Programming with SPSS

Syntax and Macros

SPSS Inc.
233 S Wacker Drive, 11th Floor
Chicago, Illinois 60606
312.651.3000

Training Department
800.543.6607

v10.0 Revised 12/31/99 ss

SPSS Neural Connection, SPSS QI Analyst, SPSS for Windows, SPSS Data
Entry II, SPSS-X, SCSS, SPSS/PC, SPSS/PC+, SPSS Categories, SPSS Graphics,
SPSS Professional Statistics, SPSS Advanced Statistics, SPSS Tables, SPSS
Trends, SPSS Exact Tests, and SPSS Missing Value are the trademarks of SPSS
Inc. for its proprietary computer software. CHAID for Windows is the trademark
of SPSS Inc. and Statistical Innovations Inc. for its proprietary computer
software. Excel for Windows and Word for Windows are trademarks of Microsoft;
dBase is a trademark of Borland; Lotus 1-2-3 is a trademark of Lotus
Development Corp. No material describing such software may be produced or
distributed without the written permission of the owners of the trademark and
license rights in the software and the copyrights in the published materials.

General notice: Other product names mentioned herein are used for
identification purposes only and may be trademarks of their respective
companies.

Copyright(c) 2000 by SPSS Inc.

All rights reserved.

Printed in the United States of America.

No part of this publication may be reproduced or distributed in any form or by

any means, or stored on a database or retrieval system, without the prior written
permission of the publisher, except as permitted under the United States
Copyright Act of 1976.
SPSS Training

Programming with SPSS Syntax and Macros

Table of Contents

Chapter 1 Introduction and Syntax Review

A Data Manipulation Example 1-1
A Macro Example 1-3
Rules and Aids for SPSS Syntax 1-6
Advice for Those Working with Syntax 1-9
Summary 1-10

Chapter 2 Basic SPSS Programming Concepts

Command Types in SPSS 2-2
The Three Types of SPSS Programming 2-3
SPSS Data Definition 2-5
SPSS Programming Constructs 2-6
Do If & End If 2-6
Do Repeat & End Repeat 2-7
Loop & End Loop 2-9
Scratch Variables 2-11
Vector 2-12
Summary 2-16

Chapter 3 Complex File Types

ASCII Data and Records 3-2
File Types 3-2
Syntax Basics 3-3
Data File Structure 3-4
Grouped Data 3-4
Mixed Data 3-5
Nested Data 3-6
Reading a Mixed File 3-7
Errors in the Data 3-10
Grouped File Type Without Record Information 3-12
Summary 3-17

Chapter 4 Input Programs

Syntax Components 4-2
Example 1: Change the Case Base of a File 4-2
End of Case Processing 4-7
End of File Processing 4-9
Checking Input Programs 4-10
Incomplete Input Programs 4-11

Table of Contents - 1
SPSS Training

Example 2: Reading Files with Missing Identifiers 4-14

When Things Go Wrong 4-20
Summary 4-21

Chapter 5 Advanced Data Manipulation

Reading a Comma-Delimited File 5-2
Reading Multiple Cases on the Same Record 5-7
An Existing SPSS Data File with Repeating Data 5-11
Print Command for Diagnostics 5-15
Practical Example: Consolidating Transactions 5-17
Summary 5-24
Appendix: Identifying Missing Values by Case 5-25

Chapter 6 Introduction to Macros

Macro Basics 6-2
Macro Arguments 6-3
Macro Tokens 6-3
Viewing a Macro Expansion 6-7
Keyword Arguments 6-8
Using a Varying Number of Tokens 6-10
When Things Go Wrong 6-15
Summary 6-18

Chapter 7 Advanced Macros

Looping in Macros 7-2
Producing Several Clustered Bar Charts 7-2
Double Loops in Macros 7-5
String Manipulation Functions 7-7
Direct Assignment of Macro Variables 7-7
Conditional Processing 7-7
Creating Concatenated Stub and Banner Tables 7-8
Additional Recommendations 7-13
Summary 7-13

Chapter 8 Macro Tricks

Combining Input Programs and Macros 8-2
Ordering Tables and Charts 8-6
The Case of the Disappearing Command 8-9
Summary 8-14

Exercises Exercises
Exercises E-1

Table of Contents - 2
SPSS Training

Chapter 1 Introduction and Syntax

Review

Topics A Data Manipulation Example

A Macro Example
Rules and Aids for SPSS Syntax
Advice for Those Working with Syntax

T
INTRODUCTION his course has two major topical areas. We will review how to use
SPSS Syntax to perform complex data manipulations that are not
available under the SPSS menu system. This will be of interest to
those who need to read complex data files from legacy computer systems
(for example, legacy health care data, transaction oriented sales systems)
and those who find they need to reorganize their data in order to perform
a desired analysis. Examples of the latter include marketing and
customer relationship studies in which a number of products (or services
of an company) are rated on each of many attributes. All information
from a respondent is typically stored in a single record, but needs to be
spread across multiple records in order for factor analysis and perceptual
mapping to be performed. When preparing data for churn (customer
retention – for telecoms, credit card issuers, insurance companies)
studies, comparisons might need to be made across transactional records
sorted by customer ID and date. SPSS Syntax permits a richer array of
data manipulations in this content than would the menu system. In
short, we will examine uses of SPSS Syntax to facilitate analysis of files
with complex structures or files that must be restructured for a desired
analysis.

The second topical area concerns automation in SPSS through the

SPSS macro language. SPSS macros can generate SPSS Syntax, which is
then executed. For this reason, macros are very handy in situations
where SPSS Syntax needs to be run repeatedly, but with minor and
systematic changes each time. For example, you might wish to produce
thirty Interactive graphs, each a clustered bar chart containing a
demographic variable and one of thirty rating scale variables. Within the
SPSS menu system, changes would have to be made in the Interactive
graph dialog box for each graph. Instead, an SPSS macro could generate
the SPSS Syntax for each interactive graph within a loop, substituting a
new rating scale variable name per iteration. In this way, the SPSS
macro language can automate what would otherwise be time-consuming
tasks for the analyst.

Introduction and Syntax Review 1 - 1

SPSS Training
Since these topics involve SPSS Syntax, we will use the dialog boxes
within SPSS infrequently. A prerequisite for this course is familiarity
with SPSS Syntax at the level of our Introduction to SPSS Syntax
training course. In this chapter, we will present a sample of data
manipulation with SPSS syntax and a macro example, and provide a
brief review of and some recommendations for SPSS Syntax.

A DATA To illustrate the type of data manipulation that can be performed with
MANIPULATION SPSS Syntax, we will display the beginning and final form of a data file
recording SPSS training course purchases. Within the Training
EXAMPLE department, there was interest in examining patterns of training courses
taken by SPSS customers, and an analysis was performed using SPSS
Clementine. However, a requirement of the analysis was a data set in
which all courses taken by a customer (an SPSS ID) were contained in a
single customer record.
The original data file, extracted from a transaction database,
contained one record per course taken, since an instance of a course being
taken by a customer constituted a sales transaction. We show this below.

Figure 1.1 Training Sales Data - Transaction File

Each record in this file is a sales transaction involving a specific

training course. The two fields displayed are customer ID and course
taken (which contains city, sequence within the year, and training course
code information). Additional fields, such as date and price, were
previously removed since they were not needed for this analysis. Here
different courses taken by an individual SPSS customer are scattered
throughout the file. Even if the file were sorted by customer ID, the fact
that the training course history for a single customer is spread across a

Introduction and Syntax Review 1 - 2

SPSS Training
number of records, that varies from customer to customer, would create
difficulties for the analysis procedures.

Figure 1.2 Training Sales Data - One Record Per Customer

The training data has been reorganized so there is a single record per
customer ID and a separate variable for each training course. These
course variables are coded 1 if a customer signed up for the course and 0
if not. This structure makes it easy to explore associations among
training courses taken by customers. The SPSS syntax to perform the
data reorganization involved two steps: creating a vector of variables in
which each variable represented a specific course, and aggregating this
file to the customer ID level. The logic behind these operations is
reviewed in Chapter 5.

A MACRO We mentioned earlier that SPSS macros generate SPSS Syntax. A

EXAMPLE common use of macros is to produce a series of syntax commands that
vary in specific ways, for example, a set of Interactive Graph or Tables
commands in which each command runs an analysis based on a different
variable. Thus one macro produces the same result as many syntax
commands (which it creates) or interactions with a dialog box. To
demonstrate, we will display results from a macro that produces a set of
Interactive graphs, substituting different variables in clustered bar
charts (this macro is discussed in Chapter 7).

Introduction and Syntax Review 1 - 3

SPSS Training
Figure 1.3 Create Bar Chart Dialog Box

The dialog above will create a clustered bar chart displaying attitude
toward government action on health for different marital status groups.
Note that only a single variable can be placed in horizontal and Color
boxes. (Note: actually multiple variables can be placed in a single box, but
this action will not produce multiple charts.) Thus creating a series of
charts, in which either the horizontal axis or Color variables change,
would require repeated visits to this dialog, substituting one variable at a
time. However, the macro below can build many clustered bar charts.

Introduction and Syntax Review 1 - 4

SPSS Training
Figure 1.4 Macro to Produce Multiple Bar Charts (Interactive Graphs)

The details of this macro (Clu2IBar) will be discussed in Chapter 7.

However, we point out that the IGRAPH command, which was pasted by
clicking the Paste pushbutton in the Create Bar Chart dialog box (see
Figure 1.3), is nested within two loops, each of which iterates over a list
of variables supplied by the user. The invocation of the macro (last line in
program), supplies two variable names for the horizontal axis variable
and three variable names for the cluster variable. Thus six IGRAPH
commands will be generated, resulting in the six bar charts shown below.

Introduction and Syntax Review 1 - 5

SPSS Training
Figure 1.5 Bar Charts Produced from Clu2IBar Macro

The six Interactive Graphs in the Outline pane were produced from
the macro. In this way, macros can automate the running of sets of
similar analyses. The second section of this course reviews SPSS macros
in detail.

RULES AND AIDS Since this course involves either the writing or generation of SPSS
Syntax, we begin be reviewing the rules of SPSS Syntax and how to
FOR SPSS
obtain syntax help.
SYNTAX
The syntax rules for editing and writing SPSS commands are as
follows:

1. Each new command must begin on a new line and end with a
period (.) or a blank line.
2. *Each command must begin in the first column of a new line.
3. *Continuation lines of a command must be indented at least
one space.
4. Variable names must be spelled out fully.
5. Subcommands must be separated with a forward slash (/).
The slash before the first subcommand is usually optional.
6. *Each line of command syntax cannot exceed 80 characters.

*Not required when running from a Syntax window, but required when
using the INCLUDE command or the SPSS Production Facility

Introduction and Syntax Review 1 - 6

SPSS Training
Syntax commands produced by clicking the Paste pushbutton from an
SPSS dialog box will conform to these rules, so the important issue is to
remember them when editing or entering syntax.

There are several useful sources of help when writing SPSS syntax. A
quick reminder of the keywords and requirements for an SPSS command
are only a tool-button click away. To demonstrate:

From within SPSS:

Click File..Open..Syntax
Move to the c:\Train\ProgSynMac directory
Double click on TransactionAgg
Scroll down to the Vector command
Click on the Vector command (so the insertion pointer touches it)

Click the Syntax Help tool

Figure 1.6 Syntax Help

In this syntax summary for the Vector command, subcommand

names and keywords are shown in upper case (some simple commands,
like Vector, have no subcommands); lower case elements describe
specifications that you supply (e.g. varlist indicates a list of variable
names that you provide). Sections of the command enclosed in square
brackets [ ] are optional, while those in braces indicate sets from which a
single choice can be made. To focus on the required elements of the
command, scan only sections not enclosed within square brackets.

Introduction and Syntax Review 1 - 7

SPSS Training

While the syntax information about VECTOR is complete, there is no

explanation about what each specification does. Although some might be
obvious from their names, many are not. Complete documentation about
SPSS for Windows syntax commands can be found in the SPSS 10.0
Syntax Reference Guide (included on the CD-ROM containing the SPSS
10.0 program). If copied to your hard drive during SPSS installation, you
can access the guide from the main menu by clicking Help..Syntax
Guide..Base (or one of the optional modules). Commands are listed
alphabetically and the subcommand options are fully explained.
Experienced syntax command users, needing only reminders, can work
from the Syntax Help windows. For others, the Syntax Reference Guide is
necessary.

Note The sequence below assumes the SPSS 10.0 Syntax Reference Guide has
been installed on your machine. If not, it can be installed from the SPSS
for Windows 10.0 CD-ROM.

Click Help..Syntax Guide..Base

Click the arrow beside Commands in the Outline pane
Scroll down to VECTOR
Click the arrow beside VECTOR in the Outline pane
Click on VECTOR in the Outline pane

Figure 1.7 SPSS Base 10.0 Syntax Reference for Vector Command

Introduction and Syntax Review 1 - 8

 SPSS Training
In addition to the syntax summary accessed through the Syntax Help
tool, the SPSS 10.0 Syntax Reference Guide contains discussion,
explanation and examples. All are useful when investigating the
possibilities of an SPSS command. For those working often with SPSS
Syntax, we strongly recommend installing the reference guide on your
machine or purchasing a copy of the SPSS 10.0 Syntax Reference Guide in
book form.

Click File..Exit to exit Adobe Acrobat and the SPSS 10.0 Syntax
Reference Guide

ADVICE FOR THOSE Finally, it is worth mentioning, at the risk of being obvious, several
WORKING WITH recommendations for those working with SPSS Syntax.
SYNTAX
Display Syntax commands as Log Items
By default, SPSS does not display syntax in the Viewer window, although
it is written to the SPSS journal file. If SPSS issues any error or warning
messages, it is useful to see which command they follow. For this reason,
while writing, editing and testing SPSS syntax, we recommend you set on
the option to display syntax as a log item in the Viewer window. We will
do this explicitly in the next chapter, but view the Options dialog here.

Click Edit..Options
Click the Viewer tab

Figure 1.8 Viewer Options

Introduction and Syntax Review 1 - 9

SPSS Training
The checkbox in the lower left corner of Viewer tab within the
Options dialog controls whether SPSS syntax commands display in the
log.

Develop Basic Syntax Using Dialog Boxes

Although this course proves the exception to the rule (discussing Input
Programs, Vectors and Loops), most SPSS commands can be generated by
clicking the Paste pushbutton of the relevant dialog box. It is to your
advantage to use dialogs, when possible, to construct the basic SPSS
syntax and then edit is as needed. This will minimize errors by reducing
your opportunity to make typing mistakes.

Use File New to Clear Data

If errors lead to complex SPSS data operations (Input Program) not
completing, SPSS can be left in a waiting state. That is, it will not
properly process new instructions until it has closure on the interrupted
sequence. To clear the current data state of SPSS, you can run the NEW
DATA command or click File..New..Data. When writing complex Input
Programs (see Chapter 4 and 5), you might consider beginning the
program with a NEW DATA command to insure that any problem data
state has been cleared prior to running your program. We illustrate this
in Chapter 4.

Test After Each Step

It is a difficult challenge to foresee all possible data problems that a
program might face and it rarely the case that any program, for that
matter, runs correctly the first time. For these reasons it is important to
systematically test and check results at each stage of the process. Full
programming methodologies have been developed to this end. Here we
merely wish to recommend that, during development, you include
displays or procedures to check the results of each set of operations, so
that when something goes awry you have a way of isolating and
identifying the problem. The Data Editor display in SPSS is useful for
this purpose, as are the Frequencies, Crosstabs, Case Summaries and
List procedures, and the Print transformation. These will be used
repeatedly in the examples we present in this course, and we can assure
you that the consultants in the SPSS Consulting group use them heavily.

Delete Items in Viewer Window

If an error occurs, and you have read and understood the warning
messages in the Viewer window, it is often a good idea to delete the
results in the Viewer window before rerunning your program. This is
because syntax commands may be appended to the last Log item in the
Outline pane, making it difficult to distinguish the old warning messages
from the new results.

SUMMARY In this chapter we introduced, with examples, the major focus areas of
this course: Syntax for complex data manipulation and SPSS Macros. We
also briefly reviewed the available help for SPSS Syntax and offered some
advice for those working with SPSS Syntax.

Introduction and Syntax Review 1 - 10

SPSS Training

Chapter 2 Basic SPSS Programming

Concepts

Topics Introduction
Command Types in SPSS
The Three Types of SPSS Programming
SPSS Data Definition
SPSS Programming Constructs
A Note About Program Execution
Analysis Tip: Reordering Variables

A
INTRODUCTION ll SPSS procedures are built upon a powerful programming
language that has been consistent, though greatly extended, since
SPSS was first developed as a mainframe statistical software
program. This course will teach you how to use this language and other
features for file and data input and manipulation, and for overall control
of SPSS execution.
The SPSS language, called syntax, is generated by the program every
time a user clicks on the OK button in a dialog box to execute a
procedure. Behind the scenes, SPSS builds syntax to send to the SPSS
central engine to execute a particular procedure or transformation. Using
the Paste button in a dialog box places a copy of that syntax in a Syntax
window so that it can be edited or saved and used again.
SPSS syntax is also often called a command or set of commands. The
grammar or rules associated with commands are fairly simple, and we
will review them as necessary throughout the chapters.

Note about Data SPSS for Windows 10.0 can run entirely on your desktop machine.
Alternatively, an SPSS Client, through which you request analyses and
File Access
view results, can run on your desktop, while the analyses are run by the
When Running SPSS Server, possibly located on a different machine. In this course,
SPSS from a except for the directory you use to access the training data files, it makes
Remote Server no difference whether the SPSS Server is located on your desktop or a
different computer. The SPSS Server Login dialog (click File..Switch
Server) allows you to connect to a remote SPSS Server (if installed on
your network).
If you are running SPSS from a Remote (not Local) server, then to
use the data files accompanying this course, they must be copied either to
the server running SPSS or to a directory that can be accessed by
(mapped from) the server. The directory references in this guide assume
you are running SPSS as a local server and can thus directly access files
stored on your hard drive.

Basic SPSS Programming Concepts 2 - 1

SPSS Training

COMMAND Most users successfully program in SPSS without a complete

TYPES IN SPSS understanding of the various command and program states of SPSS, and
you can too. Nevertheless, it helps to know a little about this subject,
particularly to help put the various capabilities of SPSS in context. Many
users know the difference between transformations and procedures, the
two main types of commands, but there are a few others:

File Definition Commands: As their name implies, all these

commands are used to input data into SPSS. They include
familiar commands such as GET, DATA LIST, or GET
CAPTURE ODBC, but also others like FILE TYPE, IMPORT,
or MATCH FILES.

Input Program Commands: These are specialized commands,

also used to input or create data in SPSS. These commands
are more esoteric and include RECORD TYPE, REPEATING
DATA, and END CASE. We have more to say about this
below when discussing the data step in SPSS.

Transformation Commands: These commands are quite varied

in their operation, but the key element they share in common
is that they neither input data nor analyze data. Instead, they
modify data (COMPUTE, RECODE), create new variables
(VECTOR, NUMERIC), write out data (WRITE), or label data
(VARIABLE LABELS). To be specific, transformations do not
cause SPSS to read the data file.

Procedure Commands: Almost all of these commands analyze

data. However, the actual definition of a procedure in SPSS is
a command that causes data to be read. Thus, SAVE is also a
procedure because it causes the data to be read and an SPSS
system file ( an .SAV extension) to be created.

Utility Commands: These commands handle a variety of chores.

They include commands to add comments (COMMENT,
DOCUMENT), to define a new file (NEW FILE), and to define
macros (DEFINE--!END DEFINE).

Knowing about command types will be helpful in understanding how

and why a program operates. For example, XSAVE, an alternative to the
SAVE command, can be used within a loop because it is a transformation,
not a procedure.

Basic SPSS Programming Concepts 2 - 2

SPSS Training

THE THREE Logically, there are three general methods of programming in SPSS. Two
TYPES OF SPSS of them involve syntax, while the third uses a version of the Basic
programming language (Sax Basic).
PROGRAMMING
Standard Syntax: These programs are the most common and simply
involve writing a series of SPSS commands to accomplish a set of tasks.
An example of a simple program is shown in the box (this program uses
the FILE TYPE command to read a non-standard ASCII data file). In a
standard syntax program, each command does one thing, and it does not
refer to other SPSS syntax. Standard programs are executed either
through the Run button, the INCLUDE command, or the SPSS
Production Facility.

* SPSS Example to read a nested file * .

FILE TYPE NESTED FILE 'C:\TEST.DAT' / RECORD RECID 1
(A) CASE 3 (F).
RECORD TYPE 'H'.
DATA LIST /H1 to H10 5-14.
RECORD TYPE 'F'.
DATA LIST /F1 to F5 15-19.
RECORD TYPE 'P'.
DATA LIST /CASEX 3 P1 to P3 20-22.
END FILE TYPE.

LIST

Macros: Many programs allow users to define macros, which are

typically a series of commands grouped together as a single command to
make everyday tasks easier and more convenient. Macros can often be
assigned to a toolbar or menu to make them readily accessible. Normally,
macros are saved as a series of instructions in a special macro language.

