composed of lines of constant
data values obtained from a matrix input argument. We can optionally spec-
ify the number of contour lines, axis scaling, and the data value at which to
draw contour lines. For example, the following statement creates a contour plot
having 20 contour lines and using the peaks m-file to generate the input data.

>> contour(peaks,20)

Contour cycles through the line colors listed in the paragraph about the plot
function.
To create a 3-D contour plot of the same data, use contour3:

>> contour3(peaks,20)

27
Subplots
We can display multiple plots in the same window or print them on the same
piece of paper with the subplot function. subplot(m,n,p) breaks the figure
window into an m-by-n matrix of small subplots and select the p-th subplot for
the current plot. The plots are numbered along first the top row of the figure
window, then the second row, etc. For example:

>> t = 0:pi/10:2*pi;
>> [X,Y,Z] = cylinder(4*cos(t));
>> subplot(2,2,1)
>> mesh(X)
>> subplot(2,2,2)
>> mesh(Y)
>> subplot(2,2,3)
>> mesh(Z)
>> subplot(2,2,4)
>> mesh(X,Y,Z)}
gives as a result:

5 5

0 0

−5 −5
40 40
30 30
20 20 20 20
10 10
0 0 0 0

1 1

0.5 0.5

0 0
40 5
30 5
20 20 0
10 0
0 0 −5 −5

28
7 Scripts and Functions
Matlab is usually used in a command-driven mode, when we enter single-line
commands, Matlab processes them immediately and displays the results. Mat-
lab can also execute sequences of commands that are stored in files. Such files
are called m-files because they have a file type of .m as last part of the filename.
Two types of m-files can be used: Script and Function. Script m-files automate
long sequences of commands. Function m-files provide extensibility to Matlab.
They allow us to add new functions to the existing functions. Much of the
power of Matlab derives from this ability to create new functions that solve
user-specific problems.

7.1 Script m-files

When a script is invoked, Matlab simply executes the commands found in the
file. The statements in a script file can operate on existing data in the workspace,
or they can create new data on which to operate. Although scripts do not return
output arguments, any variables that they create remain in the workspace, to
be used in subsequent computations.
Scripts are useful for performing analyses, solving problems, or designing long
sequences of commands that become cumbersome to do interactively.
To create a script, choose New from the File menu and select m-file (alter-
natively you just type edit in the Matlab prompt and press Enter ). This
procedure brings up a text editor window, the Matlab editor/debugger. Let’s
type the following sequence of statements in the editor window:

A=[1 2];
B=[3 4];
C=A+B

This file can be saved as the m-file example.m on drive xx by choosing Save
from the File menu. Matlab executes the commands in example.m when you
simply type example in the Matlab command window (provided there is a file
example.m in your working directory or path1 ).

>> example
C =
4 6

7.2 Function m-files

An m-file that contains the word ”function” at the beginning of the first line
is a function file. A function m-file differs from a script file in that arguments
may be passed, and variables defined and manipulated inside the file are local
to the function and do not operate globally on the workspace. Functions files
1 Do modify the default adjustments over File/Set Path or choose the working directory

directly in the command toolbar.

29
are useful for extending Matlab, that is, creating new Matlab functions using
the Matlab language itself.
We add new functions to Matlab’s vocabulary by expressing them in term of
existing commands and functions. The general syntax for defining a function is
function[output1,..]=<name of function>(input1,..);
% include here the text of your online help statements;
The first line of a function m-file defines the m-file as a function and specifies its
name. The name of the m-file and the one of the function should be the same.
It also defines its input and output variables.
Next, there is a sequence of comment lines with the text displayed in response
to the help command:
>> help <name of function>
Finally, the remainder of the m.file contains Matlab commands that create the
output variables.
Let’s consider an example:
function y=mean(x)
% Average or mean value
% For vectors, mean(x) returns the mean value.
% For matrices, mean(x) is a row vector containing the mean of each column.
[m,n]=size(x);
if m==1
m=n;
end
y=sum(x)/m;
The existence of this file defines a new function called ”mean”. The new function
is used like any other Matlab function. Let’s test the new function:
>> b=[1:99; 101:199]’;
>> mean(b)
ans =
50 150
In the list below there are some details about the mean.m file:
• The first line declares the function name, the input arguments, and the
output arguments. Without this line, the file is a script file, instead of a
function file.
• The % symbol indicates that the rest of the line is a comment and should
be ignored.
• The first lines document the m-file and display when we type:
>> help mean

30
 • The variables m, n and y are local to mean and do not exist in the
workspace after the command mean has terminated its task. Or, if previ-
ously existing, they remain unchanged.

