Table of Contents:
1. WARNING: Serious potential bug within MATLAB programs
2. Special Math408 MATLAB functions
3. Printing and copying program output
4. Keeping a diary (writing a text file from selected command-window output):
5. Formatting tables using fprintf()
6. Within-program functions and subroutines
7. Using MATLAB from a command-prompt window
1. WARNING: Serious potential bug within MATLAB programs
If you enter
within a MATLAB program, you will get x=6 and not the correct value for 1+2+3+4+5, which is x=15. The second line will be completely ignored, without even an error message.
You can get an idea of how MATLAB is thinking if you enter the same two lines in the MATLAB command window. Your entry and MATLAB's response will be
What has happened is that MATLAB interprets ``x = 1 + 2 +3'' as a complete command, and returns x=6. In the next line ``+4+5;'', MATLAB thinks that you just want to add 4 and 5 together, with the trailing semicolon (;) meaning that you are not interested in the answer.
While this behavior causes no problems at the command line, within a MATLAB program it can cause incorrect results with absolutely no indication of any problems.
2. Special Math408 MATLAB Functions:
The files
normpval.m normint.m pvchisq.m pvtdist.m pvfdist.m binomtail.m getmidranks.m
on the Math408 Web site provide functions of the same names that can be
used in MatLab programs.
Similarly, pvtdist.m can be used to find two-sided P-values based
on the Student-t distribution. Specifically, pval=pvtdist(tt,19)
returns the two-sided P-value for T=tt, where T (given H_0) has a
Student's t-distribution with 19 degrees of freedom. See the comments in
the first few lines of each of these function files for the syntax of
using the function. Functions similar to the first five functions listed
above are in the MATLAB Statistics Toolbox, but your particular copy of
MATLAB may not have this toolbox installed.
To use one of the functions listed above in a program, copy the MATLAB
function file (such as pvtdist.m or normpval.m) into the
default directory (or Current Directory) of your MATLAB program. The
corresponding function (pvtdist() or normpval()) will then
be immediately available to you. The name of the Current Directory is
shown in an edit-and-browse window at the top center or top right of the
MATLAB screen. (Function files can also be copied into any directory on
the ``MATLAB path'', if you wanted to keep MATLAB function files in
another directory.)
You can also find the name of the Current Directory by entering
>> pwd
in the MATLAB command window. (>> stands for the MATLAB cursor.) Here
pwd stands for ``print working directory''. (The name is borrowed
from UNIX.) The contents of the Current Directory will sometimes be
displayed on the left-hand side of the MATLAB screen. Alternatively, you
can display the contents of the Current Directory by entering
>> dir OR Desktop | Current Directory
Here ``dir'' is entered in the MATLAB Command Window, and | stands for
``and then click on''. That is, click on that option in the Desktop
drop-down menu. The MATLAB command
>> what
lists the available MATLAB program and function files in the current
directory.
In general, MATLAB program files are of one of two types: MATLAB programs
like Firstm.m and WcxSignRank.m on the Ma408 Web site, and MATLAB function
files like normpval.m etc. MATLAB function files have a slightly different
syntax than MATLAB program files. The names of function files on the
Math408 Web site begin with a lower-case letter and those of example
MATLAB programs begin with an upper-case letter.
File | Print...
where | stands for ``and then click on''. That is, click on that option in
the drop-down File menu.
The difficulty here is that this prints THE ENTIRE Command Window output,
perhaps including the output of several previous versions of your program
that did not work. To CLEAR the Command Window, enter
>> clc OR Edit | Clear Command Window
where, again, >> stands for the cursor in the MATLAB Command Window. You
can now enter the name of your program file (such as >> myfile
for myfile.m) and get clean output.
Another difficulty is that the above operations do not clear MATLAB's
internal variable stack. That is, if you calculate a variable named
Zvalue (for example), then accidentally delete the code from your
program that calculated Zvalue, then the previous value of
Zvalue will be used and/or displayed whenever you run
myfile. You may only notice that something is wrong if you run
myfile again the next day after exiting MATLAB and then get an
Undefined Variable error message. To clear MATLAB's interval variables,
enter
>> clear OR Edit | Clear Workspace
As you might expect, not clearing the variable stack before running a
MATLAB program can cause incorrect program output due to typos or due to
unnoticed references to variables that you no longer want. These types of
errors are very hard to detect in the output, but can be eliminated by
periodically clearing the variables as above.
>> diary mylog.txt
Thereafter, anything that is displayed in the Command Window will also be
written to mylog.txt. To stop writing (and close the file so that
you can print it), enter
>> diary off
Entering >> diary mylog.txt again (or >> diary on) continues
to append text to mylog.txt. To start over, enter
>> delete mylog.txt
This deletes the text file, so that you can start afresh.