SPSS Macros are a bit different and not exactly parallel to the more
common definition of a macro in other programs. First, they are written
in SPSS syntax (plus a few special macro commands) and are essentially
executed like any other syntax file. Second, they generate customized
SPSS command syntax, i.e., standard syntax, to reduce the time and
effort needed by the program writer to perform complex and repetitive
tasks. There is no special macro editor in SPSS or macro facility to
execute a macro; again, a macro is simply a specialized syntax file. Below
is an example of a macro that automates the production of a bar graph
and the insertion of today’s date into the title of a graph (this program
actually defines two macros). The macro begins with DEFINE and ends
with !ENDDEFINE.

Basic SPSS Programming Concepts 2 - 3

SPSS Training

DEFINE !GRAPHIT ( ARG1 !TOKENS(1) / ARG2 !TOKENS(1)/

ARG3 !TOKENS(1)) .
GRAPH /BAR(SIMPLE)=COUNT BY !ARG1
/TITLE=
"EXAMPLE OF MACRO GRAPHING TITLES AND DATES"
/SUBTITLE !QUOTE(!CONCAT(!UNQUOTE(!ARG2),
"Something ",!UNQUOTE(!ARG3)))
/FOOTNOTE= !CONCAT(""Today is ",!EVAL(@DATEIT),""" ).
!ENDDEFINE .
DATA LIST FREE / A.
BEGIN DATA
1232124
END DATA.
* The following defines a macro entity with today’s date *.
DO IF $CASENUM=1 .
WRITE OUTFILE 'TMP'/
'DEFINE @DATEIT()',$TIME(ADATE),'!ENDDEFINE .'.
END IF .
EXECUTE.
INCLUDE 'TMP' .
!GRAPHIT ARG1=A ARG2="ARGUMENT 2"
ARG3="ARGUMENT 3".

SPSS Scripting Facility: The scripting facility also allows you to

automate tasks in SPSS. It has far more power and capabilities than
SPSS macros. You can accomplish the same tasks that you would with a
macro, but can do much more, including the creation of dialog boxes, the
customization of output in the Viewer window, or the writing out of
selected portions of output to a separate file for use in other programs.
Scripts can be set to run automatically or run at user choice.

Unlike syntax programs or macros, scripts are written in a special

language, Sax Basic, that is similar to the macro language in other
programs, such as Visual Basic for Applications. Of course, scripts can
also use SPSS syntax in their definition, and they can be assigned to a
menu choice like macros. Scripts are often somewhat lengthy compared to
standard syntax, so we won’t display an example in this chapter. SPSS
Scripts are covered in the Programming with SPSS Scripts training
course.

Basic SPSS Programming Concepts 2 - 4

SPSS Training

SPSS DATA A substantial portion of this course is devoted to the manipulation of files
DEFINITION and data with SPSS. As such it will be helpful to understand a bit about
SPSS data file definition. More information is available in the Commands
and Program States Appendix in the SPSS Base 10.0 Syntax Reference
Guide (this guide is available on the CD-ROM containing SPSS and can
be copied to your hard drive when SPSS is installed).

To do something in SPSS you need to define a working data file,

possibly transform the data, and then analyze it. The first task is
accomplished with a file definition command. A simple instance might be
the GET command to open an SPSS data file, but a more complex one is
INPUT PROGRAM to define a non-standard data file. In either case, file
definition commands use an input program state to accomplish their job.
In an input program state, SPSS must determine how to read a data file,
what the definition is of a case, when to create a case, and when to create
the data file.

Often these decisions are straightforward for SPSS. When you click
on File...Open..Data, and name a file with an extension of SAV, SPSS
knows how to read the file, that each logical record in the file is to be
written to one row in the Data Editor, and that it should read the whole
file. Or in the simple program below, the execution of the DATA LIST
command (accessed by choosing File...Read Text Data; note that as of
SPSS 10.0 a GET DATA command is pasted instead) tells SPSS that
what follows is a normal file, where each line of data is to be written to a
new row, or case, in the Data Editor, and that the last case should be
written and the file created when the END DATA command is
encountered.

DATA LIST FREE / X.

BEGIN DATA.
123
END DATA.
LIST.

However, even in these simple commands, SPSS enters an input

program state that is more complex than it first appears. In fact, you can
explicitly place SPSS into this state by using the INPUT PROGRAM
command. Thus, the following program is equivalent to the first.

INPUT PROGRAM.
DATA LIST FREE / X.
END INPUT PROGRAM.
BEGIN DATA.
123
END DATA.
LIST.

Basic SPSS Programming Concepts 2 - 5

SPSS Training
The difference is that first, SPSS is explicitly put into the input
program state, and second, SPSS is told when to quit the input program
state (with END INPUT PROGRAM). There is, of course, no reason to use
INPUT PROGRAM in this uncomplicated example, but for complex data,
using an input program may be necessary to successfully read data into
SPSS. We will see more of INPUT PROGRAM in Chapter 4.

The key to understanding and using complex file definition

commands in SPSS is to understand that you, the user, are in charge of
telling SPSS what constitutes a case or row in the Data Editor, when to
create that case, and when to end the input program and create the
working data file. As an illustration, it is possible through an input
program to have SPSS read only a portion of a file rather than first
creating a larger working data file which is then reduced by selecting
certain cases to retain.

SPSS data files must be rectangular (denormalized, in database

terminology). This means that there must be a value for every variable
for every case. Or to put it another way, each row of the Data Editor
defines a case to SPSS. Often, data come in a format that doesn’t match
this layout, and that is one of the most common uses for the input
program capability. In addition, SPSS supplies several predefined
complex file definitions that read common types of non-rectangular files
(we will discuss these in Chapter 3). Changing the case definition of a file
is a common technique to solve a variety of problems.

SPSS As with other programming languages, SPSS programs, whether they be

standard syntax, macros, or scripts, all have several standard constructs
PROGRAMMING
that can be used to do many things. These include the ability to loop, to
CONSTRUCTS create an array of elements, to repeat actions, and to do actions only if
some condition is true. We illustrate these concepts here with standard
syntax; then later you will see these constructs used again in macro0s.

DO IF & END IF A DO IF & END IF structure is used to execute transformations on

subsets of cases based on some logical condition. It is often used to
replace a long series of IF statements. The logical structure of a DO IF
command sequence is:

DO IF (test for condition)

transformations
ELSE IF (test for another condition)
transformations
ELSE IF or ELSE
further transformations
END IF

Basic SPSS Programming Concepts 2 - 6

SPSS Training
The clear advantage is that not all statements are executed for each
case, as is true for a series of IF statements. Consider regression analyses
that found that the relationship between gross domestic product (GDP)
and birth rate (BTHR) is not the same for first- and third-world countries
(which is definitely true). The results of the separate regression analyses
can be applied to a file of first- and third-world countries efficiently with
these commands (where WORLD is the selection variable).

DO IF (WORLD = 1).
COMPUTE BTHR = 10.872 + .0014 *GDP.
ELSE IF (WORLD = 3).
COMPUTE BTHR = 46.148 -.004 * GDP.
END IF.

Although we could have accomplished the same with two IF

commands, the advantage is that the Else If and second Compute
commands are not executed for first-world countries.

DO REPEAT & A DO REPEAT construct allows you to repeat the same group of
END REPEAT transformations on a set of variables, thereby reducing the number of
commands that you must enter. SPSS must still execute the same
number of commands; the efficiency comes for the user, not SPSS. To
illustrate its use, let’s access the 1994 General Social Survey file, stored
in the c:\Train\ProgSynMac directory.

Displaying First, to simplify the instructions in this course, we will request that
Variable Names in variable names (and not the default variable labels) be displayed in
dialog boxes. From within SPSS:
Dialog Boxes
Click Edit..Options
Click the Display Names option button in the Variable Lists
section of the General tab
Click the Alphabetical option button in the Variable Lists
section of the General tab

In order to display SPSS commands in the Viewer window when we

run analyses, we change one of the Viewer options.

Click Viewer tab in the SPSS Options dialog box

Click checkbox beside Display commands in the log
Click OK

Now to read the data.

Click File...Open..Data
Move to the c:\Train\ProgSynMac directory (if necessary)
Double-click on GSS94

There are several questions in the file that ask about whether
national spending on various programs or areas should be increased, stay
the same, or be reduced. Imagine that we wish to compare pairs of

Basic SPSS Programming Concepts 2 - 7

SPSS Training
questions (urban problems and welfare) to see whether or not a
respondent gave the same answer to each. The program in Figure 2.1
accomplishes that task. Open it by

Clicking on File...Open..Syntax (move to

c:\Train\ProgSynMac folder if necessary)
Double-click on CHAPT2

Figure 2.1 DO REPEAT & END REPEAT Program

The Do Repeat structure requires that stand-in variable names be

used to represent a list of variables or constants. The stand-in variables
exist only within the DO REPEAT structure. Between the DO REPEAT
and END REPEAT commands, transformation commands can be used,
referencing the stand-in variables. The PRINT keyword on the END
REPEAT command tells SPSS to list the commands generated by the DO
REPEAT structure. (This is a good idea except when SPSS generates
hundreds of commands.)

The COMPUTE command uses another feature of SPSS syntax, true/

false comparisons. The COMPUTE statement tells SPSS to compare the
values of the elements in FIRST to SECOND, in pairs. When, for
example, NATSPAC is equal to NATENVIR, the test is true and SPSS
returns a “1” to the variable SAME. When the two responses are not
equal, SPSS returns a false, or “0”, to SAME.

To see this in operation

Highlight all the lines from DO REPEAT to LIST, then click on

the Run button

Basic SPSS Programming Concepts 2 - 8

SPSS Training
After SPSS runs the commands the Viewer window opens, as shown
in Figure 2.2. SPSS creates four COMPUTE commands based on the DO
REPEAT structure. Scrolling down through the output from LIST (not
shown) demonstrates that when NATSPAC is not equal to NATENVIR,
DIFF1 is set equal to zero, and when the two responses are equal, DIFF1
is set to 1 (a value of 0 for either of the spending variables is defined as
missing so the COMPUTE is not done).

Figure 2.2 Output from END REPEAT PRINT

Although in this instance little if any work was saved by the use of
DO REPEAT, in many circumstances the savings can be substantial.

LOOP & END The DO REPEAT structure is an iterative construct because SPSS
LOOP iterates over sets of elements to carry out the user instructions. A more
generic form of iteration is provided by the looping facility in SPSS,
represented by the LOOP & END LOOP commands. They can be used to
perform repeated transformations on the same case until a specified
cutoff is reached, which can be defined by an index on the LOOP
command, an IF statement on the END LOOP command, or other
options. By default, the maximum number of loops is 40, defined on the
SET command. Almost any transformation can be used within a loop.

We begin with a very simple loop to illustrate its syntax.

Click on Window...CHAPT2 - SPSS Syntax Editor to return to

the Syntax Editor window
Scroll down to the program shown in Figure 2.3

Basic SPSS Programming Concepts 2 - 9

SPSS Training
Figure 2.3 LOOP & END LOOP Example

On the LOOP command, we tell SPSS to loop five times with the
index clause of #I=1 to 5. This tells SPSS to repeat the COMPUTE
command five times for each person in the GSS file. Usually indices are
increased by one, as in this example, but that is not always the case. Nor
must they begin at 1.

The COMPUTE command itself tells SPSS to add one to the previous
value of Z, which has initially been set to 0 before the loop. The loop then
finishes with the required END LOOP command to tell SPSS the
construct has finished.

A NOTE ABOUT Notice that the program ends with an EXECUTE command. When
running syntax from a Syntax window, SPSS does not immediately
PROGRAM
process transformations by reading the data file. Instead, it stores
EXECUTION transformations in memory and waits until a command is encountered
which forces a pass of the data. This is in comparison to running SPSS
commands from a dialog box, where the command is executed
immediately after the OK button is clicked. The EXECUTE command
forces a pass of the data and executes any preceding transformations.

Highlight the commands from the first COMPUTE to

EXECUTE
Click on the Run button

To see the effect of this program

Basic SPSS Programming Concepts 2 - 10

SPSS Training
Switch to the Data Editor window
Scroll to the last column in the Data View sheet

Figure 2.4 Data Editor with variable Z added

SPSS has added Z to itself plus 1 five times, and since Z initially was
zero, Z is now 5 for every case in the file. To reiterate, the LOOP
command works within a case rather than across cases. We will see many
uses of looping in programs, and the concept of looping will be repeated in
macros and scripts.

SCRATCH The variable #I used to index the loop does not exist in the GSS file. If it
VARIABLES did, we would see it next to Z in the Data Editor. It hasn’t been created by
SPSS because it was declared a scratch variable. This is done by
specifying a variable name that begins with the # character. Scratch
variables are used in transformations or data definition when there is no
reason to retain them in the data file. They cannot be used in procedures.

Basic SPSS Programming Concepts 2 - 11

SPSS Training

VECTOR A vector is a construct used to reference a set of existing variables or

newly created variables with an index. The vector can reference either
string or numeric variables.

Here is how to create a vector from existing variables.

VECTOR SAT = SATCITY TO SATHEALT.

The vector SAT is created from the five questions in the General
Social Survey that ask about a respondent’s satisfaction with various
aspects of his/her life. This vector is not visible in the Data Editor as a
separate variable or set of variables because it is a logical construct from
existing variables. These variables must be contiguous in the file; that is,
they must be located next to each other when viewed in the Data Editor.

Conversely, the syntax

VECTOR X(10).

will create 10 new variables with names from X1 to X10, all

initialized to system-missing. To illustrate this point

Switch to the CHAPT2 - SPSS Syntax Editor window

Click VECTOR X(10)., then click the Run button

Go to the Data Editor (click Goto Data tool )and scroll to

the last columns

Figure 2. 5 Data Editor with Variables X1 to X10

The ten new variables all have system-missing values for each case.

Basic SPSS Programming Concepts 2 - 12

SPSS Training
A more interesting use of a vector is illustrated by the syntax shown
in Figure 2.6. In the DO REPEAT example we compared the value of one
spending variable to another to see if responses were identical. We can
accomplish a similar task with vectors and loops. In this instance we wish
to compare the responses on the variable NATCITY to responses on four
other variables (NATCRIME, NATEDUC, NATRACE, AND NATARMS).
And instead of creating a new variable that indicates whether the
response on NATCITY is identical or not to the other four variables, we
will compute the difference.

Switch to the CHAPT2 - SPSS Syntax Editor window

Scroll down to the program shown in Figure 2.6

Figure 2.6 Program with Vector and Loop to Compute Differences

Between Variables

The VECTOR command creates two new vectors. GROUP is

composed of the five variables from NATCITY to NATARMS (again, they
must be contiguous). DIFF_ has four elements and so creates four new
variables, DIFF_1, DIFF_2, DIFF_3, and DIFF_4. We will place the
difference for each pair in this vector.

The loop increments by 1 but begins at 2 rather than 1. It loops until

5 (or a total of four times) because there are four variables to compare to
NATCITY. On the first pass through the loop, the COMPUTE command
compares NATCITY (the first element of GROUP) to the second element
of GROUP (NATCRIME) and puts the difference in DIFF_(1). And so on
for three other iterations.

Basic SPSS Programming Concepts 2 - 13

SPSS Training
Highlight all the lines from VECTOR GROUP to LIST
Click the Run button

Figure 2.7 List Output Showing DIFF_1 to DIFF_4

Where a case has valid values for the spending variables, we can see
that SPSS created the four new DIFF_ variables measuring the
difference between NATCITY and the other four spending items. It would
be straightforward to create additional COMPUTE statements to
compare all other possible pairs.

We have used the LIST command to check the operation of SPSS in

two of the examples. Checking to see whether syntax has done what you
expected it to do is very important when doing SPSS programming. The
SUMMARIZE command, available through the menus, can do what LIST
does and more, but LIST is easier to type and less complicated when
using syntax.

Basic SPSS Programming Concepts 2 - 14

SPSS Training

ANALYSIS TIP: A set of variables must be contiguous when placing them into a vector.
REORDERING What can you do if that is not true in an existing file? Perhaps the easiest
method to rearrange variables is to use the trick of matching a file to
VARIABLES itself. Figure 2.8 displays syntax from CHAPT2.SPS that illustrates this
technique.

Switch to the CHAPT2 - SPSS Syntax Editor window

Scroll down to the Match Files example

Figure 2. 8 Match Files Program

Normally, MATCH FILES is used to match one file to another.

However, here the working data file (referenced by an asterisk on the
FILE subcommand) is matched to itself because no other file is named.
Usually files are matched using one or more link variables (for example,
ID number), but here it is not necessary since we match one file to itself.
The key portion of the MATCH FILES command is the KEEP
subcommand, where we list the variables we wish to retain in the order
we want them to appear in the Data Editor. The EXECUTE command is
required because MATCH is a transformation, not a procedure.

Highlight the lines from MATCH to EXECUTE

Click the Run button

Switch to the Data Editor window and scroll to the last
columns

Basic SPSS Programming Concepts 2 - 15

SPSS Training
Figure 2.9 Data Editor with Spending Variables moved to the End

The KEEP subcommand named the spending variables last, so they

have been moved to the last columns in the Data Editor.

SUMMARY We reviewed the types of SPSS commands, the three types of SPSS
programs, and briefly reviewed data definition. We then discussed some
of the key programming techniques in SPSS, including the use of loops,
the creation of vectors, the processing of conditional statements (DO IF),
and the creation of repeating elements (DO REPEAT). These techniques
will be used repeatedly in SPSS programming. There are a few other
important programming techniques that you will see in later chapters
when the need arises. We turn in Chapter 3 to the handling of complex
data files.

Basic SPSS Programming Concepts 2 - 16

SPSS Training

Chapter 3 Complex File Types

Topics Introduction
ASCII Data and Records
File Types
Syntax Basics
Data File Structure
Reading a Mixed File
Errors in the Data
Grouped File Type Without Record Information

M
INTRODUCTION ost SPSS users find that the standard DATA LIST command is
sufficient to read the great majority of the data files they
normally encounter. This is because most data files are
rectangular, i.e., they contain the same number of records per case, the
definition of a case is consistent throughout the file, and the variables to
be defined are identical for each case. There are, however, situations in
which the above conditions do not hold. One example is a file at a medical
center with two types of records, one for inpatients and one for
outpatients, with identical variables located in different column positions
on each type of record, and some variables unique to each type of patient.
A standard DATA LIST cannot correctly read such a file and create a
separate case for each patient type.

To handle such a data file and any other that is non-rectangular,

SPSS supplies two general solutions. The first is to use a FILE TYPE
command, which allows the use of predefined file types that read
grouped, mixed, or nested files. The second solution is to allow the user to
take complete control of the process of reading data with an INPUT
PROGRAM command. This chapter discusses the use of file types; the
next chapter covers input programs. A third solution is to read the file
with a standard DATA LIST command, then use other programming
techniques to restructure the file, such as VECTOR and LOOP. We will
also illustrate this approach in subsequent chapters.

Before reviewing the various file types, we need to discuss some

background information.

Complex File Types 3 - 1

SPSS Training

ASCII DATA AND SPSS assumes that complex files are in ASCII format so that they can be
RECORDS read with a DATA LIST command (within the complex file types). Files
that are stored in a spreadsheet or database format cannot be read
directly by SPSS with these techniques. In that case, you have two
options. You can write out an ASCII file from the other software and then
read it into SPSS with a complex file definition. Or you can read the file
into SPSS as you normally would, temporarily creating a working file
with an incorrect format for analysis. You can then use various
programming techniques to restructure the file.

The use of complex file types requires an understanding of a record.

For SPSS, a record refers to a physical line of data in an ASCII data file.
Technically speaking, a record ends with a carriage return and a line feed
(these are invisible to users in most software). In practice, if you open a
data file in a text editor, such as Notepad, each line will correspond to a
record in the data file. However, this is not always the case in word
processing software that wraps lengthy lines, so be careful when dealing
with a file for which you do not have a codebook that lists record length.

It is common to have several records for each case you plan to create
in the final SPSS data file or to have several cases on one physical record.
Understanding what constitutes a record and what the case definition
should be in the final SPSS file is part of the art of successful data input
programming.

In general, ASCII text data files can be in either fixed or delimited

format. Because complex file types must be able to locate case and/or
record variables, though, complex data should be stored in a fixed-format
ASCII text file.

FILE TYPES The three available file types within the FILE TYPE command are:

Grouped: This is a file in which all records for a single case

are located physically together. Each case usually has one
record of each type. Each record should have a case
identification variable. This type of file is often identical to a
standard rectangular file, the difference being that a grouped
file type allows additional checking for errors because of
missing and out-of-sequence records, since SPSS normally
assumes that the records are in the same sequence within
each case.

Mixed: This is a file in which each record type defines a case.

Some information may be the same for all record types but
can be recorded in different locations. Other information may
be recorded only for specific record types. Not all record types
need be defined, so this is often a very efficient method to
read only part of a data file.