8 Controlling the Flow

The structure of the sequences of instructions used so far was rather straightfor-
ward; commands were executed one after the other, running from top to bottom.
Like any computer programming language Matlab offers, however, features that
allow you to control the flow of command execution.

8.1 The FOR Loop

The most common use of a FOR loop arises when a set of statements is to be
repeated a fixed, predetermined number of times n. The general form of a FOR
loop is:

for variable=expression;
statements;
end;
The variable expression is thereby a row vector of the length n, that is pro-
cessed element-by-element (in many cases it is the row vector (1:n)). Statements
stands for a sequence of statements to the program.
Two simple examples:

for i=1:10;
x(i)=i;
end;
disp(x’);

The loop assigns the values 1 to 10 to the first 10 elements of x.

for i=1:10;
for j=1:10;
A(i,j)=1/(i+j-1);
end;
end;
disp(A)
The semicolon after the inner statement suppresses repeated printing, while
disp(A) displays the final result.

31
8.2 The WHILE Loop
Matlab has its own version of the WHILE loop, which allows a statement, or
group of statements, to be repeated an indefinite number of times, under the
control of a logical condition. The general form of a WHILE loop is the following:
while condition;
statements;
end;

While the condition is satisfied, the statements will be executed.

Consider the following example:

EPS=1;
while (1+EPS)>1;
EPS=EPS/2;
end;
EPS=EPS*2
This example shows one way of computing the special Matlab value eps, which
is the smallest number that can be added to 1 such that the result is greater
than 1 using finite precision (see: help eps, we use uppercase EPS so that the
Matlab value eps is not overwritten). In this example, EPS starts at 1. As
long as (1+EPS)>1 is true (nonzero), the commands inside the WHILE loop
are evaluated. Since EPS is continually divided in two, EPS eventually gets so
small that adding EPS to 1 is no longer greater than 1 (Recall that this happens
because a computer uses a fixed number of digits to represent numbers. Matlab
uses 16 digits, so you would expect EPS to be near 10−16 .) At this point,
(1+EPS)>1 is false (zero) and the WHILE loop terminates. Finally, EPS is
multiplied by 2 because the last division by 2 made it too small by the factor
of two.

8.3 The IF Statement

A third possibility of controlling the flow in Matlab is called the IF statement.
The IF statement executes a set of instructions if a condition is satisfied. The
general form is:
if condition 1;
% commands if condition 1 is true
statements 1;
elseif condition 2;
% commands if condition 2 is true
statements 2;
else;
% commands if condition 1 and 2 are both false
statements 3;
end;

32
The last set of commands is executed if both conditions 1 and 2 are false (i.e.
not true). When you have just two conditions you skip the elseif condition
and immediately go to else. Consider the following example:
Assume a demand function of the form

 0 , if P ≤ 2
D(P ) = 1 − 0.5P , if 2 < P ≤ 3

2P −2 , otherwise

The code will be:

P=input(’Enter a price : ’); % displays the message
% on the screen and waits
% for an answer
if P<=2;
D=0; % statement 1
elseif (P>2)&(P<=3);
D=1-0.5*P; % statement 2
else; % otherwise
D=2*P^(-2); % statement 3
end;
disp(D);

8.4 How to Break a FOR or WHILE Loop

The command break terminates the execution of a FOR or WHILE loop.

9 Random Numbers
Matlab has a number of functions which allow to draw random numbers. Here
are a few useful ones:

rand(m,n) creates an m×n matrix of U(0,1)

randn(m,n) creates an m×n matrix of N(0,1)
betarnd(A,B,m,n) creates an m×n matrix of Beta(A,B)
chi2rnd(nu,m,n) creates an m×n matrix of χ2 (ν)
exprnd(mu,m,n) creates an m×n matrix of exp(µ)
gamrnd(A,B,m,n) creates an m×n matrix of Γ(A, B)
logrnd(mu,sig,m,n) creates an m×n matrix of log N (µ, σ)
trnd(nu,m,n) creates an m×n matrix of t(ν)

10 Some Examples
10.1 loaddata.m
% load_data.m, f.canova 6-6-2004

33
%
% example on how to load data, run a linear regression calling a ols
% routine, plotting the time series and computing some statistics of the
% data