However, MATLAB has borrowed (and slightly improved) a function from
C/Java for better displays of matrices or tables. If you enter in a
program (or in the MATLAB command window):
In the `format string' ' %d %3d %4.1f\n' in the fprintf() command,
If fprintf() has a matrix argument after the format string, each COLUMN of
the matrix is matched to the format items in the format string. In this
case, you will usually want to use the TRANSPOSE of a matrix as the
argument in the data. For example, for the 4x3 matrix xx defined above
A FINAL COMMENT: In MATLAB, you MUST ENCLOSE the format string with single
quotes (' ') and NOT double quotes (" "). The latter will give a
format error. In contrast, in C or JAVA, you MUST USE double quotes to
enclose the format string, and single quotes will give a format error.
This difference takes a bit of getting used to if you are more used to C
or MATLAB and switch to the other.
MATLAB also has a feature for defining and using a subroutine or function
within the same MATLAB program file. For example, if you say
near the top of your MATLAB program file, then f(x,y) will evaluate as
x^2+y^2 in the rest of your program, and in particular f(2,3)=13.
This feature works by MATLAB parsing the single input string in the
inline() function for strings of characters that are valid MATLAB variable
names, which are an alphabetical character (a-z,A-Z) followed by
alphanumeric characters (a-z,A-Z,0-9) up to the next non-alphanumeric
character. Thus
defines exactly the same function. Enter
Start | All Programs | Accessories | Command Prompt
If the window opens with marginally-legible light-grey text on a black
background, enter
color F0
(that is, ``color F-zero'') in the command-prompt window. You will now get
legible black lettering on a parchment-white background in the
Command-Prompt Window, exactly as in a Notepad window.
Navigate to your directory of choice by using cd (``change
directory'') commands in the command-prompt window. Then enter
matlab at the command line. The program MATLAB will now appear with
that directory as the Current Directory.
You can call MATLAB from the command line to execute a particular program
file and write the (Command-Window) output to a textfile by entering
matlab -r myfile.m -logfile myoutputfile.txt
MATLAB will run the program file myfile.m as soon as it loads and
write the output to myoutputfile.txt. The program MATLAB will
remain open, but can be closed in the usual way unless you want to use it
further.
Last modified February 26, 2009
3. Printing and Copying Program Output:
You can print the contents of the MATLAB Command Window by clicking on
4. Keeping a diary (writing a text file from
selected command-window output):
You can write the Command Window to a text file, or part of the Command
Window to a text file, by first entering
5. Formatting tables using fprintf()
If you enter
xx = [ 1 23 3;
2 720 4;
3 11 1;
4 12 2 ];
then you can display the matrix xx more-or-less exactly as it is displayed
above by entering xx or disp(xx), either in the MATLAB Command Window or
in a MATLAB program. However, if
xx = [ 1 23 3;
2 720 4;
3 11 1.5;
4 11 1.5 ];
then entering xx or disp(xx) will display
1.0000 23.0000 3.0000
2.0000 720.0000 4.0000
3.0000 11.0000 1.5000
4.0000 11.0000 1.5000
The first column is now confusing as row or observation numbers, and the
final column is confusing as midranks.
fprintf(' %d %3d %4.1f\n', 3, 11, 1.5)
then you will get as output
%d stands for `display as integer'
%3d stands for `display as integer, using AT LEAST 3 spaces, right justified'
%4.1f stands for `display as number with possible fractional part,
using AT LEAST 4 spaces, with 1 SPACE beyond the decimal point'
\n stands for end-of-line (or new line), which you must say explicitly
Arguments following the format string in fprintf() are matched one-by-one
with the format instuctions in the format string.
fprintf(' %d %3d %4.1f\n', xx')
displays
1 23 3
2 720 4
3 11 1.5
4 11 1.5
Note that `%3d' has causes the second column to align correctly, and
`%4.1f' does similarly for the third column. In this example, using `%4.1f'
instead of `%3.1f' has introduced an extra space between the second and
third columns, but also provides protection in case you later enter a row
with data `7 311 11.5'. That is, the 3rd column will still line up at the
decimal points.
6. Within-program functions and subroutines
The usual way to include a subroutine or a special function in a MATLAB
program is to write a MATLAB function file. This is convenient to do, but
if you have a directory with several MATLAB program files, each with
several MATLAB function files, it may be difficult to see what belongs to
whom.
help inline or
doc inline at the MATLAB command line for more information.
7. Using MATLAB from a Command Prompt (or Console or Terminal) Window:
If you want to run MATLAB so that it opens with a given directory (for
example, c:\Temp) as its Current Directory, first open a
Command-Prompt Window by clicking on
The Windows batch file cmatlab.bat on the Math408 Web site
automates this process. That is, enter cmatlab myfile to
(i) load MATLAB, (ii) run myfile.m, and (iii) write
Command-Window output to myfile.txt. The batch file
cmatlab.bat includes options to make MATLAB run faster by not
loading the full MATLAB GUI window, but does leave open a more primitive
version of MATLAB, which you might still want to close.