Complex File Types 3 - 2

SPSS Training
Nested: This is a file in which the record types are related to
each other hierarchically. An example is a file containing
school records and student records, where all the students
attending one school have their records placed together after
the school record. Usually the lowest level of the hierarchy,
the student in this example, defines a case. Information from
the higher-level records, perhaps overall GPA at the school, is
usually spread to the lower-level record when the case is
defined. All record types that form a complete set should be
physically grouped together, with an optional case identifier
on each record. It is worth noting that record types can be
skipped when reading the data, resulting in the creation of
cases at a higher level in the hierarchy.

SYNTAX BASICS Complex file type programs are begun by the command FILE TYPE and
closed with the command END FILE TYPE. These two commands enclose
all definitional statements. One of the three keywords GROUPED,
MIXED, or NESTED must be placed on the FILE TYPE command. The
commands that define the data must include at least one RECORD TYPE
and one DATA LIST command, though it is common to have several. One
set of RECORD TYPE and DATA LIST commands is used to define each
type of record in any data file. The definition of a case, again, depends
upon which FILE TYPE is specified.

The RECORD subcommand is required and names the column

location of the record identification information and, optionally, the
variable that will store this information. A CASE subcommand is also
available (and required for a grouped file) that specifies the name and
location of the case identification information.

This syntax illustrates the basic structure,

FILE TYPE (Grouped, Mixed, or Nested) FILE = 'Your File' /

RECORD = RECID 4 CASE = ID 1-3.
RECORD TYPE 1.
DATA LIST / your variables and column locations here.
RECORD TYPE 2.
DATA LIST / more variables here.
etc.
END FILE TYPE.

All three file types have subcommands available that warn the user
when records and cases are encountered that don't meet the definitions of
the file type, record, and case. This warning can include situations when
records are missing.

After the FILE TYPE--END FILE TYPE structure is processed, a

rectangular active file is created, no matter the original structure of the
raw data file.

Complex File Types 3 - 3

SPSS Training

DATA FILE To further illustrate the three types of files, we display samples of data
STRUCTURE files that can be read as grouped, mixed, or nested files.

GROUPED DATA A grouped data file often looks identical or very similar to a standard
rectangular data file. However, a grouped file often has one or more of the
following problems:

1. A different number of records for each case

2. Records out of order
3. Records with the wrong record number
4. Duplicate records

These situations all mean that SPSS will not read the file
successfully with a simple DATA LIST command. The structure of a
simple grouped data file is shown in Figure 3.1.

Figure 3.1 Grouped Data File for Hospital Patients

These data are from a hospital and contain information on tests and
procedures administered to each patient. Each patient’s data begins with
a record that lists identifying information. The second and subsequent
records include information on a test that was given, the date of the test,
and the cost. Each record after the first defines a test, but we would like
the case definition to be a patient. The problem is that a different number
of tests is given to each patient, so we cannot specify the same number of
records for each patient.

A CASE subcommand is required on the FILE TYPE command in

addition to the RECORD subcommand. Records with a missing or
incorrect case identification information cannot be corrected and placed
with the correct case, but SPSS will warn you about the problem.

All defined variable names for a grouped file must be unique because
multiple records will be put together to form one case. By default all

Complex File Types 3 - 4

SPSS Training
instances of missing, duplicate, out of range (called "wild") or out-of-order
records will result in warnings from SPSS.

MIXED DATA A mixed raw data file looks quite different than a rectangular data file.
Again, a MIXED file type is used when each record type defines a
separate case (though not all record types need be defined). Figure 3.2
depicts a portion of the file MIXED.DAT that contains job information on
employees from a large company.

Figure 3.2 Mixed Data File For Employees

Columns 2-4 contain an identification number for each employee.

This is necessary for the company but not important for a FILE TYPE
MIXED definition. Column 6 contains the required record identification
information, in this case either a 1 or 2 (there is also a record number 3
not shown). The company had three separate record-keeping systems for
employee information for each division. The data from all three have
recently been placed in one file for reporting purposes.

A standard DATA LIST cannot be used to read this file because some
of the same information is in a different location for each record type.
Salary is recorded in a different location for each record type, and other
variables are not recorded in each system. We will attempt to read this
file in the first example.

Complex File Types 3 - 5

SPSS Training

NESTED DATA A nested data file also looks quite different than a rectangular data file.
A FILE TYPE NESTED command is used when the records in a file are
hierarchically related. One example is a file with two types of records,
one for each department in a company, and one for each of the employees
in that department. All the employee records for one department are
placed consecutively together, after the record for the department in
which they are located and before the next department record.

The variable names on all the records must be unique because one
record of each type will be grouped together to form a case. Since not all
record types need be mentioned on the RECORD TYPE command, it is
possible to define a case at a higher level in the hierarchy, e.g., a
department rather than an employee. In fact, a case can be defined at any
level in the hierarchy of record types. Figure 3.3 depicts a nested data file
for a school district.

Figure 3.3 Nested Data File for School District

The file contains data on the performance of high school students

organized by homeroom and school. It contains three types of records:

Record 1: The high school record contains identifying

information on the school plus the school's overall GPA, SAT
verbal and SAT math scores.

Record 2: The homeroom record contains identifying

information on the homeroom plus the average GPA for all
students in the homeroom.

Complex File Types 3 - 6

SPSS Training
Record 3: The student record contains identifying information
including the student's sex and academic track, plus GPA,
SAT verbal and SAT math scores.

There is only one variable in common for all three records, the record
identifying information in the first column.

The data file has no case identifying information, which is typical of

real-life situations. That isn't a problem since SPSS simply stores the
higher-level record information, spreading it to each record 3 when it
creates a case for each student, retaining this information until another
record type 1 is encountered. SPSS can still successfully create the cases
even when intermediate-level records are missing (a homeroom record, in
this instance).

READING A We will read the employee mixed data file from Figure 3.2 (named
MIXED FILE MIXED.DAT) into SPSS and create a rectangular file with a case for each
employee.

Here is a codebook table for the three types of employee record

systems.

RECORD 1 RECORD 2 RECORD 3

VARIABLE LOCATION LOCATION LOCATION

ID 2-4 2-4 2-4

RECORD ID 6 6 6
AGE 8-9 8-9 8-9
SEX 11 11 11
SALARY 13-17 22-26 13-17
TENURE 19-20 19-20 19-20
JOBCODE 22 17 not recorded
LOCATION 24 15 22
JOBRATE 26 13 not recorded

Not only are variables like SALARY recorded in different locations,

but two variables, JOBCODE and JOBRATE, were not recorded on record
3, which was the oldest record-keeping system. When the same variable,
e.g., AGE, is defined by more than one record type, the format type and
length should be the same on all records. SPSS uses the first appearance
of the variable for the active file dictionary.

The appropriate commands to read this file are included in Figure 3.4
and are in the file MIXED.SPS.

Click on File..Open..Syntax
If necessary, switch directories to c:\Train\ProgSynMac
Double-click on MIXED

Complex File Types 3 - 7

SPSS Training
Figure 3.4 Mixed File Type Program

The command FILE TYPE begins the file definition and puts SPSS
into an input program state. The MIXED subcommand tells SPSS that
this is a mixed data file. The data file is named here, not on the DATA
LIST commands that follow. The only other required subcommand is
RECORD to specify the record identification variable. The equal sign is
not required following RECORD or FILE. The record variable is in
column 6 and will be named SYSTEM. For the employee data it is
important to retain information that tells us under what record system
the data were created because of duplicate IDs under each system; often,
though, the record variable doesn’t need to be retained in the final file. In
that case, it can be declared a scratch variable by beginning its name
with “#”.

Each employee data system, corresponding to a type of record in the

data file, gets its own RECORD TYPE command. The value specified on
the command (1, 2, or 3) refers to an actual value in the file MIXED.DAT
in the record identification position, here column 6 (refer to Figure 3.2).
The DATA LIST command following each RECORD TYPE command
defines the variables for that record type. Notice how SALARY is in
columns 13-17 for record types 1 and 3 but in columns 22-26 for record
type 2.

An optional subcommand on the FILE TYPE command is WILD,

which tells SPSS to issue a warning when it encounters undefined record
types in the data file. The default is NOWARN, so SPSS simply skips all
record types not mentioned and does not display warning messages.

Complex File Types 3 - 8

SPSS Training
The input program state ends with the END FILE TYPE command,
which is followed by labeling commands and then a Frequencies
command. FILE TYPE--END FILE TYPE are not procedures and do not
cause the raw data file to be read, so they must be followed by either an
EXECUTE command or another procedure.

To run all the syntax

Click Run..All

SPSS displays the commands in the Viewer window (not shown) and
then the frequency table for SYSTEM. We can see that there are 212
employees in the file, created from 212 records, and that there are 52 of
record type 1, 122 of record type 2, and 38 of record type 3 in the data file.
All the information on, for example, salary, has now been placed in one
column despite its two different locations in the data (if you wish, switch
to the Data Editor to verify this).

Figure 3.5 Frequency Table for System

Complex File Types 3 - 9

SPSS Training

ERRORS IN THE We will illustrate what occurs with undefined record types by once again
DATA reading the file MIXED.DAT. Because warning messages are turned off
by default, we were unaware that there are in fact 213 records, or
employees, in the file. However, the 213th case has an error in its record
type, as shown in Figure 3.6.
Its record type should be a “3” but is instead a “4.”

Figure 3.6 Error in MIXED.DAT

SPSS skipped this employee’s record because it was not defined on a

RECORD TYPE command, but didn’t warn us. Let’s tell SPSS to do that,
then reread the file.

Click Window..MIXED - SPSS Syntax Editor to return to the

syntax file
Add the subcommand /WILD=WARN to the end of the FILE
TYPE command, but before the period (.)

Figure 3.7 Modified File Type Command to Add Warnings From SPSS

Complex File Types 3 - 10

SPSS Training
After you have carefully added this subcommand, rerun all the
commands by

Clicking Run..All

When SPSS switches to the Viewer, you will now see a warning
message and a note in the log under the FREQUENCIES command, as
shown in Figure 3.8. The warning message is clear, telling us that the
record type (4) was ignored when building the file. You can verify this by
looking at the frequencies output, which lists only 212 cases.

The exact position of the problem is noted in the message that begins
with “Command line:”. The critical information is that it was on case 213
that SPSS encountered an unknown record ID, whose value is 4. SPSS
also conveniently lists the actual line of data from the file MIXED.DAT
for reference. Obviously, warnings can be helpful in finding and fixing
errors in data entry or definition.

Figure 3.8 Warning Messages for Undefined Record Type 4

It is now possible to use the file created via the FILE TYPE MIXED
command to report on either the total group of employees, or differences
between employees across record-keeping systems.

Complex File Types 3 - 11

SPSS Training

Analysis Tip If you plan to read a file with known errors, you might think that you
want to be warned every time there is a problem defining an SPSS data
file. However, this is not always the case. If it is a large data file with
many errors, SPSS could possibly generate hundreds, even thousands, of
warning messages. It is unlikely that you will care to scroll through all
that output. In recognition of this, the maximum number of warnings
SPSS will display has been set relatively low, to a value of 10. When you
do want to see more warnings, use the SET command with this syntax:

SET MXWARNS = 100. (or to whatever value is appropriate)

GROUPED FILE Reading either a grouped or nested file into SPSS isn’t much different
TYPE WITHOUT than reading a mixed file in terms of the syntax. However, one situation
that causes problems yet is still relatively common, and therefore worth
RECORD exploring, is when you wish to use FILE TYPE GROUPED but don’t have
INFORMATION a record identification variable. This is fairly common, especially because
any rectangular data file can be read with either a standard DATA LIST
or via a FILE TYPE GROUPED format. The advantage of the latter is
that SPSS will fix any problems with out-of-order records and read the
file correctly if there are missing records. However, a record type variable
is needed in each instance, and you may not have created one for what
you knew was a standard rectangular file format.

Our example here is a little more complicated. Figure 3.9 displays a

small data file that stores information on students in a statistics class
and their scores on each of three assignments. There is an identification
variable for the student in column 2, but no numeric record identifier to
tell SPSS that the first record has information on the first quiz, the
second record on the first homework assignment, and the third record on
the first test. Moreover, student 2 did not complete the first homework
assignment and so only has two records, which is the real problem.

If this file had been created with one record for each student, and
each assignment in a separate column, it would be a straightforward task
to read it into SPSS.

Figure 3.9 Grouped Data File

Note We should point out that the score type (quiz1, etc.) field can be used as a
record type identifier, although it is a string field. However, we will
ignore this in order to demonstrate another method of reading the file.

Complex File Types 3 - 12

SPSS Training
What we want to do is read this data file and create three cases, one
for each student. We also want to create three variables, one for each type
of assignment, and just as important, we want SPSS to realize that the
second student’s second record has his score for the test, not the
homework, assignment.

There are two methods to approach this problem without using a

more complex INPUT PROGRAM command sequence.

1) Read the data into SPSS, create a record identifier, write the
data back out as an ASCII file, then read it back in using
FILE TYPE GROUPED.

2) Read the data into SPSS, create a record identifier, then

manipulate the data to create the necessary variables.

The latter choice is clearly preferable because it requires fewer passes

of the data file. The file GROUPED.SPS contains the commands to
implement the second method.

Click on File..Open..Syntax (move to c:\Train\ProgSynMac

folder)
Double-click on GROUPED

Figure 3.10 Program to Read a Grouped File

The DATA LIST command reads GROUPED.DAT as if it were a

rectangular file. The interesting techniques in this program are what
happens after the file is read.

Complex File Types 3 - 13

SPSS Training
First, we create a RECORD variable with the LAG function. Initially,
RECORD is set to 1 for each case. Then, when the current case has the
same ID as the previous case, RECORD is set equal to the value of
RECORD for the previous case plus one. When the previous case was a
different student, then this statement is not executed and RECORD
remains at 1, i.e., it is reset to 1 for each student’s first record. This
creates the desired record identification variable. To see this

Highlight the lines from DATA LIST to the first LIST

command
Click on the Run button

Figure 3.11 List Output Showing Record ID

Analysis Tip When creating programs that read or transform data, it is very helpful in
debugging your programs to list out the data after an action or series of
actions is executed. If you don’t do this, then it will be very difficult to
figure out where things went wrong. Following this advice, the LIST
command has been placed at four spots in the program. It is better to err
on the side of excess here.

Our problems are not yet solved, though. Student 2 didn’t complete
the homework, but the value of RECORD for his test1 score is 2, not 3.
The set of IF statements fix this problem. They assign the correct value of
RECORD for each type of assignment, so that student 2’s test1 score will
now be listed as record type 3.

Switch to the GROUPED - SPSS Syntax Editor window

Highlight the lines from IF to the next LIST command

Click on the Run button

In the Viewer (not shown) we can now see that the value of RECORD
for the second line (or case) for student 2 has been changed to a 3.

Complex File Types 3 - 14

SPSS Training
One task remains, and that is to take these 8 cases in the Data Editor
and turn them into 3 cases. As mentioned above, we could write this file
out and then read it back into SPSS, but that is cumbersome. A better
method is to use the AGGREGATE command. For analysis purposes, it is
best to create a separate variable or column for each assignment. If we
simply aggregate the current working file by student ID, that won’t
happen. Why?

To create these new variables (three in this instance), we can take

advantage of the VECTOR command we saw in Chapter 2. The VECTOR
command creates a new vector, SCORE_, with three elements. The
COMPUTE statement then assigns the value of SCORE for a case to an
element of SCORE_ based on the value of RECORD. In other words, for
the first record type (quiz1), SCORE_1 gets the value of SCORE for that
case, the quiz1 score. The values of SCORE_2 and SCORE_3 for that case
are system-missing. For the second record type (for hmwk1) SCORE_2
gets the value of SCORE for that case, and SCORE_1 and SCORE_3 are
system-missing.

It’s probably easier to see this in action.

Switch to the GROUPED - SPSS Syntax Editor window

Highlight the lines from VECTOR to the next LIST command
Click on the Run button

Figure 3.12 List Output Showing Vector SCORE_

The values for each assignment or test have been placed in separate
variables, with the quiz1 score in SCORE_1, the hmwk1 score in
SCORE_2, and the test1 score in SCORE_3. However, there are still
three cases for each student, with lots of missing data, and all we need is
one case for each student to calculate appropriate statistics.

Complex File Types 3 - 15

SPSS Training
A file with that structure can be created using the AGGREGATE
command, which changes the case base in a file and calculates summary
statistics for the new case base. For example, in a file of customers that
buy five different products, AGGREGATE can create an SPSS data file
where the case base is product (and so there are only five cases), with
information such as the number of customers who bought each product,
the mean age of customers who bought that product, and so forth.

There are three necessary subcommands for AGGREGATE, as shown

in Figure 3.10. The OUTFILE subcommand tells SPSS whether to save
the new data file to disk or to make it the current working file in the Data
Editor. The asterisk means to replace the current working file with the
new one. The BREAK subcommand defines the case base for the new file.
By breaking on the variable ID, which here has only three unique values,
we will create a file with three cases. The aggregated variables
subcommand creates the summary variables in the new file. Its format is

LIST OF NEW VARIABLES = FUNCTION (LIST OF

EXISTING VARIABLES)

where you define new variable names on the left of the expression, an
aggregate function on the right, followed by the same number of existing
variables that will be used to create the new variables. In our example,
we use the MAX (maximum) function to create three variables called
QUIZ1, HMWK1, and TEST1, based on the maximum value of SCORE_1
to SCORE_3 for each ID value. Why does this accomplish our task? Could
we have used a different function?

To finish the program

Highlight the AGGREGATE and LIST commands

Click on the Run button

Figure 3.13 List Output Showing New Assignment Variables

The output from LIST demonstrates that there are only three cases in
the new file, one for each student. Three new variables have been created,
one for each assignment. This makes it easy to calculate statistics for
each assignment. And student 2 has been correctly assigned a missing
score for HMWK1. Since AGGREGATE only creates the new summary

Complex File Types 3 - 16

SPSS Training
variables we define, the variables ASSIGN, SCORE, RECORD, and
SCORE_1 to SCORE_3 are gone, which is fine. You may also want to look
at the Data Editor (not shown) to see the file format.

Analysis Tip Unlike the definition of complex file types, every command in this
program could have been created from the dialog boxes except VECTOR.
As this course is generally concerned with SPSS programming, we
instead worked from a Syntax file. Either approach is acceptable,
although seeing the syntax often helps your understanding, and it
certainly lets you apply the same type of program in the future to another
data file.

SUMMARY We reviewed the types of complex files, their structure, and the SPSS
syntax used to read these data files. We illustrated the use of complex file
types by reading a mixed data file, then discussed how data errors are
handled by SPSS. We then showed how to read a grouped file with an odd
structure and no numeric record type information.

Complex File Types 3 - 17

SPSS Training

Complex File Types 3 - 18

SPSS Training

Chapter 4 Input Programs

Topics Introduction
Syntax Basics
Changing the Case Base of a File
End of Case Processing
End of File Processing
Checking Input Programs
Incomplete Input Programs
Reading Files with Missing Identifiers
When Things Go Wrong

T
INTRODUCTION here are situations where you encounter non-rectangular raw
data files that cannot be read directly with the complex file types
provided by SPSS. For those situations, SPSS offers an input
program facility, as mentioned in Chapter 2, that has the capability to
read essentially any type of ASCII data file. The ability to read a file will
at times depend upon the cleverness of the user, as really odd files may
require creative solutions.

An input program can also be used to create data that match a target
distribution, often for purposes of teaching or illustration. In other words,
an input program can create data from nothing (this is the one time that
SPSS provides a free lunch, so to speak).

For very large files, input programs also offer great efficiencies, even
if the file is a standard rectangular data file. An input program can be
used to select only certain cases as the file is read, saving one pass of the
data. Or it can concatenate raw data files, saving on having to create
SPSS data files of each. And input programs can perform the equivalent
functions of a grouped, mixed, or nested file type, but with added
flexibility.

The user is in charge of case definition when writing an input

program, so careful attention must often be paid to where in the program
stream a case should be created. At times, you may also need to tell SPSS
when to stop reading the data and create a working data file.

Input Programs 4 - 1
SPSS Training

SYNTAX The commands INPUT PROGRAM and END INPUT PROGRAM enclose
COMPONENTS data definition and transformation commands that build cases from input
data. At least one file definition command, such as a DATA LIST, must
be included in the structure. Essentially any transformation commands
can be placed within an input program structure, but no procedures. This
means that you can use COMPUTE, IF, DO IF, REPEATING DATA,
LOOP, or any of the other transformation commands that may help to
create a working data file.

It is very important to understand that SPSS processes the input

program commands on a case-by-case basis. This may be hard to grasp
intuitively, given that an input program creates the definition of a case as
it is executed, but we will illustrate this concept below to provide a
concrete example. Without comprehending this point, input programs
may remain somewhat mysterious in their operation.