% this program call CAL.m, OLS,m, MPRINT.m

% a) read in the data from an ascii file

load ypr.dat; % loading data

[nobs,nvar] = size(ypr); % # Observations in time series
y = ypr(:,1); p = ypr(:,2); r = ypr(:,3); dates = cal(1948,1,4);
%detrend y, p remove linear trend
tr = (1:nobs)’; cr = ones(nobs,1); yt = ols(y,[cr tr]);
y1 = y-yt.beta(2)*tr; % output gap
pt = ols(p,[cr tr]);
p1 = p-pt.beta(2)*tr; % detrended inflation gap

% collect detrended data in X

X = [y1 p1 r];

% b) plot the time series

header = ’Fig. 1: Time series’; F1 = figure(1); clf
set(F1,’numbertitle’,’off’) set(F1,’name’,header) tit =
strvcat(’y’,’p’,’r’); tt = (1:nobs)’; for i=1:3
subplot(2,2,i)
plot(tt,X(:,i))
title(tit(i,:))
end

% c) statistics on the time series

stats_r = [mean(r) std(r) min(r) max(r)]; stats_y1 = [mean(y1)
std(y1) min(y1) max(y1)]; stats_p1 = [mean(p1) std(p1) min(p1)
max(p1)]; inv.cnames = strvcat(’mean’,’std’,’min’,’max’);
inv.rnames = strvcat(’series’,’r’,’y1’,’p1’); disp(’*** Summary
Statistics of time series ***’)
disp(’---------------------------------------- ’)
mprint([stats_r;stats_y1;stats_p1],inv)

10.2 example var.m

% example_var.m; f. canova, version 1.2, june 28-6-2005
% This example shows how to run a VAR (there are better programs around
% but this does every step clearly).

34
% It estimates parameters, computes unconditional forecasts and impulse responses
% using a choleski decomposition

%%%%%%%%%%%%%%%%%%%%% practice exercises %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

% exercise 1: change the ordering of the varables in the var
% exercise 2: reduce the number of lags to 4 and eliminate the linear trend
% exercise 3: run the system with output, interest rates and inflation
% exercise 4: run impulse responses assuming a non-zero impact of variable
% 2 on variable 1 (need to modify the command that generates s
% choose e.g. the upper triangular value to be -0.5 o 0.5)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clear session clear

% read in the data (ascii format TxN matrix, T=# of observations N =# of variables)
load var.dat
% data from 1973:1 1993:12; order of data
% 1: y
% 2: p
% 3: short rate
% 4: m1
% 5: long rate

% pick the variables of the var system

indepvar=1 % column of variable 1
depvar=4 % column of variable 4
pociatok=1; %

dimension=(size(var)) [var(:,4) var(:,1)];

rawdat=[var(pociatok:dimension(1,1),indepvar)
var(pociatok:dimension(1,1),depvar)];
%display(’Rawdat:’)
%rawdat(:,:)
dimension=(size(rawdat)); %# of observations
totnobs=dimension(1,1); nvars=dimension(1,2);

%dat = log(rawdat); % converts raw data to logs

dat = rawdat;

% INITIALIZATION FOR VAR ESTIMATION AND FORECASTING

nfore = 12; % # OF FORECASTING STEPS TO BE CALCULATED @
nsteps = 36; % # of steps to compute for impulse response functions @
nlags = 6; % # OF LAGS INCLUDED IN VAR @
ndeterm = 2; % # OF DETERMINISTIC VARIABLES IN VAR @
depnobs = totnobs - nlags; % # OF DEPENDENT OBSERVATIONS @
ncoeffs = nvars*nlags+ndeterm; % # OF COEFFICIENTS IN EACH VAR EQUATION @

35
nparams = nvars*ncoeffs; % TOTAL NUMBER OF ESTIMATED PARAMETERS @
df = depnobs - ncoeffs; % DEGREES OF FREEDOM PER EQUATION @

% SET UP DETERMINISTIC VARIABLES

const = ones(depnobs,1); dummy=1:1:depnobs; trend=dummy’; determ =
[const trend];

% SET UP DEPENDENT VARIABLES

y = zeros(depnobs,nvars); y(:,:) = dat((nlags+1):totnobs,:);

% SET UP INDEPENDENT VARIABLES

x = zeros(depnobs,nvars*nlags); i = 1; while i <= nvars;
j = 1;
while j <= nlags;
x(:,j+((i-1)*nlags)) = dat(nlags+1-j:totnobs-j,i);
j = j+1;
end;
i = i+1;
end;

x = [x determ]; sizex=size(x); sizey=size(y);