EXAMPLE 1: We begin our exploration of input programs with an example that

CHANGING THE includes vectors and looping. Figure 4.1 displays a small data file. This
data file is to be used in an experiment on the effect of a drug on
CASE BASE OF A hypertension. Each person’s data consist of 15 measurements, recorded in
FILE three sets of five numbers. Five different characteristics of each person
were recorded at three time points (hence the three records). Thus for the
first person, the first five numbers are 1, 2, 3, 4, and 5, for the five
measurements at time 1. In this small test file, there are six lines of data
for two persons.

Figure 4.1 Data File with Three Records Per Case

However—and this is the problem with the file structure—we wish to

flip the data because measurements of each characteristic came in sets of
three that logically should be grouped together. Thus, the three
measurements for the first characteristic of the first person are not 1, 2,
and 3, but 1, 9, and 0. That is, they are the first numbers in each record.
The three measurements for the second characteristic are 2, 8, and 0, and
so forth for the remaining three characteristics.

Input Programs 4 - 2
SPSS Training
When we are done, the SPSS data file should look like Figure 4.2. In
the restructured file, the rows have become columns and the columns
have become rows, separately for each person. That is our goal. It is
usually very important when using an input program to literally sketch
out the equivalent of Figure 4.2 to have a clear goal in mind.
Note that typically such files would contain subject and characteristic
identifiers, and additional information. We drop these variables to better
focus on the data reorganization. However, without such identifiers you
cannot tell what is what within the file.

Figure 4.2 The Goal: A Restructured Data File

To open the program that carries out this task

Click on File..Open..Syntax
If necessary, switch directories to c:\Train\ProgSynMac
Double-click on INPUT1

Input Programs 4 - 3
SPSS Training
Figure 4.3 Input Program to Invert a File

The key points of the program are these:

1) Read the file as it currently exists to create fifteen variables for

each person;
2) Use a vector and looping to read the first, sixth, and eleventh
variable for each person;
3) Put these three bits of information into three variables in a
new case;
4) Do this four more times for each person, to create five cases for
the five characteristics being measured of each person.

The case basis can be changed as the file is read because we are still
in an input program structure. We will walk through each command of
this program to show how it operates.

INPUT PROGRAM: This command begins the input program

definition. No other specification is necessary or can be added.

DATA LIST: The command reads the file INPUT1.DAT. This is a

standard DATA LIST command, but because there are three
records per person, a RECORDS subcommand is used to
specify this to SPSS. Fifteen new variables are created, one
for each measurement. They are declared scratch variables so
that they are not retained on the new file that SPSS will
construct. The syntax for the DATA LIST uses a “/” to indicate
that SPSS should read the next record in INPUT1.DAT.

VECTOR: This command creates a vector from the fifteen

variables.

Input Programs 4 - 4
SPSS Training
LOOP: We loop from 1 to 5 (using the scratch variable #I) because
we want to create five new cases for each person.

COMPUTES: Each pass through the loop creates three new

variables, A, B, and C. For the first value of #I, which is 1, A
is equal to the value of #V1. B is equal to #V6, and C is equal
to #V11. It is important that you understand why this
happens and that this operation puts the values of 1, 9, and 0
in these three variables.

At this point, SPSS has 19 variables created, #V1 to #V15, #I,

and A, B, and C.

END CASE: This command tells SPSS to create a case at this

point and pass it on to the commands immediately following
the input program structure, when the data have all been
read. When the case is created, all 16 scratch variables are
ignored, so SPSS creates a case with only A, B, and C, as
desired.

When SPSS first encounters the END LOOP command, it resets the
value of #I to 2 and passes through the three COMPUTE statements and
the END CASE command again with that value. Recall from Chapter 3
that looping occurs within a case in SPSS, not across cases. Since the
initial DATA LIST statement essentially created a (temporary) case from
each set of three records, the five passes through the loop create five new
cases from these three records, or five cases per person.

When looping is done for the first person, SPSS goes on to read the
next set of three records and do the same for this person, and this will
continue until the end of the data file is encountered.

END INPUT PROGRAM: When the physical end of the file

INPUT1.DAT is encountered, SPSS then writes out a final
version of a working data file to the Data Editor, with only
the variables A, B, and C. However, an input program is a
transformation and so is not executed until SPSS encounters
a procedure or the EXECUTE command. Here, we force
execution with the LIST command (an SPSS procedure).

Let’s run the complete program.

Highlight all the lines in the program.

Click the Run Button

The commands are echoed in the Viewer, and the output from LIST
shows that the file has been inverted as desired. The variables A, B, and
C for the first case have values of 1, 9, and 0.

Input Programs 4 - 5
SPSS Training
Figure 4.4 List Output Showing New File Structure

Given the file restructuring capabilities of SPSS, we certainly could

have read the data file without an input program, then changed the case
base with such commands as AGGREGATE afterwards. But an input
program is more efficient than such techniques, especially as a file grows
larger. When you have the choice between these two approaches, an input
program is usually the better choice.

Input Programs 4 - 6
SPSS Training

END OF CASE A basic problem that many people have when writing input programs is
PROCESSING how to properly handle end of case and end of file processing. We can use
the program INPUT1.SPS to illustrate the complexities of these
decisions.

In Figure 4.5, the program has been slightly modified by switching

the order of the END CASE and END LOOP statements.

Figure 4.5 Program With End Loop Before End Case

What type of data file will this structure create? How many cases will
be created? Why?

The answer is shown in Figure 4.6.

Input Programs 4 - 7
SPSS Training
Figure 4.6 New SPSS Data File

What happens is that the new program still tells SPSS to loop five
times within a case, but the second loop replaces the values of A, B, and C
with new values, the third does the same, and so forth. After the fifth and
last loop, the values of A, B, and C for the first case are 5, 5, and 0—the
values for the last number in each of the three records. At this point, we
then tell SPSS to create a case. Since we effectively didn’t change the
case basis from what the DATA LIST command created, we end up with
only two cases, one for each person. This is not what we intended, and it
illustrates the importance of correctly placing the END CASE command.

Fortunately, it’s relatively easy to experiment in this fashion and

usually straightforward to see the effect of command placement.

Input Programs 4 - 8
SPSS Training

END OF FILE When you write an input program, you can control all aspects of case and
PROCESSING file creation. Often, though, file creation is determined by SPSS
automatically when it runs out of raw data after reading to the end of an
ASCII data file. At that point, it creates the working data file. However,
you can control this decision, and we illustrate this capability by
modifying INPUT1.SPS slightly.

Click Window...INPUT1 - SPSS Syntax Editor

Add a blank line after END LOOP
Type the command END FILE. in this line
Make sure END CASE precedes END LOOP

The syntax editor should appear as Figure 4.7

Figure 4.7 Modified Program with End File Command Added

At first glance, it might appear that this END FILE command is

redundant. After all, it occurs after both the END CASE and END LOOP
commands. What effect might this command have? Run the program by

Highlighting all the lines

Clicking on the Run Button

The output from LIST says that SPSS created five cases, not ten.
Inspection of the output shows that only data from the first person in the
file was used to create the working data file. The program operated
correctly for the first person, creating five rows with three variables, but
it dropped the second person (and it would have dropped any subsequent
records).

Input Programs 4 - 9
SPSS Training
The END FILE command was encountered before END INPUT
PROGRAM. This means that SPSS passed the decision of when to finish
reading the data file to the user. Our program operates as follows:

1) Read the first three records;

2) Loop five times for the first person and create five cases with
three variables each;
3) Don’t read the next set of three records. Instead, stop reading
the data file and write out a working file.

Figure 4.8 List Output Showing Data From Only One Person

Like END CASE, END FILE takes effect immediately, which is why
it produces this result. It also has this effect because SPSS processes files
by case. Logically, it is true that the end of file comes just after the end of
the looping, but in SPSS programming the entire input program is read
and processed for each case before the next case is read and defined.
Thus, the END FILE command is encountered for the first case, and
SPSS stops reading INPUT1.DAT.

CHECKING INPUT In Chapter 3 we strongly advised you to place LIST commands liberally
PROGRAMS throughout your programs when creating and debugging them.
Nevertheless, an input program is a transformation, and so procedures
cannot be placed within its structure.

This makes it imperative to work with small but representative files

when writing an input program, as we did in the first example. The
larger the file, the more that can go wrong and the harder to figure out
what went wrong.

Input Programs 4 - 10
SPSS Training

INCOMPLETE There are two types of mistakes to make in writing input programs. The
INPUT first is a logical mistake, such as placing the END CASE command in the
wrong location. The second, just as common, is to make a typing error
PROGRAMS
when creating commands. In either instance, strange things can happen
that are directly related to the fact that an error occurred in the middle of
an input program structure.

When SPSS is in the input program state, it hasn’t yet created a

working data file. Moreover, SPSS will be in the middle of various
definitions, which are not necessarily reset when the program fails. In
other words, when there is an error in an input program, SPSS will often
remain in the input program state. This isn’t good, because when you fix
an error and rerun your program, SPSS will become confused when it
sees an INPUT PROGRAM command when it is already in that state.

To illustrate this problem, we will make the most common type of

syntax error and see what it causes.

Click Window..INPUT1 - SPSS Syntax Editor

Remove the line END FILE.
Remove the period after the END CASE command
Highlight all the lines
Click on the Run button
If necessary, switch to the Viewer window

When this program is run, SPSS understandably complains, although

it doesn’t actually state that the END CASE command is missing a
period. Instead, it becomes confused about the text “END LOOP” which it
thinks the user meant to add to the END CASE command (which has no
subcommands).

Figure 4.9 Errors Created in an Input Program

Input Programs 4 - 11
SPSS Training
Notice that the second error, on END INPUT PROGRAM, explicitly
states that there is an error created while there is an unclosed complex
file structure. If you take a look at the Data Editor (not shown), you will
see that it is empty, although it does have the three variables A, B, and C
created.

Now let’s fix the error and run the program again.

Click Window...INPUT1 - SPSS Syntax Editor

Add a period after the END CASE command
Highlight all the lines
Click on the Run Button
If necessary, switch to the Viewer

In Figure 4.10, you can see what occurred. After processing the first
command, INPUT PROGRAM, SPSS complained with an error that such
a command is not allowed in the set of file definition commands, i.e.,
those commands that normally follow an INPUT PROGRAM command.

Figure 4.10 Errors Created When Rerunning the Input Program

The remainder of the program is not correctly processed, so once

again the Data Editor is empty of cases.

Fortunately, there is an easy solution to this problem, which SPSS

mentioned in Error Message 101: Use the NEW FILE command. This
command clears the working data file and removes any unclosed loops or
file structures from memory. Let’s add it to the program.

Input Programs 4 - 12
SPSS Training
Click Window...INPUT1 - SPSS Syntax Editor
Add the command NEW FILE. on its own line before INPUT
PROGRAM (Alternatively, click File..New..Data)

Your screen should appear similar to Figure 4.11.

Figure 4.11 New File Command Added

Highlight all the lines

Click on the Run Button

The program has now worked flawlessly with no errors. The output
from LIST shows the correct file structure of 10 cases. Figure 4.12
provides a look at the file in the Data Editor. The NEW FILE command is
so important that, until you become accomplished at writing input
programs, and whenever you first start writing such a program, we
recommend you include it as the first command and run it each time
along with the other commands.

Input Programs 4 - 13
SPSS Training
Figure 4.12 Data Editor With Five Cases Per Person

EXAMPLE 2: After beginning with a relatively uncomplicated example, we next

attempt to read a deceptively simple file. This example will illustrate
READING FILES several features of input programming, including the ability to reread
WITH MISSING lines of data.
IDENTIFIERS
Occasionally you may encounter files which require special treatment
because they have no case and/or record identifiers on every record.
Figure 4.13 is an example of such a data file, recording order information
for a consumer products company. Each customer’s data begins with a
header record that contains information on billing ID, the location of the
customer, and the general product type. In this small data file, a
customer can make up to three orders, represented by the records
following the header record. The order records contain the price and the
type of order, 1, 2, or 3.

Input Programs 4 - 14
SPSS Training
Figure 4.13 INPUT2.DAT File

In this file, there are multiple records per customer, and an unequal
number per customer, since not every customer has made three orders.
The goal is to create an SPSS data file with one case per customer. In
addition, the three order types can come out of order—for the first
customer, the third order appears second—and we also want to record
these correctly in sequence in the SPSS data file. As always, it is
important to have a picture of the goal in mind, so Figure 4.14 displays
the SPSS Data Editor with the target file structure. Each customer,
identified by the billing number, represents one case, and the orders are
now in the correct sequence, so that order number 3 for the first customer
is placed in the variable VALUE3, even though it appeared on the second
record.

Figure 4.14 Target SPSS File Structure

Input Programs 4 - 15
SPSS Training
Would it be possible to read this file with a FILE TYPE GROUPED?
If not, what are we missing? What about FILE TYPE MIXED? Or FILE
TYPE NESTED?

The complications in the file are that the header record has no record
information, and the order records have no ID variable. We must take
this into account when reading the file.

Our task will be made much simpler by using the REREAD

command. To determine the type of record, it is necessary to read it. That
sounds so obvious it’s not worth stating, but here is the difficulty: once
SPSS has read a record, it normally goes to the next record in the data
file. However, that’s no good here, because once we’ve read a record and
determined whether it is a header or order record, we then want to
reread it to store the necessary information in the SPSS data file.
REREAD tells SPSS to do exactly that.

Let’s open the program that accomplishes the task of reading this
data file.

Click on File..Open..Syntax
If necessary, switch directories to c:\Train\ProgSynMac
Double-click on INPUT2

Figure 4.15 displays the program in the Syntax Editor window. The
general method that will read this file is to:

1) Read a field from each record to test for record type.

2) If the record is a header record, build a case with the END

CASE command from the information currently stored. In
other words, when a header record is encountered, we know a
new customer’s records are being read, so we create a case for
the previous customer from the records just above this header
record. And if the record is the last record in the file, build a
case and end the file with the END FILE command.

3) Read the header record with the appropriate DATA LIST.

4) If the record is not a header record, read the order type and
cost as scratch variables so they are not retained in the final
file.

5) Store the results into a vector created earlier in the program.

Input Programs 4 - 16
SPSS Training
Figure 4.15 INPUT2.SPS File

This logic is displayed graphically in Figure 4.16. As before, we will

work through the program one line at a time, explaining the logic behind
each command and the structure of the syntax.

Figure 4.16 Flow Chart For INPUT2.SPS

Input Programs 4 - 17
SPSS Training
The program begins with the INPUT PROGRAM command.

NUMERIC: This command initializes the variable REGION and

the three VALUE variables that will be used to create a
vector for the orders. These variables are initialized to
system-missing, an important feature, because the value of
REGION will be used to determine when the first header is
being read for the first case.

VECTOR: This command defines the vector ORDER from the

three VALUE variables.

DATA LIST: This is the first of three DATA LIST commands. It

defines the file to be read, and it includes the END
subcommand, which defines the variable #LAST (a scratch
variable so it is not retained). #LAST is a logical variable
which takes on only two values. It has a value of 0 except
when the end of file is reached (after the last record), where it
is set to 1. This will allow us to tell SPSS to create the last
case and to end the file. We read only the variable #CHECK
from each record, from column 3. That column has non-
missing information only for header records. We could have
chosen any column that differentiated a header record from
an order record.

DO IF #LAST: This first DO IF checks for the value of #LAST.

Since #LAST is a logical variable, the DO IF is executed when
this condition is true, i.e., when #LAST is equal to 1. And that
occurs when no records remain and the end of file is reached.

END CASE: Notice that this command is indented four spaces. It

is good programming practice to make programs more
readable by indenting lines that fall within a DO IF structure
(or DO REPEAT or LOOP). SPSS commands can be indented
without the period in column 1 when running interactively
from the Syntax Editor, but programs run in the production
facility or with the INCLUDE command should include the
period (or, alternatively, a plus sign or dash) to indent a
command.

The END CASE command, for the last record in the file, tells
SPSS to create a case. We must do this check because the last
record is not followed by another header record, which is the
normal method used to tell SPSS when to build a case.

END FILE: The command tells SPSS to create the working data
file, again because this is the last record.

COMPUTE: The COMPUTE command creates a logical variable

that allows us to check the record type, using the variable
#CHECK from the first DATA LIST. When #CHECK is
system missing, #TESTREC is set to true, or 1. This occurs
for order records. Why? When #CHECK is not system–
missing, #TESTREC is set to false, or zero.

Input Programs 4 - 18
SPSS Training
DO IF #TESTREC=0: This second DO IF is the heart of the
program. It checks to see whether #TESTREC is 0, which
means we are reading a header record, and so must create a
case.

DO IF NOT(SYSMIS(REGION)): The third DO IF, nested in the

second, creates a case when REGION is not system missing.
This is true for every header record except the first, where
REGION was initialized to system missing by the NUMERIC
command. So for all header records except the first, this DO
IF command creates a case.

REREAD: After creating a case, it’s time to get on with building

the next case, beginning with the header record we just read.
Thus, the REREAD command tells SPSS to go back to the
beginning of this record.

DATA LIST: This second DATA LIST rereads the record with a
format appropriate for a header record.

ELSE IF #TESTREC=1: This block handles order records. The

header test variable takes on values of 0 and 1. When it is
equal to 1, that means that #CHECK is system missing, so we
are reading an order record, which has no information in
column 3.

REREAD: Once again, we must tell SPSS to reread the record

because SPSS had to read it once to see whether it was a
header or order record.

DATA LIST: The third DATA LIST command reads the two
variables that are on order records. It uses scratch variables
because this information will either be passed to other
permanent variables or is unnecessary to retain.

COMPUTE: This command plugs the cost of each order into the
correct element of the vector ORDER. For example, if
#TYPE=2, then the second element of ORDER (VALUE2) is
set to the value of #COST for that record.

END IF: This closes the last DO IF.

The program ends, as always, with an END INPUT PROGRAM

command, then a LIST command to force program execution and check
our work.
To see this in operation:

Highlight all the lines in the program

Click on the Run button

When an input program works correctly, it all seems so deceptively

simple. In Figure 4.17 we see the output from LIST. We have created

Input Programs 4 - 19
SPSS Training
three cases as planned, retaining region, billing number, and product
type, and we have placed the costs in the correct VALUE variables. The
second customer has a missing value for VALUE2 because he made no
order of this type.

Figure 4.17 List Output With Three Cases

WHEN THINGS Creating a program such as INPUT2.SPS is not easy, though, without
lots of experience at programming with SPSS or with other programming
GO WRONG
languages. Expect to make several errors when trying to create this
program, which, as noted previously, can be difficult to check because
procedures can’t be placed in the middle of the program to check your
work. One important thing to do is to make sure you are working in an
empty Viewer window, and to empty it out on a regular basis. Otherwise,
as you execute the program, syntax errors from one run of the program
will blend with errors from subsequent runs, and it will be more difficult
to recognize what needs to be fixed.

One error that is sometimes made is to ignore the need to tell SPSS to
stop reading the data file. What makes the decision to use an END FILE
command tricky is that, in general, there is no requirement within an
input program to tell SPSS when to stop reading the file. So in our first
example, no END FILE command was necessary. This is because
normally SPSS reads the last line of data, inputs it as instructed, sees an
end–of–file marker in the raw data file, and writes out a working data file
automatically.

In this program, however, we took control of the process with the

END subcommand on DATA LIST. We did this so we would be able to
create the last case, but SPSS then passed end of file processing to the
user.

SUMMARY This chapter provided a brief look at input programs via the use of a
simple example. We discussed the basics of writing such programs, and
the importance of controlling end of case and end of file processing. We
also discussed how to check and correct input programs. We then
attempted a complex input program that read a data file with missing
record and identification fields.

Input Programs 4 - 20
SPSS Training

Chapter 5 Advanced Data Manipulation

Topics Introduction
Splitting a String Variable on a Delimiter
Reading Multiple Cases on the Same Record
An Existing SPSS Data File with Repeating Data
Print Command for Diagnostics
Practical Example: Consolidating Transactions
Appendix: Identify Missing Values by Case

INTRODUCTION
W
e have discussed how to read various non-rectangular files
in SPSS, using either input programs or complex file types. In
this chapter, we consider some examples of how to further
manipulate data in SPSS, and we discuss a few additional commands
often used in data transformation. The examples were chosen to both
illustrate potentially useful applications and to show the operation of
these commands.

Advanced Data Manipulation 5 - 1

SPSS Training

SPLITTING A Often users encounter ASCII data files that are comma delimited, rather
STRING than in fixed or freefield format. Comma delimited means that the values
of variables are separated by commas. Many database software programs
VARIABLE ON A
create such files. An example of this type of file is displayed in Figure 5.1.
DELIMITER
Figure 5.1 Comma–Delimited Data

SPSS does have an option to automatically read this type of file,

under freefield input, but this example serves to illustrate methods used
when working with string fields. However, there should be ten values per
record, but the first and second records have only 9 (because there are
only 8 commas). Standard freefield input will misread such a data file.
(Note the LIST option on the DATA LIST command and the default Text
Wizard dialog choices (click File..Read Text Data) will read the data
correctly.)
Also, on occasion a list of string values, separated by a delimiter
character, are stored as a single field in a data file. For example, radio
stations listened to, problems encountered with a product, products used
in the last month, etc.
Such files can be read without too much fuss. The method is based on
these steps:

1) Read all of each record as a long string variable;

2) Search through this string for a comma, extracting the text
before this position as a variable;
3) Replace the string with the contents following the comma, and
continue the process;
4) Repeat until there are no more commas, which allows for
variable-length input fields.