% ESTIMATE VAR
xxx = inv(x’*x); % (x’x)^(-1) matrix
sizexxx=size(xxx);
b = xxx*(x’*y); % ols estimator (this is a system wide estimator)
sizexy=size(x’*y); sizeb=size(b);
res = y - x*b; % residuals
sizeres=size(res);
vmat = (1/depnobs)*(res’*res); %vcov of residuals
stds = sqrt(diag(vmat)); % Std devs of residuals
%disp(’Std devs’); stds

s = chol(vmat)’; % Cholesky decomp of vmat

seb = sqrt( diag(xxx)*(diag(vmat)’) ); % STANDARD ERRORS estimated coefficients
%disp(’STANDARD ERRORS OF B ’); seb
vcorr = zeros(nvars,nvars); % CORRELATION MATRIX OF residuals

i = 1; while i <= nvars;

j = 1;
while j <= nvars;
vcorr(i,j) = vmat(i,j)/(sqrt(vmat(i,i)*vmat(j,j)));
j = j+1;
end;
i = i+1;

36
end;

display(’Estimates, std errors and t statistics for variable’);

% order is estimates for the two equations, st. err of the two equations,
% t- stat for the two equations
[b(:,:) seb(:,:) b(:,:)./seb(:,:)]

%disp(’COVARIANCE MATRIX OF RESIDUALS:’); vmat

disp(’CORRELATION MATRIX OF RESIDUALS:’); vcorr

% CONSTUCTION OF A COMPANION MATRIX (FOR COMPUTING

% FORECASTS) AND COMPANION MATRIX LESS DETERMINISTIC COMPONENTS (FOR
% COMPUTING DOMINANT ROOTS, IMPULSE RESPONSES

dimcmp = ncoeffs; % dimension of companion matrix

iden = eye(dimcmp);
dimcmpld = nvars*nlags; % "ld" trailer: "less deterministic components"
idenld = eye(dimcmpld);

% CONSTRUCT FINAL OBSERVATIONS, TO BE USED BY THE COMPANION MATRIX TO COMPUTE FORECASTS

yT = zeros(ncoeffs,1); i = 1; while i <= nvars;

j = 1;
while j <= nlags;
yT(((i-1)*nlags+j),1) = y(depnobs-j+1,i);
j = j+1;
end;
i = i+1;
end;

yT(ncoeffs-1,1) = 1; % ADJ SPECIFIES CONSTANT IN yT

yT(ncoeffs,1) = depnobs+1; % ADJ SPECIFIES TREND IN yT

% CONSTRUCT COMPANION MATRIX FROM B (NCOEFFS X NVARS).

% THIS IS COMPLETELY GENERAL WITH REGARDS TO THE DIMENSIONALITY OF THE VAR
dimb=size(b);
bld = b(1:(dimb(1,1)-ndeterm),:); % REMOVES DETERMINISTIC COMPONENTS @
ald = zeros(dimcmpld,dimcmpld); % COMPANION MATRIX LESS DETERM COMPS @
a = zeros(dimcmp,dimcmp); % COMPANION MATRIX @

dimbld=size(bld’); dimidenld=size(idenld); j = 1;

while j <= nvars;

ald(((j-1)*nlags+1):(j*nlags),:) =[ bld(:,j)’;...

37
 idenld(((j-1)*nlags+1):(j*nlags-1),:)];
a(((j-1)*nlags+1):(j*nlags),:) = [b(:,j)’;...
iden(((j-1)*nlags+1):(j*nlags-1),:)];
j = j+1;
end; dimald=size(ald);

a(dimcmp-1,:) = iden(dimcmp-1,:); % ADJ SPECIFIES ROW FOR CONSTANT

a(dimcmp,:) = iden(dimcmp,:); % ADJ THIS AND NEXT LINE SPECIFY...
a(dimcmp,dimcmp-1) = 1; % ADJ ... ROW FOR TREND