To accomplish these tasks we need to utilize not only the VECTOR

and LOOP commands, but also the string transformation capability of
SPSS. The program that illustrates this procedure is named
COMMA.SPS.

Click on File...Open..Syntax
If necessary, switch directories to c:\Train\ProgSynMac
Double-click on COMMA

Advanced Data Manipulation 5 - 2

SPSS Training
Figure 5.2 Comma.sps Syntax File

We are going to read the small dataset from Figure 5.1, which is
included as inline data in the program. This method will easily generalize
to any size file, and can be used to split an existing string variable into
new variables. We now discuss program execution, step by step.

DATA LIST: Each record of data is read in as one long string

variable, 80 characters in width, with the name VAR. Doing
so makes it possible to easily search through all the data and
avoids the problem of missing data between commas. This
trick—reading all the data as a long string—is a very common
technique to read troublesome files.

VECTOR: The command creates a vector to store the actual

variables. 10 elements of SLOTS are created because there
are ten variables for each case. Each is given a format of A5
(alpha field 5 columns wide) because a number of the fields
contain string data.

COMPUTE: For each case, #I is initialized to 0, but note that this

is outside the loop that will be created next.

LOOP: This Loop command has no index value, so it might

appear as if it will not loop at all, or loop forever, but that is
not true. Looping will continue until the maximum number of
loops is reached (40), or until a condition is met on the END
LOOP command.

Advanced Data Manipulation 5 - 3

SPSS Training
The two COMPUTE commands find the next comma and increment
#I. The INDEX function in SPSS as used in the first COMPUTE has this
format:

INDEX(variable,test string)

It returns an integer that indicates the starting position of the first

occurrence of the test string in the variable. It returns 0 if the test string
is not found within the variable.

COMPUTE #COMMA: The INDEX function searches through

VAR for a comma. The syntax places the test string value in
apostrophes (‘,’). It sets #COMMA equal to the position of the
first comma it finds. Thus for the first case and the first
iteration through the loop, the value of #COMMA is 2.

COMPUTE #I: The value of #I is incremented by 1. The reason

will become apparent in a moment.

We next encounter a DO IF–END IF structure. The two COMPUTE

commands within this structure are executed when #COMMA is greater
than 0. When does this occur? Answer: when the value that the INDEX
function returns is greater than 0, which occurs when a comma can be
found in VAR. In other words, the COMPUTE commands are executed
when there is a value to record from VAR.

The first COMPUTE uses the SUBSTR function. SUBSTR has this
format:

SUBSTR(variable,position,length)

It returns the substring of the variable beginning at the position

indicated and running for the length specified.

COMPUTE: SLOTS(#I): Let’s see how this operates for the first
iteration of the loop. #COMMA is equal to 2, so this says to
take a substring of VAR, beginning in column 1, for one
column. This means to read the first column of information (a
“1”) and place it in the first element of SLOTS, because #I =1
(now you can see why we increment #I once each loop). So
now SLOTS1=1 and the first variable for the first case has
been created.

COMPUTE VAR: This command replaces the full 80-character

VAR with a shorter version, taking a substring beginning at
column 3, thus leaving out the “1” and the comma that follows
it for the next iteration.

END LOOP: The looping only ends when #COMMA is equal to

zero. This occurs when the INDEX function can’t find any
more commas, which means we don’t need to read any more of
VAR.

Advanced Data Manipulation 5 - 4

SPSS Training
If you’ve followed the program to this point, it might appear that our
data file reading problem has been solved, but that’s not quite the case. A
line of data does not end with a comma, but this means that the last field
will not be stored by SPSS because the program as written requires a
comma at the end of a field (so that the preceding characters can be
stored in the vector SLOTS). The last field must be handled separately.

COMPUTE SLOTS(#I): This command takes the current value of

VAR and puts it in the element for SLOTS equal to the
current value of #I. Thus if the last bit of data is the tenth
variable, as for the third case, this command places 78 into
SLOTS10.

The LIST command then executes the preceding transformation

statements and checks our work. Before executing the program, let’s run
through one more example. After the program places the value of “1” in
SLOTS1, it does the following:

1) Computes VAR to be equal to

“alpha,23453,,43553,,,sadfs,5435”
2) Loops again, since #COMMA is equal to 2
3) Computes #COMMA=6
4) Computes #I=2
5) Computes SLOTS2= substring of VAR from position 1 to 5 =
“alpha”
6) Recomputes VAR, etc.

To see this in action:

Highlight all the lines from DATA LIST to the first LIST
command
Click the Run tool button

Figure 5. 3 Ten New Variables Created

The output from LIST shows that the values from the raw data have
been correctly placed in the appropriate variables, with missing data (a
blank for string variables) inserted in the right spots. The tenth variable
value is missing for both the first and second cases. To further illustrate
how the program is functioning, switch to the Data Editor window.

Click on the Goto Data tool

Advanced Data Manipulation 5 - 5

SPSS Training
The variable VAR is retained in the file and has the value of 5435 for
the first case, 34 for the second case, and 78 for the third case. It stores
the last valid value in each record because that was the last field that did
not end with a comma.

Figure 5.4 Data Editor With New Variables and VAR

Note that the VAR column is narrowed in the Data Editor so both it
and the SLOT variables are visible.
This file is now a regular SPSS data file and ready to be used for
analysis, but all the variables are strings. To make the variables that are
numbers into numeric variables, we use the NUMBER function. The DO
REPEAT command simply affords an easy means to create a series of
eight COMPUTE statements. The NUMBER function requires two
arguments, the variable to be converted, and a format by which to read
the variable (here F5 for 5 characters, no decimal digits).

Click Window..COMMA - SPSS Syntax Editor

Highlight the commands from DO REPEAT to LIST
Click on the Run tool button

After these commands are executed, the Viewer window displays all
eight COMPUTE commands created from the DO REPEAT (not shown).
Figure 5.5 displays a portion of the output from LIST with the numeric
versions of eight of the SLOTS variables. Why do these variables all have
two decimal digits displayed?

Advanced Data Manipulation 5 - 6

SPSS Training
Figure 5.5 List Output with Numeric Variables

This program can be readily generalized by increasing the width of

the long string variable VAR on the DATA LIST, by using more than one
of these variables if the record length is more than 256 characters (the
maximum width for a long string variable), and by setting the number of
elements for SLOTS as high as necessary to create as many fields as exist
in the data. MXLOOPS may have to be set higher than the default of 40
as well (use SET MXLOOPS= value.).

READING In Chapter 4 we discussed how to rearrange a file and change the

MULTIPLE columns into rows. At times you may be faced with the reverse situation:
CASES ON THE many cases stored in one record. This can occur whenever all information
for one person or organization is stored in one record of data during data
SAME RECORD entry for efficiency’s sake. You may later wish to place each occurrence
within an ID into a separate case for analysis purposes (such as to
calculate aggregate statistics across cases). An example of such a file is in
Figure 5.6.

Advanced Data Manipulation 5 - 7

SPSS Training
Figure 5.6 A File With Multiple Cases Per Record

Each case has identifying information in the first two fields. It then
has up to five fields of valid data, which could be customer orders,
procedures in a hospital, exam scores, and so forth. This type of structure
can be thought of as repeating data, because each case contains repeating
bits of information that are of the same type.

SPSS supplies a predefined command that will read these types of

data. The REPEATING DATA command must be used within an input
program structure, and it will build one new case for each repeating
group. All the repeating groups must contain the same type of
information, but there can be a different number of repeating groups for
each record.

The user must tell SPSS how many repeating groups of data are in
each record, but often files of this type do not include such information, as
in Figure 5.6. This value can be created from the data, though, as we
demonstrate below. The sample program that reads these types of data is
named Repeat1.sps.

Click on File..Open..Syntax (move to c:\Train\ProgSynMac)

Double-click on Repeat1

The method of reading the file is not complicated.

1) Read the data with a standard DATA LIST command;

2) Compute a variable that records the number of repeating data
groups;
3) Use the REPEATING DATA command to create the cases.

You must supply the number of repeating data groups for the
REPEATING DATA command, but there is no penalty in initially
estimating too many of these on the DATA LIST. We illustrate this by
having SPSS read up to six variables for the critical information even
though no more than five appear on any input record.

Advanced Data Manipulation 5 - 8

SPSS Training
Figure 5.7 REPEAT1.SPS Syntax File

Specifically, the program does these tasks:

DATA LIST: Read in the variables ID and GROUP, which appear

once per record, and then read six variables for the repeating
data.

COMPUTE: This command uses the NVALID function, which

returns the number of nonmissing values in a set of variables
for each case. This puts the number of occurrences of
repeating data into the variable TOTNUM.

REPEATING DATA: The command has only three required

subcommands. The STARTS subcommand specifies the
beginning column of the repeating data segments, here
column 6. The OCCURS subcommand tells SPSS how many
repeating data groups are on each record, and the DATA
subcommand tells SPSS the names, locations, and formats (if
necessary) for each variable within the repeating groups.
There can be many variables within each repeating group, but
we’ve kept things simple in this example and have only one.

END INPUT PROGRAM closes the input program structure, and

it is followed by the inline data.

Run the program by:

Advanced Data Manipulation 5 - 9

SPSS Training
Highlighting all the lines
Clicking on the Run Button

Figure 5.8 List Output With All Variables

Twenty one cases have been created by SPSS, for the 21 separate
repeating data values in the input file. There are four repeating groups
for the first record (117, 837, 947, and 594), so SPSS created four cases
from the first input record. Although no input record had six repeating
data groups, our specification of six input variables on the DATA LIST
command didn’t cause any difficulty. SPSS simply created a sixth
variable, V6, that is missing for every case. The critical information, in
addition to ID and GROUP, is the information stored in VARVALUE,
which came from the repeating data.

At this point, we could delete the variables V1 to V6 and TOTNUM

from the file (though the latter might be useful for analysis), and save the
file as an SPSS data file. We then can compute such quantities as the
mean of VARVALUE for the whole file, which would have been very
difficult to calculate with the initial file structure.

A file in this format also allows us to study the data by customer by

aggregating based on ID.

Advanced Data Manipulation 5 - 10

SPSS Training

AN EXISTING What if an SPSS data file has already been created with the same format
SPSS DATA FILE as the ASCII data file in Figure 5.6? That is, what if you are faced with
an SPSS data file that has only six cases, with repeating data on each
WITH case? Such a file is REPEAT.SAV.
REPEATING
DATA Click on File..Open..Data (move to c:\Train\ProgSynMac)
Double–click on REPEAT
Click No when asked to save current data

What is certainly true is that you cannot use the REPEATING DATA
command to restructure the file, because that command, and input
programs in general, operate on raw ASCII data, not SPSS data files.

Figure 5.9 SPSS Data FIle, Repeat.sav, With Repeating Data

One option is to write the data out as an ASCII file and read it back
in as above, but this is cumbersome and time–consuming with large files.
A better alternative is to use SPSS programming to take this existing file
structure and modify it.

To do so we will need the XSAVE command. XSAVE produces an

SPSS data file, just as the SAVE command does. The key difference is
that XSAVE is not a procedure but a transformation; consequently,
XSAVE can be placed within DO IF and LOOP structures. This often
improves processing time by reducing the number of data passes, and
sometimes allows you to do things that are basically impossible with only
the SAVE command.

The program that restructures the file is named Repeat2.sps.

Click on File..Open..Syntax (move to c:\Train\ProgSynMac)

Double-click on Repeat2

Advanced Data Manipulation 5 - 11

SPSS Training
Figure 5.10 Repeat2.sps Syntax File

Not surprisingly, the program relies upon a vector and looping to

accomplish its job. It works in this fashion:

1) Create a vector to store the repeating data

2) Loop as many times as there are repeating data groups
3) Create the variable to store the repeating data group, and if it
is not missing, write a case to the file STACKDAT.SAV,
retaining only the necessary variables.

Specifically, the commands do the following:

VECTOR: The new vector VREPEAT is created from the existing

repeating data variables, here V1 to V6.

LOOP: We loop as many times as there are repeating data

groups. However, requesting more loops than there are
groups will not cause an error or the program to function
improperly.

COMPUTE VARVALUE: On the first loop, the value (117) of the

first element of VREPEAT (V1) for the first record is placed
into VARVALUE.

DO IF: This is executed if the current element of VREPEAT is not

missing. In other words, the statements in the DO IF
structure are executed if there is a valid repeating data value.
This check means that no cases will be created when there is
missing data, even when a record has fewer than 6 repeating
data values.

Advanced Data Manipulation 5 - 12

SPSS Training
XSAVE: The XSAVE command has the same format as SAVE. We
name the output file and tell SPSS which variables to retain.
Users often picture a SAVE command as being applied to the
whole file at once, but if that were true, XSAVE couldn’t write
out the data appropriately here. Instead, both SAVE and
XSAVE write out one case at a time, until all the data from
the current working SPSS file have been read. This means
that XSAVE writes out the first case to STACKDAT.SAV,
whose values are these:

ID GROUP VARPOSN VARVALUE

1 1 1.00 117.00

and waits for input from the next iteration of the loop.

In this manner, SPSS creates 21 cases from the original 6, one for
each valid repeating data group. The idea of looping through a record and
writing out the cases to an SPSS data file with XSAVE is the central idea
of this, and similar, SPSS programs.

The program closes with END IF and END LOOP. We then need to
get the file STACKDAT.SAV to see the new structure.

Highlight all the lines from VECTOR to EXECUTE (do not run
the GET FILE command yet)
Click on the Run tool button

Switch to the Data Editor (click the Goto Data tool )

Figure 5.11 Data Editor Showing REPEAT.SAV With New Variables

Advanced Data Manipulation 5 - 13

SPSS Training
We are first looking at the existing file, not STACKDAT.DAV, so you
can understand better how this program functions. Two new variables
are created, VARPOSN and VARVALUE. The latter is system–missing
for every case because no input record had a sixth valid field of repeating
data. What may be surprising is that VARPOSN is equal to 7, not 6, even
though the LOOP command is designed to iterate six times. What SPSS
does is execute the loop six times, then increment the value of VARPOSN
to 7, and test for the loop ending condition (which is VARPOSN=6). Since
7 is greater than 6, SPSS doesn’t execute another loop, but it has set the
value of the index variable to one higher than the end of the loop
sequence. This is how loops always operate.

Now let’s look at the new file structure.

Click Window..Repeat2 - SPSS Syntax Editor

Highlight the two lines GET FILE and LIST
Click on the Run button

The output from LIST shows that we have created the desired file
structure. There are 21 cases in total, and we have only the variables we
need, unlike the program with the REPEATING DATA command, in
which variables were dropped to create the final file.

Figure 5.12 List Output from STACKDAT.DAV

These examples have illustrated how there is often more than one
way to accomplish the same result in SPSS. If a file is small, it may make
little difference which method is used, but for larger files, efficiency is
often important (see Chapter 8). This example also demonstrates how an

Advanced Data Manipulation 5 - 14

SPSS Training
existing file can be restructured, just as we did in Chapter 4 when we
changed columns into cases.

PRINT In the examples run thus far we frequently made use of the List
command to display values for specified variables in the active data file.
COMMAND FOR
One disadvantage of the List command is that it is a procedure. As a
DIAGNOSTICS result, List cannot be placed within a loop or Do If structure, and would
be of limited value in determining what is happening within such
structures. Also, as a procedure the List command forces a data pass.
Thus if we insert three List commands in a program, three data passes
must be performed, which is inefficient.
To answer these issues, SPSS has a transformation command called
Print. It provides the same general features as List, but is a
transformation. Print can be placed within Do If and loop structures, and
can provide detailed feedback about what occurs. However, be aware that
if a Print command is placed within a loop, then it will print a line of
output during each iteration of the loop for every case in the file. Such
output is extensive for large data files, so care must be taken. Below we
briefly demonstrate the use of the Print command by inserting one into
the program run earlier that read a comma-delimited file.

Click Window..Comma - SPSS Syntax Editor

Insert a blank line before the COMPUTE VAR=… line
Type PRINT / '#Comma= ' #comma (F2) ' Var= ' var
(a50). (all on one line)

Figure 5.13 Comma.sps Program with Print Command Added

Advanced Data Manipulation 5 - 15

SPSS Training
By default the Print command will print to the Viewer window, but it
can be directed to a file. Text can be added to a Print command by
enclosing it in quote marks. We use this feature to identify the variables
whose values will be printed. Values for two variables (#comma and VAR)
will appear. Note that unlike the List command, Print can print values
for scratch variables. Formats can be given for the variables printed.
Here we print only the first 50 characters of the string variable VAR in
order to prevent the line from wrapping.

Click Run..All
Scroll up in the Viewer window to the lines beginning with
#comma=

Figure 5.14 Output from Print Command

The output shows the values for #comma (column position for the
next comma) and VAR (the string containing the data). In each iteration
of the loop we see how the first element (preceding the comma) of VAR is
stripped off, leaving the tail end (beyond the comma) to be processed in
the next step. Now we have access to just what is occurring within the
loop. Recall that the List command, although useful, can display only the
final result. Thus the Print command provides a glimpse into the
workings of loops, Do Ifs, and other transformations and input program
structures. Print is extremely useful when debugging programs
containing such structures or complex transformations.

Advanced Data Manipulation 5 - 16

SPSS Training

PRACTICAL To demonstrate how SPSS data manipulation commands can be

EXAMPLE: combined to answer a practical question, we present a question posed
within the SPSS Training Department. There was interest in exploring
CONSOLIDATING
what combinations of training classes customers take in the course of a
TRANSACTIONS year. However, attendance at an SPSS training class is recorded as a
single sales transaction within the sales database, so a customer who
took five training classes would appear on five sales transaction records.
For the analysis, we needed a single record (case) per training customer,
on which all training classes attended during the year would be recorded.
In this section, the steps taken to reformat the data are detailed. We
should note that additional data cleaning operations were performed
(selecting only training class transactions within the specified time
period; checking that only valid training course codes were present, etc.),
but are not shown here in order to allow us to focus on the data
reorganization aspect of the problem.

We begin with a sales transaction file containing sales for training

classes.

Click File..Open..Data (move to c:\Train\ProgSynMac

directory)
Double click on TrainTransact
Click No if asked to save the contents of the Data Editor

Figure 5.15 Training-Class Sales Transactions

Each record, representing a single training class transaction, contains

a course code and the customer SPSS ID number. Other fields (date,
price, etc.) have been previously removed from the file. The COURSE
variable is a 7-character string. The first character is a city location code.
The second and third characters record the sequence number of the

Advanced Data Manipulation 5 - 17

SPSS Training
course within the city during that year. Finally, the last three characters
(positions 4-6) represent a 3-digit training course topic code number (for
example 108 is “The Basics: SPSS for Windows 8.0”).

The goal is to reorganize the data so each customer appears in a

single record containing a variable for every course, coded 1 if the
customer took the course, and 0 otherwise. This target data structure is
shown below.

Figure 5.16 Training Course Data Organized by Customer ID

Now it is relatively easy to examine whether taking one course is

related to another. It should be noted that if sequence or city information
were deemed important, then this structure would not be adequate. This
serves to reinforce the notion that no single data structure is convenient
for all possible analyses, and the ability to restructure data, whether
within a database or through a program like SPSS, is important.

So how do we go get to this data structure? The steps are summarized

below.

1) Extract the 3-digit training course topic number from the

COURSE variable and recode it to an integer value within a
contiguous range (1-30).
2) Create a vector of course variables and for each training course
transaction, set the appropriate course variable to 1.
3) Aggregate the course transaction file to the customer ID level.

Click File..Open..Syntax (move to c:\Train\ProgSynMac)

Double click on TransactionAgg
Scroll down to the GET command

Advanced Data Manipulation 5 - 18

SPSS Training
Figure 5.17 TransactionAgg.sps Syntax File (Beginning)

The first task is to obtain the 3-digit training course topic code from
the COURSE variable.

STRING: We create a new string variable (SCODE) three

characters long (A3). Since SPSS assumes newly created
variables are numeric, we must declare SCODE to be a string
with the STRING command before using it in the COMPUTE
command that follows. SCODE will store the training course
topic code.

COMPUTE SCODE: The SUBSTR (substring) function will

extract, from the COURSE variable, a 3-character string,
beginning in position 4 and extending three characters. The
result of this substring extraction is placed in the SCODE
variable. If COURSE is “C22108 “, then SCODE will be “108”.

COMPUTE CODE: Since it will be easier to deal with numeric

course topic codes, that is, it is easier to type 108 than ‘108’,
we convert the string training course topic code stored in
SCODE into a numeric form. This is accomplished with the
NUMBER function and F3 (Fixed number 3 columns long)
represents the format for the numeric variable CODE. The
two COMPUTE commands could be consolidated into a single
command, with the NUMBER function being directly applied
to SUBSTR(COURSE,4,3), but the logic may be more clear
when broken down into single steps.

LIST: Since the transaction file contains thousands of records, we

limit the number of cases displayed in the LIST command
using the CASES subcommand. Only the first 30 cases will
appear in the Viewer window.

Advanced Data Manipulation 5 - 19

SPSS Training
Highlight the five lines from the STRING to the LIST command
Click the Run button

Switch to the Data Editor (click Goto Data tool )

Figure 5.18 Training Course Topic Code as a Separate Field

Switch to the TransactionAgg – SPSS Syntax Editor window

Scroll down to the *Define Record Types. command

Figure 5.19 Next Portion of TransactionAgg.sps Syntax File

Advanced Data Manipulation 5 - 20

SPSS Training
* : Comments can be added to SPSS Syntax programs. Comments
begin with an asterisk in column 1 and the remainder of that
line, and any continuation lines (up to a command
terminating period), are treated as a comment. Comments
can also begin with /* and terminate with */, and such
comments can be inserted within a command line.

RECODE: The RECODE statement converts the training course

topic 3-digit codes into integer values from 1 to 30, which will
be used as vector indices. These integer values are stored in a
variable called RECORD (short for record type). Care must be
taken to note which original training course topic code is
mapped into which integer code. In this program they are
recorded in a VALUE LABELS command just above the
RECODE. Note that an AUTORECODE could be used in
place of the RECODE here, but the output from the
AUTORECODE, or the value labels from the new variable,
would then need to be examined to determine the mapping of
training course topic codes to the numbers 1 through 30.

VECTOR: A vector named COURSE is defined and its range set

to 30, since there are 30 course topics used in the RECODE.
Setting it too high would simply mean that there would be a
few extra variables with all system-missing values; setting it
too low would produce error messages since a vector subscript
would be out of range.

COMPUTE: Since there is a single training class transaction per

case, there is no need for a loop. Rather, the variable from the
COURSE vector in the RECORD position is set equal to one.
For example, the first case had a training course topic code
(CODE) of 108. The RECODE command would assign it a
value of 4 for the RECORD variable, so the fourth COURSE
vector variable (COURSE(RECORD)) is assigned a value of 1.

Thus as each training transaction is processed, a single variable in

the COURSE vector is set equal to one, while the others remain system
missing. We see this below.

Highlight from the RECODE to the LIST command

Click the Run button

Switch to the Data Editor (click Goto Data tool )

Advanced Data Manipulation 5 - 21

SPSS Training
Figure 5.20 Assigning Each Training Course to a Variable

Consistent with the RECODE command, we see that sales

transactions of training course topic code 108 (The Basics: SPSS for
Windows 8.0) now have a code of 1 in the COURSE4 variable (the fourth
variable in the COURSE vector).

We face the now familiar problem of information from a single

customer being spread across multiple records. As before, we will use the
Aggregate procedure to consolidate all the training transactions for a
customer into a single case.

Switch to the TransactionAgg – SPSS Syntax Editor window

Scroll down to the AGGREGATE command

Figure 5.21 Last Portion of TransactionAgg.sps Syntax File

Advanced Data Manipulation 5 - 22

SPSS Training
The AGGREGATE command does most of the work here.

AGGREGATE: The AGGREGATE command will produce a file

containing one case for each customer (/BREAK = SPSS_ID)
and it will replace the current file in the Data Editor window
(/OUTFILE = *). The variable names in the aggregate file (i.e.
Basics61, DE, …, Amos) are chosen to describe the relevant
SPSS training products. These will be based on the
COURSE1 to COURSE30 variables in the current file. Notice
that the FIRST (first valid value) aggregate summary
function is used; it will return the first valid value within the
group formed by the BREAK variable’s value. For a course,
this value will be 1 if the customer took the course and system
missing otherwise.

FORMAT: This changes the display and writing formats for the
course variables to be as fixed numbers one column wide.
Recall the default would be as a fixed number eight columns
wide and with two decimals.

RECODE: If a training course were not taken by a customer we

want the appropriate course variable’s value to be 0, not
system missing. This is because cases with system-missing
values are excluded from many SPSS procedures (for
example, Crosstabs). The RECODE command, which replaces
system missing with 0 for the course variables, accomplishes
this.

Highlight from the AGGREGATE to the LIST command

Click Run..Selection
Switch to the Data Editor (click the Goto Data tool )

Figure 5.22 Training Course Transactions Aggregated to Customer

Level

Advanced Data Manipulation 5 - 23

SPSS Training
We now see, on an individual customer basis, which training courses
were taken. Within SPSS we could now use Crosstabs, Tables, or the
Categories procedure Homals, to explore training course patterns.
Alternatively, Clementine contains dedicated procedures (rule association
analysis) to explore such data.

SUMMARY We discussed four programs that perform advanced data manipulation:

reading comma–delimited files, showing two ways to read repeating data
on one record, and converting transaction level data to customer level
data (performing complex file manipulation to create a missing data
report is in the Appendix). These few examples should provide some
indication of the power of the SPSS transformation language.

Advanced Data Manipulation 5 - 24

SPSS Training

APPENDIX: SPSS via the Frequencies procedure can easily provide output to note the
IDENTIFYING categories and amount of missing information for each variable in a file.
What if, instead, you wished to know which variables were missing for
MISSING each case, and to automatically produce a report of this information? To
VARIABLES BY see an example of this type of analysis, examine Figure 5.23.
CASE
In this small file of 7 cases and 10 variables, we can see that
variables X7 and X9 are missing for case 1, X4 and X10 for case 2, and so
forth. This type of information can be very handy in searching for missing
data patterns and when doing multivariate analysis to understand what
is causing some cases to be dropped from an analysis.

Figure 5.23 Missing Data Report by Variable

Producing such a report from an SPSS data file, though, is a bit of a

challenge. It will not require us to learn any new commands (except
possibly AUTORECODE), but it will require us to visualize the file
structure and the intermediate steps necessary to reach our goal.
Programming this problem is an excellent example of the capability of
SPSS in the hands of a knowledgeable user.

To create the report in Figure 5.23, we need to have a file that has
one row for every case, which is already true in a standard SPSS data
file, so that certainly doesn’t sound difficult. But the tricky part is that
the best way to create the output in Figure 5.23 is to have a series of
variables—as many as in the original file—that are all strings, with the
values of these strings the variable names with missing data from the
original file.

To make this clear, we’ve displayed the target file in the Data Editor
in Figure 5.24. The fact that the values in the columns beginning with
the variable VNAMES1 are left justified indicates that these are string
variables. This file is what we need to create.

Advanced Data Manipulation 5 - 25

SPSS Training
Figure 5.24 SPSS Data File to Create Missing Variable Report

So how do we get there? The answer involves several types of

manipulation, but is summarized below.

1) Save a copy of the original file, but make sure there is an ID

variable.
2) Loop through this same file, creating a series of indices that
identify which variables are missing for each case, and write
this information out, plus the case ID, to a separate file.
3) Continue with the original file, select the first case, flip the file,
and compute a new variable that contains each variable's
position in the original file. Save this as a new file.
4) Get the file saved after looping and create a counter that
increments by 1 for each additional variable missing for each
case, sort the file by the variable position indicator, then
match this file to the file saved in step 3 with a table match.
5) Take the resulting file and compute the new string variables
that will hold the crucial information. Then aggregate this
file, which has as many cases for each original case as there
were missing variables, to a file with one case for each
original case.

We don’t expect that these actions will be intuitively obvious, unless

you are a natural born programmer. We’ll work through the program,
block by block, to see how it functions.

Advanced Data Manipulation 5 - 26

SPSS Training
Click on File..Open..Syntax (move to c:\Train\ProgSynMac)
Double-click on Missvar.sps

The first portion of the program reads in a small dataset with both
numeric and string variables. In this file, blanks represent missing data
in the lines between BEGIN DATA and END DATA. There are eight
numeric variables and two string variables. An ID variable is computed
by using $CASENUM, the SPSS system variable that records case
number.

Figure 5.25 Missvar.sps Syntax File

The program is going to use the VECTOR command, but a vector

must be composed entirely of numeric or string variables, not a mixture.
Accordingly, the AUTORECODE command is used to change variables X9
and X10 into the numeric variables NUM9 and NUM10. AUTORECODE
creates new values beginning with 1, where that value represents the
first value of X9 or X10 in ascending alphabetic order. A blank goes first
in the sort order, so it become a 1, and since the blank is missing data, we
next include the MISSING VALUES command to code 1 as missing for
NUM9 and NUM10.

This file is then saved by dropping the original X9 and X10 (the
strings), renaming NUM9 and NUM10 back to X9 and X10 so we can
retain the original variable names, and using the KEEP subcommand to
place all the variables in consecutive order in the file before ID. This is
done so that they can all be placed in one vector (because variables on a
vector must be contiguous in the data file).

Advanced Data Manipulation 5 - 27

SPSS Training
Run this syntax.

Highlight the lines from DATA LIST to SAVE

Click on the Run button

In the output from AUTORECODE (not shown), we can see that a

blank has been recoded to a 1 for both NUM9 and NUM10.

With the original file saved in correct format, we now need to create a
file that stores information on which variables are missing for which
cases.

Switch to Missvar.sps (click Window..Missvar - SPSS Syntax

Editor)
Scroll down to the next section of the program (as in Figure 5.26)

Figure 5.26 Next portion of Missvar.sps

The critical part of the whole program is next. First we must get the
file DATA.SAV, which has the new file structure we need.

VECTOR: We create a vector with the 10 variables (in a file with

more variables you would simply create a much larger vector).

LOOP: We loop 10 times because there are 10 variables. On each

loop we test with the MISSING function on a DO IF to see if
that vector element for that case is missing.

XSAVE: If the variable is missing, we write a case to the new

SPSS data file (MISSING.SAV), keeping only ID and
VARNUM. ID tells us the case and VARNUM tells us which
variable is missing, so this file is an index of that information.

Advanced Data Manipulation 5 - 28

SPSS Training
The EXECUTE statement forces execution of the transformations,
including XSAVE.

Highlight the lines from GET FILE to EXECUTE

Click on the Run button

The Viewer echoes back the syntax from these commands. The
important thing is the information in MISSING.SAV. Figure 5.27 shows
this file. If you refer to Figure 5.24 you can see that there were two
missing variables for the first case, variables X7 and X9. Therefore,
MISSING.SAV has two rows for case 1, and the VARNUM for each
represents which variables have missing values, 7 and 9. The second case
has missing values for variables 4 and 10, and so forth.

Figure 5.27 MISSING.SAV Data File

We’ve now created half of what we need. In some respects, in fact,

this file is an answer to the original question, but it has as many cases as
there are occurrences of missing data (18 in this instance), and this would
be unwieldy for larger files. Also, MISSING.SAV is not in a format that
makes it easy to create reports summarized by case.

We continue by creating a small file with the variable names.

Switch to the Data Editor (click Goto Data tool )

Advanced Data Manipulation 5 - 29

SPSS Training
The working data file has the information we need, but only if we flip,
or transpose, the file. Transpose creates a new data file in which the rows
and columns in the original data file are transposed so that cases (rows)
become variables and variables (columns) become cases. SPSS
automatically creates new variable names and displays a list of the new
variable names. A new string variable, CASE_LBL, containing the
original variable names is also automatically created.

Figure 5.28 Current Working Data File

We don’t need all the data in this file, just the labeling information,
so we will select out the first case and transpose the file.

Switch to Missvar.sps (click Window..Missvar - SPSS Syntax

Editor)
Highlight the two lines SELECT IF and FLIP
Click the Run button

Switch to the Data Editor (click Goto Data tool )

We now have the file you see in Figure 5.29. This file has the variable
CASE_LBL that contains the original variable names. It also has a
column labeled VAR001 that was the first row of data, which is
unimportant. For purposes of matching this file back to the
MISSING.SAV file, we need to create a case ID (actually a variable ID).
Since the first ten values of CASE_LBL are X1 to X10, we can create such
a variable by using the SPSS system variable $CASENUM as before.

Advanced Data Manipulation 5 - 30

SPSS Training
Figure 5.29 Transposed Data File

Switch to Missvar.sps (click Window..Missvar - SPSS Syntax

Editor)
Highlight COMPUTE and SAVE
Click on the Run button

Notice we called the ID variable VARNUM because it is really an ID

variable for the position of a variable, and we gave it that name because
the same name is used in the MISSING.SAV file.

Switch to the Data Editor (click Goto Data tool )

Compare the working data file in the Data Editor (not shown) to
Figure 5.27 that displays the MISSING.SAV file. We are going to match
the CASE_LBL information from the file VARNAMES.SAV by VARNUM
to each case in MISSING.SAV. In other words, for the first case in
MISSING.SAV, with VARNUM=7, the value of “X7” will be added to the
file. For the second case, the value of “X9” will be added, etc. This action
will place the correct variable names on MISSING.SAV. We do this
through a table match.

In addition, a counter will be created that increments by one for each

case (each case represents a missing value) but resets to 1 when the ID
variable in MISSING.SAV changes. We do this for purposes of reporting,
not for the match itself.

Advanced Data Manipulation 5 - 31

SPSS Training
Switch to Missvar.sps (click Window..Missvar - SPSS Syntax
Editor)
Scroll down until you see the commands in Figure 5.30

Figure 5.30 Matching Portion of MISSVAR.SPS

We must also sort the MISSVAR.SAV file by VARNUM because it is

not in that order now. The variable COUNTER is created in the standard
manner.

Highlight the lines from GET FILE to EXECUTE

Click on the Run button

Switch to the Data Editor (click Goto Data tool )

The resulting file is, of course, ordered by VARNUM. The COUNTER

is out of its original order, but that won’t be a problem. We’ve added the
Variable CASE_LBL to MISSING.SAV so that we can see, for example,
that for cases 4, 5, and 6, each is missing on variable X2.

Advanced Data Manipulation 5 - 32

SPSS Training
Figure 5.31 Data File After Table Match

The hard part of our work is over. If you refer to Figure 5.24, you will
see a file with only seven cases, one for each of the original cases, and a
column for each of the ten original variables. To create such a file from
the current working data file, we need to use the AGGREGATE
command, breaking on ID, and spreading the CASE_LBL information
into separate variables.

We accomplish the second task first by using a vector.

Switch to Missvar.sps (click Window..Missvar - SPSS Syntax

Editor)

Referring back to Figure 5.30, the VECTOR command creates the

vector VNAMES which has 10 string elements, each with format A8. We
then place the CASE_LBL values into VNAMES, using COUNTER as the
placeholder. To see how this works, look at Figure 5.31.

When COUNTER=1, this variable is the first missing variable for

that case (in order from X1 to X10). This is true for cases 4, 5, and 6 for
X2, so the value “X2” is placed in VNAMES(1) for those three cases.

Highlight the commands from VECTOR to EXECUTE

Click the Run button

Switch to the Data Editor (click Goto Data tool )

Advanced Data Manipulation 5 - 33

SPSS Training
In the Data Editor (not shown), you can see how 10 new variables
have been created from the vector VNAMES and how the values of
CASE_LBL have been placed in the correct slots in VNAMES.

The final step is to use AGGREGATE. The MAX function is one way
to take the values of VNAMES1 to VNAMES10 and place them into
variables with the same names (we could also have used the MIN
function). Although the VNAMES variables are strings, the MAX function
can be applied. The N function puts the number of cases (here the
number of missing variables) into the variable NMISS.

Switch to Missvar.sps (click Window..Missvar - SPSS Syntax

Editor)
Click on the AGGREGATE command
Click on the Run button

Switch to the Data Editor (click Goto Data tool )

We have created the target file structure, with one case for each case
in the original data file, and 10 string variables which store the names of
the missing variables for each case. A report can now be produced like
Figure 5.23.

Figure 5.32 Data File With Missing Variable Information

To make the syntax easier to write, we used variable names X1 to

X10. You are likely to have less generic variable names, and if so, you will
simply need to specify them at the appropriate places in the program. If
you don’t want to do the AUTORECODE for the string variables and add
them to the numeric variables, you can do a separate report for each type

Advanced Data Manipulation 5 - 34

SPSS Training
of variable. And if you have many variables, you could use more than one
vector command.
Hopefully, this program provided you some hints and examples of
how to create elaborate SPSS programs that do data manipulation.

Note: The SPSS Missing Values module produces a report similar to the output
from this program. It also contains algorithms for imputing (substituting
values) for missing data.

Advanced Data Manipulation 5 - 35

SPSS Training

Advanced Data Manipulation 5 - 36

SPSS Training

Chapter 6 Introduction to Macros

Topics Introduction
Macro Basics
Macro Arguments
Macro Tokens
Viewing a Macro Expansion
Keyword Arguments
Using a Varying Number of Tokens
When Things Go Wrong

M
INTRODUCTION acros in most software programs are small routines of
commands that automate one or more tasks to make your work
more efficient and easier. Using a macro means not having to
construct commands each time you need to do a particular analysis or
data transformation task.

The SPSS Macro facility provides a means of constructing SPSS

commands and controlling their execution. Thus, the output of a macro is
a new set of tailored SPSS commands. Macros are most useful when one
or more SPSS programs contain repeating blocks of identical or similar
commands, but a macro can be useful in several different contexts. For
example, a macro can be used to:

1) Issue a series of the same or similar commands repeatedly, using

looping constructs.
2) Produce output from several procedures with a single command;
3) Change specifications on several procedures at once.

As noted in Chapter 2, SPSS macros are a bit different than those you
may have encountered in other software. Although the SPSS macro
facility has its own language, you don’t invoke a macro editor to create
macros. And SPSS macros don’t directly automate actions you take with a
mouse. Instead, SPSS macros are created like any other syntax file, then
executed in the same manner as other syntax. SPSS macros always
contain regular SPSS syntax in addition to the specialized macro
commands.

Introduction to Macros 6 - 1
SPSS Training

MACRO BASICS The structure of a macro is outlined below. Macros begin with the
DEFINE command and end with the !ENDDEFINE command. Macro
subcommands and keywords all begin with an exclamation point to
distinguish them from regular SPSS commands. The macro name (which
can also begin with an exclamation point) will be used in the macro call.

DEFINE macro name macro arguments and specifications

are placed here

macro body

!ENDDEFINE
.

The macro body can include SPSS commands, parts of commands, or

macro statements.

Once it has been constructed, using a macro is a straightforward

process. First, the macro has to be processed like any other syntax file.
This defines the macro to SPSS. Then you must do a macro call where
you specify the name of the macro and any necessary arguments (such as
variable names). The call, too, is equivalent to an SPSS command, so it is
also executed like any other syntax.

The macro definition commands are placed into a standard SPSS

syntax file, though they do not have to be in the same syntax file as the
macro call. As a consequence, the syntax to use an existing macro
requires just two commands, as shown next. The INCLUDE command
places the commands in the named file into the SPSS command stream.
In this instance, the INCLUDE command brings the macro definition
commands in MACRO.SPS into SPSS. The macro call is next, using the
macro name (MYMACRO) defined in MACRO.SPS. Any arguments on
MYMACRO are placed after the name, and the macro call ends with a
period, like all SPSS commands.

INCLUDE 'PathToMacro\MACRO.SPS'.
MYMACRO arguments.

Highlighting these two commands and clicking on the Run button will
execute the macro. Again, no special macro editor needs to be invoked to
execute macros in SPSS. And a macro call does not have to immediately
follow a macro definition.

Interestingly, the macro facility does not build and execute

commands by itself; rather, it expands strings in a process called macro
expansion. After the strings are expanded, the commands that contain
the expanded strings are executed as part of the SPSS command
sequence. This may seem rather esoteric knowledge, but knowing that
the macro facility manipulates strings is important when creating
complex macros and deciphering their errors.

Introduction to Macros 6 - 2
SPSS Training

MACRO There are many subcommands and specifications available in the macro
ARGUMENTS facility, but macro arguments are central to macro creation. Arguments
are input to the macro that will be supplied by the user, and they can be
of two types: keyword and positional.

Keyword arguments are assigned names in the macro definition, and

when the macro is called, they are literally identified by name and thus
can be given in any order.

Positional arguments are declared after the keyword !POSITIONAL;

when the macro is called, they are identified by their relative position
within the macro definition. The first positional argument is referred to
by “!1” in the macro body, the second by “!2”, and so forth.

MACRO TOKENS Tokens are used in conjunction with arguments in macros. A token is a
character or group of characters that has a predefined function in a
specified context. That definition probably doesn’t seem too illuminating
because tokens are so varied in their operation. In essence, a token
declaration tells SPSS how to know or recognize which elements on the
macro call should be associated with this argument. The simplest token
definition is one that assigns the next n values (or tokens) to the
argument. It is better to see this in action where we can also discuss
other issues of macro syntax.

In the macro below, we’ll focus first on the argument and tokens. The
arguments to a macro must be enclosed in parentheses. The use of the
!POSITIONAL argument keyword tells SPSS that arguments, or input, to
the macro will be positional in the macro call. The !TOKENS (4)
specification declares that there will be four positional tokens.