% COMPUTE ROOTS OF THE VAR FROM COMPANION MATRIX

EI=eig(ald); size(EI); ROOTS=abs(EI); ROOTS = sortrows(ROOTS,1);
MAXRT = ROOTS(dimcmpld,:);

display(’Roots of the VAR’); ROOTS

% COMPUTE NFORE-STEP-AHEAD FORECASTS

fore = zeros(nfore,nvars); % WILL CONTAIN FORECASTS
an = a; % AN WILL BECOME A^N BELOW
forecst = an*yT;

j = 1; while j <= nvars;

fore(1,j) = forecst(((j-1)*nlags+1),1);
j = j+1;
end;

k = 2; while k <= nfore;

an = an*a;
forecst = an*yT;
j = 1;
while j <= nvars;
fore(k,j) = forecst(((j-1)*nlags+1),1);
j = j+1;
end;
k = k+1;
end; fore

% Here, we graph the last few observations of each series, and the forecastS.

clear dummy dummy=1:1:2*nfore; xrange =dummy’; clear dummy;

dimxrange=size(xrange); dimy=size(y); depnobs-(depnobs-nfore+1);

dimfore=size(fore);

yfore=[y(depnobs-nfore+1:depnobs,:);fore(:,:) ];
dimyfore=size(yfore);

38
message=strcat(’Last few observations of each series, and the ’...
’corresponding forecast -indep. variable # ’,num2str(indepvar,3))
figure; plot(xrange,yfore(:,1),’-b’,xrange,yfore(:,2),’-r’)
title(message) grid on ; xlabel(’Month’);
ylabel(’Observation/Forecast’)

% ************ Generate impulse response functions. *************************

% We generate nvars*nvars responses: one for each variable in response to
% each shock. Each response is of length nsteps. The following code is
% inefficient (it calculates A^nsteps nvars times), but transparent. The
% responses are a function of the ordering selected for
% the Cholesky decomposition of vmat.

%steps = seqa(0,1,nsteps+1);
pom= 0:1:nsteps; steps=pom’;clear pom;

respzeros=zeros(nsteps+1,nvars+1,nvars); i = 1;
while i <= nvars; % i indexes the shocks
esp = zeros(nsteps+1,nvars); % WILL CONTAIN IMPULSE RESPONSE FCNS
shock = zeros(nvars,1);
shock(i) = 1/s(i,i); % Normalizes shocks to (orthogonalized)
shock = s*shock; % std dev units
resp(1,:) = shock’;
bigshock = zeros(dimcmpld,1); % expands shock to be compatible with
j = 1; % companion matrix
while j <= nvars;
bigshock((j-1)*nlags+1) = shock(j);
j = j+1;
end
an = ald;
k = 2;
while k <= nsteps+1; % k indexes the step size
bigresp = an*bigshock;
j = 1;
while j <= nvars;
resp(k,j) = bigresp((j-1)*nlags+1,1);
j = j+1;
end;
an = an*ald;
k = k+1;
end;
dimsteps=size(steps);
dimresp=size(resp);
aaa=nsteps+1;

39
 display(’impulse responses to shock to variable’);
i
[steps resp(:,:)]
respzeros(:,:,i)=[resp(:,:) zeros(nsteps+1,1)];
i = i+1;
end;

%respzeros=[resp zeros(nsteps+1,1)]

for i=1:nvars;
figure;
message=strcat(’Impulse responses - shock to variable # ’,...
num2str(i,3),’ system with variables ’, num2str(indepvar,3),...
’and ’, num2str(depvar,3));
plot(steps,respzeros(:,1,i),’b.-’,steps,respzeros(:,2,i),...
’-r’,steps,respzeros(:,3,i),’k--’);
title(message) ;
grid on ;
if i==1 axis([0 20 -5.0 1.5]); end;
xlabel(’Month’);
ylabel(’Response’);
legend(’Variable # 1 ’,’Variable #4’,’zero value’);
end

10.3 example gmm.m

% Program name: example_gmm.m;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Example GMM Estimation %
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% The program uses the firm’s euler equation to estimate the ajustment cost
% parameter (gamma) and the capital share (alpha). See Adda-Cooper section 8.4.3
% for a discussion of this model.
% NOTE: the program uses fminserach which does not produce standard error
% of the estimates

%it call model1.m, varcov.m, and the data book2.m

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% exercise 1: fix alpha to 0.5 and find optimal gamma
% exercise 2: jointly estimate alpha, gamma, delta
% exercise 3: use more instruments (instruments are defined in the
% file model1)
% exercise 4: check if results are different with a two-step estimator.