DEFINE DESMACRO (!POSITIONAL !TOKENS (4)).

DESCRIPTIVES VAR = !1.
!ENDDEFINE.

DESMACRO EDUC SPEDUC AGE AGEWED.

Next comes the body of the macro, which is a simple DESCRIPTIVES

command, with one exception. We replace the usual list of variables with
the keyword !1, which refers to the first (and only) positional argument in
the macro. When SPSS expands this macro, it will place the four
positional tokens in place of the !1. This means, in this instance, that they
must be variables for the command to execute properly.

The macro closes with the !ENDDEFINE command. The macro

DESMACRO is now defined (or at least will be once SPSS runs these
statements). We call the macro by using its name and then supplying four
positional arguments, which are variable names.

Introduction to Macros 6 - 3
SPSS Training
The diagram in Figure 6.1 may make the relationship between
arguments and tokens more clear. The keyword !POSITIONAL, an
argument, refers to where in the macro body the tokens will be placed.
Hence !1, the first positional argument, is placed after the equals sign.
The !TOKENS (4) specification on the argument is not referred to in the
macro body; instead, it tells SPSS how many elements in the positional
argument will be input by the user. In the macro call, the four variable
names are not just variable names, but also tokens.

Figure 6.1 Outline of Macro Structure

We will use the 1994 General Social Survey file to demonstrate how
this macro functions. We open the GSS file and then run this macro.

Click File..Open..Data
Switch to the c:\Train\ProgSynMac directory if necessary
Double-click on GSS94

Then

Click File..Open..Syntax
Double-click on MACRO1 (not shown)

Running a macro involves defining it, then calling it with any

necessary input. To do so,

Highlight the lines from DEFINE to DESMACRO

Click on the Run tool button

SPSS processes all the commands, echoing them to the Viewer. It

then runs a DESCRIPTIVES command that has been created by the
macro facility for the four variables we specified.

Introduction to Macros 6 - 4
SPSS Training
Figure 6.2 Descriptives Output from DESMACRO

That’s all there is to using a macro! Understandably, this was a lot of

work for a DESCRIPTIVES command, but it’s best to begin with simple
macros to illustrate their features and operation.

What happens when you define the same macro twice? Let’s find out
by rerunning the commands.

Switch to the MACRO1 Syntax (click Window..Macro1 - SPSS

Syntax Editor)
Highlight the lines from DEFINE to DESMACRO again
Click the Run tool button

First, note that we got another set of output from DESCRIPTIVES, so

clearly running the program again didn’t seem to cause any major
problems. However, we received a warning message that politely tells us
that DESMACRO has previously been defined, but that this new instance
will take precedence.

Introduction to Macros 6 - 5
SPSS Training
Figure 6.3 Warning Message From Defining DESMACRO Twice

When you are writing macros you will undoubtedly see warning 6804
now and then. Warnings can be turned off with the SET command, but
we don’t recommend that approach, especially when debugging your
macro definitions. Clearly, this warning means that once defined in an
SPSS session, a macro need not be defined again.

Introduction to Macros 6 - 6
SPSS Training

VIEWING A When constructing complex macros, it is usually necessary to see the

MACRO exact commands created by the macro facility. This is particularly crucial
when debugging macros. You can turn this option on with the SET
EXPANSION command,

SET MPRINT ON.

We’ll try this to see the result for DESMACRO.

Switch to the MACRO1 Syntax (click Window..Macro1 - SPSS

Syntax Editor)
Add the line SET MPRINT ON. just above the DESMACRO call
Highlight only these two lines
Click on the Run tool button

Figure 6.4 SET MPRINT ON. Added

In the Viewer, the lines beginning with "M>" appear due to the SET
command. The macro call appears just above these lines, and the
DESCRIPTIVES command constructed by the macro appears below. The
four positional tokens (EDUC, SPEDUC, AGE, and AGEWED) are
placed, in that order, on the DESCRIPTIVES command just after the
equals sign (which is where the !1 positional argument was in the macro
definition).

Figure 6.5 Macro Expansion of DESMACRO

Introduction to Macros 6 - 7
SPSS Training

KEYWORD Since arguments are central to macros, we next review the syntax for a
ARGUMENTS keyword argument. This type of argument is given a user-defined
keyword in the macro definition. In the macro body the argument name is
preceded by an exclamation point, so keyword arguments should be seven
characters or less in length. On the macro call, though, the keyword is
specified without the exclamation point (when and when not to use an
exclamation point is often a point of confusion for those new to macros).

Switch to the MACRO1 Syntax (click Window..Macro1 - SPSS

Syntax Editor)
Scroll down until you see the program in Figure 6.6

This macro is called TABX and allows the user to create a

crosstabulation of two column variables and one row variable. There are
two keyword arguments defined, ARG1 and ARG2. The first is defined
with two tokens and the second with one token. The arguments are
separated by a slash, and the whole set is enclosed in parentheses.

Figure 6.6 TABX Macro Definition

The body of the macro consists of a standard CROSSTABS command,

but as with the previous example, the variables are replaced with
arguments. The keyword !ARG2 is placed where the row variables go, and
!ARG1 is put in the column variable position. They are both preceded
with exclamation points because they are in the macro body. The
CROSSTABS command also requests column percentages and the chi–
square statistic.

Notice that !ARG1 appears after !ARG2 in the macro body. Keyword
arguments can appear in any order in the macro. When the macro is
called, the three variables are input as arguments, two for ARG1 and one
for ARG2. Two crosstabulations will be produced, PRES92 by SEX and
PRES92 by RACE.

Highlight the lines from DEFINE TABX to TABX

Click on the Run tool button

Introduction to Macros 6 - 8
SPSS Training
When the output is produced, the first thing to observe is that the
macro expansion is displayed not just for the macro call, but also for the
macro definition itself. This is not a problem, but it can be annoying. You
can turn this off by placing a SET MPRINT OFF command before
DEFINE, and then a SET MPRINT ON before the macro call.

Figure 6.7 Macro Expansion for Macro Definition and Call

If you scroll down a bit, you will see the table in Figure 6.8. The
macro facility created the CROSSTAB command shown in Figure 6.7,
which when executed by SPSS created two tables. In the first table, we
see that females were more likely to vote for Clinton in 1992, and males
more likely to vote for Perot.

Introduction to Macros 6 - 9
SPSS Training
Figure 6.8 Crosstabs of 1992 Presidential Vote and Sex

USING A What if you wish to use a different number of inputs, or tokens, to the
VARYING macro arguments? You might want to create crosstab tables with four
row variables and three column variables the next time you call TABX.
NUMBER OF This is easily accomplished with the !CHAREND or the !ENCLOSE
TOKENS keywords, both of which we will briefly review.

The next series of commands in MACRO1.SPS define a macro named

CORMACRO that calculates correlation coefficients for a series of
variables.

Switch to the MACRO1 Syntax (click Window..Macro1 - SPSS

Syntax Editor)
Scroll down until you see the CORMACRO definition

The macro definition contains two positional arguments. The first

uses the !CHAREND keyword and in quotes specifies a colon (:) as the
ending character. !CHAREND assigns all tokens up to the specified
character to the first positional argument, so no exact number of tokens is
specified. The second positional argument also uses !CHAREND,
assigning all tokens up to the slash (/) to the second positional argument.
The SPSS CORRELATIONS command is the macro body, with the two
positional arguments on either side of the WITH keyword. We have also
first turned macro printing off, then back on, to avoid extraneous output.

Introduction to Macros 6 - 10
SPSS Training
Figure 6.9 CORMACRO Macro Definition

The first call of the macro assigns the variables AGE and EDUC to
the first positional argument. This is because the two variables are
followed by a colon, which tells the macro facility that input to the first
positional argument has ended. The second positional argument has only
PRESTG80, ended by a slash.

Before running this we should verify that the Correlation Autoscript

(Autoscripts are discussed in the Programming Scripts Using SPSS
course) is not enabled. To do so:

Click Edit..Options, then click the Scripts tab

Click Enable Autoscripting checkbox, if necessary, to uncheck
it
Click OK

We’ll run the commands through the first macro call.

Highlight the commands from SET MPRINT OFF to the first

call of CORMACRO
Click on the Run tool button

You can see the macro expansion in the Viewer and the
CORRELATIONS command created by the expansion. Below this is the
correlation output, where education is reassuringly highly correlated with
occupational prestige, but not age. The key point is that SPSS knew
where to end each set of tokens by using ending characters. This means
that with this feature any number of variables can be specified.

Introduction to Macros 6 - 11
SPSS Training
Figure 6.10 Output From Call of CORMACRO

To illustrate another aspect of macro operation, run the second call of

CORMACRO

Switch to the MACRO1 Syntax (click Window..Macro1 - SPSS

Syntax Editor)
Put your cursor on the second CORMACRO call
Click on the Run tool button

The macro call does not end with a slash after TVHOURS, the second
variable for the second positional argument. However, the macro worked
perfectly, creating the two by two correlation matrix you see in Figure
6.11.

Introduction to Macros 6 - 12
SPSS Training
Figure 6.11 Output From Second Call of CORMACRO

What happened is that the macro expansion placed PRESTG80 and

TVHOURS after the WITH keyword on the CORRELATIONS command,
but then kept reading in the macro call because no slash had been
encountered. However, before a slash was encountered, SPSS read a
period, which ends any command. Macro expansion stopped, and the
command was passed to SPSS to be executed. Although you shouldn’t
always rely on a period to correctly end a macro expansion, it can work in
certain instances.

The !ENCLOSE token keyword can be an alternative to !CHAREND

in this instance. The next macro definition in MACRO1.SPS illustrates
its operation.

Switch to the MACRO1 Syntax (click Window..Macro1 - SPSS

Syntax Editor)
Scroll down until you see the CORMAC2 definition

Introduction to Macros 6 - 13
SPSS Training
Figure 6.12 CORMAC2 Macro Definition

The !ENCLOSE keyword assigns all tokens between two specified

characters to an argument. The starting and ending characters can be
any one character string. Thus, the first positional argument, !1, will
include all the tokens between the specified characters, which are the left
and right brackets. The syntax may look rather daunting, but the
characters have to be placed in apostrophes, separated by a comma, and
the whole specification enclosed in parentheses. The second positional
argument, !2, also will have an indeterminate number of tokens, also to
be defined as those between left and right brackets.

The macro body is identical to the previous macro. The macro call of
CORMAC2 puts three variables between the first set of brackets. They
will be placed on the left of the WITH keyword in the CORRELATIONS
command. The second list of variables between brackets will be placed at
the second positional argument, to the right of the WITH keyword.

Highlight the lines from SET MPRINT OFF to CORMAC2

Click on the Run tool button

The macro is expanded and creates the intended CORRELATIONS

command. SPSS executes the command and a portion of the table is
displayed in Figure 6.13.

Introduction to Macros 6 - 14
SPSS Training
Figure 6.13 Output From Call of CORMAC2

WHEN THINGS The key to understanding errors in macros is to understand the

difference between errors in macro definition and macro calls. The macro
GO WRONG
facility simply creates strings that (you hope) are valid SPSS commands,
so incorrect SPSS syntax will not prevent macro expansion. Instead, the
syntax error will show up when SPSS executes the command.

So what does happen when the user calls a macro with the wrong
type of input? One possible problem is to add extra tokens to a list. In
Figure 6.14 you see the TABX macro with an error on its call. Instead of
one token on ARG2 we have two, PRES92 and VOTE92. If we execute
this call, what will SPSS do?

Introduction to Macros 6 - 15
SPSS Training
Figure 6.14 Misspecified Macro Call: Extra Token

Not surprisingly, the command doesn’t work, but perhaps not in the
way we might have expected. SPSS expands the macro and doesn’t
complain upon expansion. As is evident in Figure 6.14, PRES92 is placed
in the correct spot before the BY keyword, and VOTE92 is ignored while
the rest of the command is built. Then, rather than throwing away this
extra specification, SPSS simply adds it to the end of the command and
then puts in a period. This causes a problem because it now appears that
VOTE92 is a part of the CELLS subcommand, which it certainly is not.
SPSS then provides a warning message when it tries to run the
constructed command.

Figure 6.15 Error in Macro Call: Extra Token

And if instead we invoke this same macro with no tokens for ARG2,
the result is another problem, as in Figure 6.16. The CROSSTABS

Introduction to Macros 6 - 16
SPSS Training
command has no variables before the BY keyword, but the macro
expansion didn’t complain about the missing argument. SPSS, of course,
did complain when it processed the command, with an accurate warning
message about the problem.

Figure 6.16 Error in Macro Call: Missing Tokens

What happens when a macro is defined with incorrect macro syntax?

The answer is that you will receive error messages, and possibly quite a
few, as the macro facility becomes confused. As you might imagine, one of
the most common errors is to leave out a parenthesis. We’ve redefined the
last correlations macro (naming it CORMAC3), but left out the first
parenthesis before the first !POSITIONAL keyword. The result is the
mess you see in Figure 6.17.

Introduction to Macros 6 - 17
SPSS Training
Figure 6.17 Error in Macro Definition

We received error 6807, which clearly states that the DEFINE

command is missing definitions within a set of parentheses. However,
this error causes a cascade of six other similar errors, all concerning an
undefined macro variable or operator. As is usual when reading error
messages, begin with the first, not the last, when looking for the problem.
The CORRELATIONS command couldn’t be constructed, given these
errors.

SUMMARY This chapter has reviewed the basics of macro definition and calls,
focusing on arguments and tokens. The macros in this chapter were
simple so that we could focus on understanding macro construction. The
next chapter will introduce more complex macro features and more
typical uses of macros.

Introduction to Macros 6 - 18
SPSS Training

Chapter 7 Advanced Macros

Topics Introduction
Looping in Macros
Producing Several Clustered Bar Charts
Double Loops in Macros
String Manipulation Functions
Direct Assignment of Macro Variables
Conditional Processing
Creating Concatenated Stub and Banner Tables
Additional Recommendations

T
INTRODUCTION he macro facility is very powerful and offers several additional
features that were not discussed in the previous chapter. For
example, since macros essentially manipulate and create strings,
the macro facility offers several functions that manipulate strings,
including some that are equivalent to the string functions available in
SPSS. Macros can also do looping to accomplish repetitive tasks and
conditional processing (IF like statements). Also, macros permit
assignment of values to macro variables either as constants or through
the evaluation of an expression.

We will review two programs that illustrate some of these features.

Advanced Macros 7 - 1
SPSS Training

LOOPING IN As we have discussed, SPSS syntax does not permit procedures to be

MACROS placed inside loops. However, macros can be used to mimic this function.
Since macros produce strings which are SPSS commands, the use of a
loop in a macro can yield a similar, repeating set of SPSS commands.
This technique can be used to advantage in many circumstances.

There are two types of loop constructs in macros: the index loop and
the list processing loop. These constructs allow the user to iterate over
just about anything, including variables, numbers, file names, or
procedures.

The index loop is similar to the LOOP command in SPSS, looping a

number of times defined by numeric arguments on the index. The syntax
is:

!DO !VAR = start !TO finish [!BY step ].

specifications
!DOEND.

where !VAR could be !I or a similar macro variable, and start and

finish are numeric values supplied by arguments.

The list processing loop also begins with !DO and ends with !DOEND,
but has this syntax

!DO !VAR !IN (list).

specifications
!DOEND.

Here, the loop iterates for as many elements as there are in the list,
which is usually a macro argument (specifically, the tokens for that
argument). On each iteration of the loop, the value of !VAR is set to one of
the values of the list, in order.

PRODUCING The first macro illustrates the list processing looping construct.
SEVERAL
Click File..Open..Data
CLUSTERED BAR
Switch directories to c:\Train\ProgSynMac if necessary
CHARTS Double click on GSS94

Now that we have the data file, we’ll open the syntax file.

Click File..Open..Syntax
Double click on MACRO2

When creating a clustered bar chart from the menus, neither

standard nor Interactive graphs permit multiple graph requests by
placing several variables in the Category Axis (Horizontal or X-axis in
IGraphs) or Define Clusters By (by default, Color in IGraphs) boxes. That

Advanced Macros 7 - 2
SPSS Training
can be frustrating when a user wants to produce several bar charts at
once. The first macro in MACRO2.SPS, displayed in Figure 7.1, solves
this problem by allowing several clustered bar charts to be specified at
once.

Note We illustrate using standard graphs because the basic Graph command is
simpler than an IGraph command. For those using Interactive Graphs,
equivalent macros using Interactive Graphs are included at the end of
the Macro2.sps Syntax file.

Figure 7.1 Macro to Produce Clustered Bar Charts

The DEFINE command names the macro CLUSBAR. Two arguments

are declared, CAT and CLUS, which represent the Category Axis and
Define Clusters By variables. We use the !CHAREND keyword to assign
the first set of tokens (variable names) up to the slash (/) to CAT, and
assign the remaining tokens to CLUS with the !CMDEND keyword.

The !DO--!DOEND list processing syntax tells the macro facility to

loop over the tokens in !CAT, one by one. The value of the loop on each
iteration—a token from !CAT—will be placed in the macro variable !I.
There will be as many iterations as there are tokens in !CAT. It’s no more
complicated than this.

A standard GRAPH command follows !DO to produce a clustered bar

chart. However, rather than specify variable names after the two BY
keywords, the value of the macro variable !I is substituted in the position
of the Category Axis variable (after the first BY keyword), and !CLUS is
placed in the Define Clusters By spot (after the second BY).

Advanced Macros 7 - 3
SPSS Training
Only one token is named for CLUS on the macro call, so there will be
one clustering variable. But on each iteration of the loop, a token from
CAT will be placed in the position of !I. The macro call names two tokens
(here variable names) for CAT, so two GRAPH commands will be
constructed, one for each token.

The macro body ends with !DOEND first to close the loop, then
!ENDDEFINE to close the definition.
On the macro call, the variables NATHEAL and NATENVIR (asking
about the level of federal spending on these issues) are read as values
(tokens) of CAT, and the variable CLASS as the value of CLUS.

Highlight the lines from DEFINE to CLUSBAR

Click the Run tool button

SPSS produced two clustered bar charts, creating the command

echoed in the Viewer for the first bar chart of NATHEAL and CLASS in
Figure 7.2. It appears that class is related to attitude toward spending on
the nation’s health.

Figure 7.2 Bar Chart of NATHEAL and CLASS

If you scroll to the second bar chart (not shown) you can see that the
macro call did indeed produce two graphs.

Advanced Macros 7 - 4
SPSS Training
The ability to loop over variable names is very powerful and allows
you to create a huge amount of output very quickly. Below we rewrite the
macro to allow additional variables for the CLUS argument.

DOUBLE LOOPS Our next example directly extends the macro we just examined to allow
multiple variables for both the Category Axis (Horizontal or X-Axis
IN MACROS
variable for Interactive Graphs) and Define Clusters By (by default, Color
variable for Interactive Graphs) variables. Again we present the example
using a simple Graph command, and the equivalent macro using an
IGraph command appears at the end of the Macro2.sps file.

Return to the Macro2 Syntax window (click Window..Macro2 –

SPSS Syntax Editor)
Scroll down to the CLUSDBR macro (the command Define
CLUSDBR)

Figure 7.3 Macro Containing Double Loop

As before, the DEFINE command names the macro CLUSDBR and

the CAT and CLUS arguments are declared.

Within the DO !I…!DOEND loop examined earlier, there is a second

loop nested: DO !J…!DOEND. This latter syntax tells the macro facility
to loop over tokens in !CLUS, one by one. The value on each iteration of
this inner loop– a token from !CLUS– will be placed in the macro variable
!J. There will be as many iterations of this inner loop as there are tokens
in !CLUS. Also, since this loop (involving !J) is nested within an outer
loop (involving !I), the inner loop will be repeated in its entirety for each
token in !CAT, which drives the outer loop.

Advanced Macros 7 - 5
SPSS Training
Both of the macro loop variables, !I and !J, appear in the GRAPH
command. As before !I is substituted in the position of the Category Axis
variable (after the first BY keyword), but now !J appears in the position
of the Define Clusters By spot (after the second BY). This position was
formerly occupied by the !CLUS argument itself.

Since there are two !DO structures, there must be two !DOENDs; the
inner !DOEND ends the inner (!J) loop and the outer !DOEND ends the
outer (!I) loop. If we reversed the order of the two loops, that is, if !DO !J
preceded !DO !I, the same set of Graph commands (and graphs) would be
produced, but in a different order. This is because the inner loop iterates
completely through its range or set of values for each successive value of
the outer loop. The macro definition is closed with an !ENDDEFINE.

The macro call names two tokens for CAT (NATHEAL and
NATENVIR) and two for CLUS (CLASS and MARITAL), so four GRAPH
commands will be constructed, one for each combination of a CAT and
CLUS token.

Highlight the lines from SET MPRINT OFF to CLUSDBR

Click the Run tool button

Figure 7.4 Bar Chart Produced from Macro with Double Loop

The Outline pane indicates, or you can discover by scrolling in the

Content pane, that the macro call produced four bar charts.
Next, we turn our attention to a more complex macro, though we will
first review some additional macro features.