40
% exercise 5: allow for serial correlation in the error (need to change
% the definition of W in varcov)
% exercise 6: cut sample size by half (use the last 190 data points)
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

clear all; % Clears all variables

clc % Clears the command window

% Start the timer

tic

% Load the data set

% invpc = investment rate (i)
% shc = productivity shock (A)
% prok = profit rate (AK^alpha-1)
load book2; data = [invpc; shc; prok];

% Parameters
beta = .95; % these are calibrated
delta = .069; guess = [1 1];
p = 1; % Normalize price to one

T = length(data);
maxiter = 10; % Convergence is fast therefore only need two iterations
% this is a parameter to play with
iter = 1;
W = eye(2); % Step One of the GMM algorithm

while iter <= maxiter;

[results J]= fminsearch(@model1,guess,[],data,beta,delta,p,T,W);
gamma = results(1);
alpha = results(2);

% Calculate the Optimal Weight Matrix (Step Two)

W = varcov(gamma,alpha,data,beta,delta,p,T);
iter = iter +1;
end

disp(’*****************’) disp(’ Results’)

disp(’*****************’) disp([’Gamma = ’ num2str(gamma)])
disp([’Alpha = ’ num2str(alpha)]) disp([’J statistic = ’
num2str(T*J)]) disp(’*****************’)

% End the timer

toc

41
10.4 draw rn.m
%%% example of how to draw random numbers

%%%% Random Number Generators in matlab

%%%% randn.m Normal distribution
%%%% rand.m Uniform distribution
%%%% gamrnd.m Gamma distribution
%%%% wishrnd.m Whishart distribution
%%%% iwishrnd.m Inverted Whishart distribution
%%%% mvnrnd.m Multivariate Normal
%%%% betarnd.m Beta Distribution
%%%% type ’help xxx.m’ for information on what these functiosn do.

% note: without the seed the draws will be different at each run.

% exercise 1: draw measurement errors from a uniform distribution with

% the same variance
% exercise 2: draw technology shocks from a t- distribution with 4 dof.
% exercise 3: draw consumption measurement error from a beta(5,2)

% parameters
ndraws=2000; % number of draws
k0=100; % initial condition on capital
A=2.83; % constant in the production function
etta=0.36; % share of capital in production function
betta=0.99; % discount factor
ssig1=0.1; % standard error of technology shock
ssig2=0.06; % measurement errors
ssig3=0.02; ssig4=0.2; ssig5=0.01;

for t=1:ndraws
uu1(t)=1+randn(1)*ssig1;
uu2(t)=randn(1)*ssig2;
uu3(t)=randn(1)*ssig3;
uu4(t)=randn(1)*ssig4;
uu5(t)=randn(1,1)*ssig5;
kk(t)=A*etta*betta*k0*uu1(t)+uu2(t);
yy(t)=A*k0^etta*uu1(t)+uu3(t);
cc(t)=yy(t)*(1-etta*betta)+uu4(t);
rr(t)=etta*(yy(t)/k0)+uu5(t);
k0=kk(t);
end

% this is the same loop in matrix format

42
%
%uu1=ones(ndraws,1)+rand(ndraws,1).*ssig1;
%uu2=rand(ndraws,1)*ssig2;
%uu3=rand(ndraws,1)*ssig3;
%uu4=rand(ndraws,1)*ssig4;
%uu5=rand(ndraws,1)*ssig5;
%kk=A*etta*betta*uu1+uu2;
%yy=A*k0*etta*uu1+uu3;
%cc=yy*(1-etta*betta)+uu4;
%k0=kk(t);

% collecting statistics of simulated data

cc1=cc(400:end); yy1=yy(400:end); disp([’ Mean
Std Skewness Kurtosis’]) disp([’Consumption ’
num2str([mean(cc1) sqrt(cov(cc1)) skewness(cc1) kurtosis(cc1)])])
disp([’Output ’ num2str([mean(yy1) sqrt(cov(yy1))
skewness(yy1) kurtosis(yy1)])])

% scalign variables
cc2=(cc1-mean(cc1))/sqrt(cov(cc1));
yy2=(yy1-mean(yy1))/sqrt(cov(yy1));

% histogram and density estimation

[f1]=ksdensity(cc2,sort(cc2)); [f2]=ksdensity(yy2,sort(yy2));
figure(1), subplot(2,1,1),hist(cc2,100),axis
tight,title(’Consumption’); subplot(2,1,2),hist(yy2,100),axis
tight,title(’Output’); figure(2),
subplot(2,1,1),plot(sort(cc2),f1,’r’),axis
tight,title(’Consumption’)
subplot(2,1,2),plot(sort(yy2),f2,’r’),axis tight,title(’Output’);

43