Advanced Macros 7 - 6
SPSS Training

STRING Since macro expansion creates strings, it is only natural that the macro
MANIPULATION facility supplies about a dozen string manipulation functions to create
exactly the command needed. Some of the most commonly–used are
FUNCTIONS
these.

!CONCAT (string1, string2, ...): Concatenates several strings.

!QUOTE(string): Puts apostrophes around the string. Used to
construct SPSS syntax commands that require apostrophes,
such as labels and titles.
!HEAD(string): Returns the first token within a string.
!TAIL(string): Returns the end of the string after removing the
first token.
!BLANKS(n): Generates a string containing the specified number
of blanks. However, the blanks must be quoted because the
macro facility otherwise compresses (deletes) blanks in macro
expansion.

DIRECT In SPSS, a COMPUTE command can assign a value to a variable, like

ASSIGNMENT OF this:
MACRO COMPUTE X = 5.
VARIABLES
The macro facility has essentially the same capability by use of the
!LET command. Its syntax is:

!LET !var = expression

where the expression may be a single token, a constant, or a macro

function. So it is possible to have statements like these:

!LET !X = 5
!LET !Y = !CONCAT(ABC,!1)

CONDITIONAL Conditional processing in a macro is accomplished with an !IF–!IFEND

PROCESSING construct. Unlike SPSS, macros use a !THEN operator to make an
assignment. Thus the syntax is:

!IF (expression) !THEN statements

!ELSE other statements
!IFEND

When the expression is true, the statements following !THEN are

executed. When the expression is false and when !ELSE is specified, the
statements following that command are executed.

Advanced Macros 7 - 7
SPSS Training

CREATING Using multiple categorical variables in both the rows and columns of a
CONCATENATED table is very common. The simplest possible TABLES syntax to create
such a table, for a set of generic variables A through F with column
STUB AND percents, is this:
BANNER TABLES
TABLES
/TABLE= A + B + C BY D + E + F
/STATISTICS cpct( :D E F ).

If creating such tables were a recurring task, it would be handy to

automate their production rather than have to use the TABLES dialog
boxes for each table. The next macro carries out this job by using several
of the macro features introduced above.

Return to the Macro2 Syntax window (click Window..Macro2 –

SPSS Syntax Editor)
Scroll down until you see the commands in Figure 7.5

The macro, called STB_BAN, will create a table with any number of
variables on the stub and banner. It also allows the user to specify a title
for the column percent statistic and an overall table title.

Figure 7.5 STB_BAN Macro

The syntax of a TABLES command doesn’t appear very tricky to

construct, but we have to put a plus sign (+) between the variables on the
stub and banner. And if the macro is to be written to allow an unlimited

Advanced Macros 7 - 8
SPSS Training
number of variables, there is no way to specify ahead of time how many
plus signs are needed. That is the chief problem the macro solves.

We work through this macro, line by line.

DEFINE: An argument called STUB is defined that has tokens

assigned to it that are enclosed in parentheses. A second
argument called BANR is defined that also has tokens
assigned that are in parentheses. These two arguments will
contain the variables used in the table. If you scroll down in
MACRO2.SPS, you can see that the macro call (listed below)
places the variables DRINK, SMOKE, and VOTE92 in STUB
and SEX, RACE, and CLASS in BANR.

STB_BAN Stub (DRINK SMOKE VOTE92 )

Banr ( SEX RACE CLASS) Colstat "Percent"
TTL *Drinking, Smoking, and Voting*.

The DEFINE command continues by defining the argument

COLSTAT and assigning it one token on the macro call. It concludes by
defining the argument TTL that will have, as a token, all the text
between two asterisks on the macro call. Looking at the macro call in
MACRO2.SPS, we can see that TTL is used to specify the table title
“Drinking, Smoking, and Voting,” and COLSTAT specifies a statistics
label for the column percent.

All of this is straightforward and similar to what we have seen before.

The interesting part of the macro begins next. First, though, look at the
TABLES command itself in the macro. It is reproduced here for reference:

TABLES / TABLE=!ST BY !BA /statistics cpct(!Colstat : !Banr)/

TITLE !QUOTE(!TTL).

Two arguments, !ST and !BA, which we have not defined yet, are
placed in the position of the stub and banner variables. This seems odd
since we have defined STUB and BANR as arguments to contain these
same variables. In fact, !BANR is used after the colon in parentheses on
the STATISTICS subcommand to tell SPSS to calculate column
percentages. We can use !BANR in the latter position because the
variables appear together with no intervening commas or plus signs, but
that is not true in the TABLE subcommand.

Four !LET commands follow the DEFINE command. They set the
macro variable !STCPY equal to STUB and !BACPY equal to BANR. In
other words, they make copies of the variables that the user inputs. The
third and fourth !LET commands set the new macro variables !ST and
!BA (the ones that will actually be used in the TABLES command) to
!NULL, which is an empty string of length 0.

The macro then has two similar sections, one for the stub variables
and one for the banner variables.

Advanced Macros 7 - 9
SPSS Training
!DO: This is a list processing loop, with the variable !S taking on
successive values of the tokens in !STUB. To follow the macro
through, we will use the first token in !STUB, which is
DRINK. That is the first value of !S.

!LET !ST: !ST, which begins as a null string, is set equal to the
concatenation of itself, plus the first token in !STCPY, which
is DRINK, plus one blank. So !ST = “DRINK ”. So far, so good.

!LET !STCPY: The !TAIL function returns all tokens except the
first, so !STCPY now consists of SMOKE and VOTE92.

!IF: This IF construct checks to see whether !STCPY is not equal

to !NULL (the null string). In this first pass through the loop,
that is true, since it contains SMOKE and VOTE92. Then the
!LET command sets !ST equal to the current value of !ST,
which is “DRINK ”, plus a plus sign and another blank. So
now !ST will be equal to “DRINK + ”. As you can see, we are
beginning to create the TABLES subcommand syntax.

Now we understand why copies were made of the original arguments;

we are modifying the list of tokens in the macro on the fly.

On the second iteration, !S is set to SMOKE. !ST is set to “DRINK +

SMOKE ”. !STCPY is set to the last token, VOTE92. Since !STCPY is not
yet null, !ST is set to “DRINK + SMOKE + ”.

On the third and last iteration, !S is set to VOTE92. !ST is set to

“DRINK + SMOKE + VOTE92”. !STCPY is set to the null string because
there is only one token in !STCPY when the !TAIL function is evaluated.
The !IF command returns a false result, so the !LET is no longer
executed, and the !IF ends along with the loop. The correct stub syntax
has been created.

Exactly the same process is used to create the syntax for the banner.
Then in the TABLES command within the macro body, quotes are placed
around the title, which is input as the token for TTL, using the !QUOTE
function.

Interestingly, !COLSTAT is placed in the position for a statistics label

but seemingly has no quotes placed around it in the macro body.
However, in the macro call, note that the value of COLSTAT (“Percent”)
is specified with quotes. When you specify a token on a macro call, it is
not like specifying a label in a regular SPSS command. Since the token
specification for COLSTAT did not use the !ENCLOSE keyword (as we
did for TTL), all the text after COLSTAT on the call is read as one token,
including the quote marks. This turns out to be an elegant solution to put
quotes around labels in macros.

Let’s define the macro.

Highlight the lines from SET MPRINT OFF. to !ENDDEFINE.

Click the Run tool button

Advanced Macros 7 - 10
SPSS Training
Switch to the Viewer window

If everything worked correctly, there will be no errors in the output.

Then let’s call the macro.

Return to the Macro2 Syntax window (click Window..Macro2 –

SPSS Syntax Editor)
Highlight the next three commands: SET MPRINT ON.,
STB_BAN, and SET MPRINT OFF.
Click the Run tool button

The macro worked perfectly, placing three variables each on the stub
and banner and creating the command you see in Figure 7.6. The token
“Percent” was substituted for !COLSTAT, and the TITLE subcommand
was created with the title “Drinking, Smoking, and Voting” placed in
quotes. The !BANR argument after the colon placed the three banner
variables in that spot.

Figure 7.6 Command Created From STB_BAN Call

Scrolling down a bit will display the table created from this
command. There are interesting relationships between the three
demographic variables and the three behavioral variables in the stub.

Advanced Macros 7 - 11
SPSS Training
Figure 7.7 Table Created From STB_BAN Call

One oddity you may have noticed is the large number of lines between
the macro definition in the Viewer, and the macro call and table. A
portion of these lines is shown in Figure 7.8. This occurs because of all
the periods ending the commands in the macro body, such as on the eight
!LET commands.

Figure 7.8 Blank Lines Created From STB_BAN Call

SPSS commands normally end with a period, but some macro

commands don’t have that restriction, including !LET, !IF, and !DO. Thus
the following syntax is perfectly correct in a macro:

Advanced Macros 7 - 12
SPSS Training
!LET !STCPY = !Stub !LET !BACPY=!BANR !LET !ST=!NULL
!LET !BA=!NULL
Not only are several commands placed on the same line, but also the
line doesn’t end with a period. You can use this feature if you wish, and it
will reduce the amount of output, but usually writing macros in this
fashion makes them harder to read and debug, so it may not be the best
practice except for the experienced user.

ADDITIONAL To ease production of macros we suggest the following steps for your first
RECOMMENDATIONS few macros.

1) Test the SPSS commands that the macro is designed to

produce and make certain they work.

2) Add a single argument at a time to the macro and thus isolate

potential problems.

3) Be careful about the order of arguments and how you instruct

SPSS to detect when a particular set of tokens for an
argument is finished.

4) Use default values for macros when debugging.

The last bit of advice refers to the ability in macros to establish a

default setting for an argument. Here is an example.

DEFINE MACDEF (VARLIST = !DEFAULT (SEX) !TOKENS(3) /

STATS= !TOKENS (2)).
FREQUENCIES VARIABLES !VARLIST /STAT = !STATS.
!ENDDEFINE.

MACDEF STATS = MEAN STDDEV.

In this simple macro for a FREQUENCIES command, we have

defined the argument VARLIST to consist of three tokens, or variables.
However, if VARLIST is not specified in the macro call, it will have a
default value of “SEX” and create frequencies output only for that
variable. And this is true even though three tokens would normally be
input for VARLIST. Using the !DEFAULT keyword is a way to reduce the
number of things being substituted and to isolate potential problems.

SUMMARY We’ve discussed more advanced macro commands and functions, and
we’ve reviewed three programs that use many of these features to
accomplish common tasks.

Advanced Macros 7 - 13
SPSS Training

Advanced Macros 7 - 14
SPSS Training

Chapter 8 Macro Tricks

Topics Combining Input Programs and Macros

Ordering Tables and Charts
The Case of the Disappearing Command

I
INTRODUCTION n previous chapters we have introduced the basic syntax of the SPSS
macro language and have demonstrated macro use with a series of
examples. In this chapter, we extend the discussion by presenting
several macro features that we feel may not be obvious, but have proven
to be very useful in practice. These examples are based on programs
written by the SPSS Consulting group for customer applications. The
macros presented here have been simplified so we can easily focus on the
salient points.

One of the macros uses an input program to generate a data file for
testing purposes. A second pairs together crosstabulation tables and
clustered bar charts based on the same variables. The final example
demonstrates how macro logic can be used to include or exclude SPSS
commands from a program. These serve to illustrate how macros are used
in practice.

Macro Tricks 8 - 1
SPSS Training

COMBINING When developing custom applications, it is sometimes necessary to

INPUT generate test data against which the application is run. This is done
because a small data sample provided by a customer may not show the
PROGRAMS AND
full range of variation that may be faced in practice. Also, in statistical
MACROS training courses it is useful to create data files that simulate various
relationships among variables.
To have SPSS create cases with no input requires, as we mentioned
earlier, an input program. Placing the input program within a macro
permits you produce data files with different characteristics by simply
changing the macro arguments.
Since the macro will create a data file, no data need be in the Data
Editor window in order to start.

Click File..Open..Syntax (move to c:\Train\ProgSynMac)

Double-click on Simdat

Figure 8.1 Macro to Produce Random Data with Values 1 to 5

The MAKEDAT macro contains two arguments, one for the number of
cases (NCASES) and one for the number of variables (NVARS). For each
argument a single token (TOKENS(1)) is expected, which makes sense,
since only a single value need be given for the number of cases or the
number of variables.

Macro Tricks 8 - 2
SPSS Training
INPUT PROGRAM: If we focus first on the input program
(between the INPUT PROGRAM and END INPUT
PROGRAM commands), we see it is composed of two loops,
one to create cases and the other to create variables within a
case.

VECTOR: First a vector VAR is declared. It will contain the

variables built by the macro. Since the number of variables is
contained in the macro argument !NVARS, this argument is
used in the VECTOR command. The size of the vector will be
controlled through the !NVARS argument.

OUTER LOOP: The outer loop, indexed by #I, iterates from 1 to

the number of cases specified (!NCASES). Since an END
CASE command is within this loop, each iteration of the outer
loop will add a case to the file. Variables are actually assigned
values within the inner loop, while the outer loop controls
case generation.

INNER LOOP: A secondary loop, indexed by #J, assigns random

values to the variables for the current case. The loop contains
a COMPUTE command that sets the #Jth variable in the VAR
vector equal to a value produced by a pseudo-random number
function (RV.UNIFORM). The RV.UNIFORM(0,5) function
will generate a continuous value, uniformly distributed, in the
range 0 through 5. By adding .5 to it, then rounding, we
produce a uniformly distributed set of integer values 1 to 5. In
this way, if #J (via !NVARS) were set to 20, twenty variables,
VAR1 to VAR20, would each be assigned a value of 1 through
5 for the current case.

END FILE: The END FILE command is critical here. This is

because no data are being input, so there is no end of file
event to trigger the creation of the SPSS data file. Without
the END FILE command, the program would remain paused
in an input program state.

VARIABLE LEVEL: The VARIABLE LEVEL command is used to

declare measurement level for SPSS variables. Since the
variables are assigned integer values 1 through 5, we declare
them to be nominal type. We know in advance the name of
the first variable created VAR1. However, the name of the
last variable is constructed by concatenating (!CONCAT) the
string ‘VAR’ and the number of variables argument (!NVARS)
and a blank space ‘ ‘. Thus if !NVARS were set to 20, the last
variable would be VAR20 .

EXECUTE: Since this program contains no procedures, the

EXECUTE command forces execution.

SET SEED: This is not required, but you can specify the starting
point of the pseudo-random number generator by providing a
large integer value for the SEED. This is mainly used to
permit you to reproduce the same data values later, if desired.

Macro Tricks 8 - 3
SPSS Training
When the MakeDat macro is called, it will create a data file
containing 50 variables (!NVARS) and 1,000 cases (!NCASES).

Click Run..All
Switch to the Viewer window

Figure 8.2 Macro Generated Commands in Log

The argument value for !NCASES (1000) appears in the first LOOP
(outer loop) command. Also, the argument value for !NVARS (50), has
been substituted into the VECTOR, second LOOP (inner loop) and
VARIABLE LEVEL commands.

Switch to the Data Editor (click Goto Data tool )

Macro Tricks 8 - 4
SPSS Training
Figure 8.3 Random Data Produced by Macro

We see values (1 through 5) for the first few variables (of 50) and
cases (of 1,000).

Extensions The example presented was simple in that no relationships were imposed
on the created variables. You might want to create a data set that
demonstrates specific relationships, for example, strong or weak
correlations. Those interested in examining a more complex variation, in
which the variables are first forced to be uncorrelated and then a
relationship imposed, should examine the Syntax file AtSimdat.sps. It
generates a data file containing a user-specified number of categorical
and continuous predictor variables that are related to an outcome
variable. In addition, the number of categories present in the categorical
variables is passed as an argument.

Macro Tricks 8 - 5
SPSS Training

ORDERING In Chapter 7 we saw a macro example in which a series of clustered bar

TABLES AND charts were generated by placing the CHART command within macro
loops. We will use a variation of this to produce output that contains, for
CHARTS
each pair of variables on a list, a crosstabulation table and a clustered bar
chart displaying the same percentages.

The challenge here is to produce just the desired tables and charts in
the proper order. Specifically, if four variables were given, A, B, C and D,
then we want a crosstab followed by a bar chart displaying A by B, A by
C, A by D, then B by C, B by D, and finally C by D. Thus each unique
pairing of the variables on the list should produce a table and chart,
which should follow each other in the Viewer window.
As you, no doubt, suspect at this point, loops will be involved.
Although we have seen loops within macros earlier, here the same list of
variables must drive two separate loops in a coordinated fashion. In order
to accomplish this, we will make use of the !TAIL macro function
reviewed in Chapter 7.

Click File..Open..Data (move to c:\Train\ProgSynMac)

Double-click on GSS94
Click No when asked to save contents of Data Editor

Click File..Open..Syntax (move to c:\Train\ProgSynMac)

Double-click on CrossChart

Figure 8.4 CrosCht Macro

The CrosCht macro takes a single argument (VLIST), a list of

variables on which the tables and charts will be based. Below we discuss
the elements of the macro.

Macro Tricks 8 - 6
SPSS Training
!LET !TAILIST: The macro variable !TAILIST is set equal to the
list of variables passed as an argument to the macro. It will
be used later to store the tail (all tokens except the first) of
the list of variables.

!DO !COLVAR !IN (!VLIST): This is the outer loop within the
macro. !COLVAR is a macro variable that stores the variable
name used as column variable in the crosstab table and the
cluster variable in the barchart. At each iteration of the loop
the next variable from !VLIST is assigned to !COLVAR.

!LET !TAILIST = !TAIL (!TAILIST): Here !TAILIST is set equal

to the tail of itself. Since !TAILIST was initialized to !VLIST,
in the first iteration of the loop it contains all variables
passed to the macro except the first. We later pair each
variable name in !TAILIST with the variable name stored in
!COLVAR to produce tables and charts. For example, if the
!VLIST argument contained variables A, B, C and D, during
the first outer loop iteration, !COLVAR = ‘A’ and !TAILIST
=’B C D”. During the second iteration, !COLVAR = ‘B” and
!TAILIST = ‘C D’. This allows us to match the head variable
on the list with the remaining variables.

!IF..!IFEND: At the last iteration of the outer loop, !TAILIST will

be empty and no tables or charts should be attempted. The
!IF statement will only generate the CROSSTAB and GRAPH
commands if !TAILIST is not null.

!DO !ROWVAR !IN (!TAILIST): The inner loop will assign, during
each iteration, a value from !TAILIST to !ROWVAR. The
variable name stored in !ROWVAR will be paired with the
variable name in !COLVAR to create the table and chart.

CROSSTABS and GRAPH: The macro variables !ROWVAR and

!COLVAR are substituted as the row (category) and column
(cluster) variables in the CROSSTAB (GRAPH) command.
Since the CROSSTABS and GRAPH commands are within
the loop, they will be run in sequence for each combination of
!ROWVAR and !COLVAR variable names. This insures that
the table and graph for a pair of variables will appear
contiguously in the Viewer window.

If the variables A, B, C, D were passed to the macro, !COLVAR and

!TAILIST would take the following values at each iteration of the outer
loop.

Iteration !COLVAR !TAILIST

1 A BCD
2 B CD
3 C D
4 D null

Macro Tricks 8 - 7
SPSS Training
In this way, at each outer loop iteration the first variable name on the
list of remaining variables (!TAILIST) is dropped from this list and
becomes the !COLVAR.

Click Run..All

Figure 8.5 Crosstab Tables and Graphs Produced by CrosCht Macro

The Outline pane displays the sequence of a crosstab followed by a

graph for the last few variable combinations. The creative use of macro
string functions, such as !HEAD and !TAIL, permit macros to process
argument lists in flexible ways.

Macro Tricks 8 - 8
SPSS Training

THE CASE OF So far we have used macros to build commands. However, macro logic can
THE be used to build or not build certain commands based on arguments
passed to the macro. This can be useful if you wish data selection or
DISAPPEARING
modification done only when requested. We illustrate this in a macro that
COMMAND runs a simple human resources report for employee data. If an age
argument is passed to the macro, then a data selection command
(SELECT IF) will be run based on the age value. If no age argument is
passed to the macro, then no data selection command is built. In addition,
if age selection is performed, then a note is added to the title of the
summary report (CASE SUMMARIES procedure).

Click File..Open..Data (move to c:\Train\ProgSynMac)

Double-click on BankNew

Figure 8.6 Employee Data

The data file contains employee information. The macro will generate
a report containing summaries of education, current salary and time in
current job position for subgroups. Age is included in the file, but is not
visible in Figure 8.6.

Click File..Open..Syntax (move to c:\Train\ProgSynMac)

Double-click on MacroSelect

Macro Tricks 8 - 9
SPSS Training
Figure 8.7 MacroSelect Syntax File

We comment on several aspects of the macro, but focus on the

instructions that build the data selection command.

REPORT1: The REPORT1 macro takes two arguments: AGE and

GROUP. The AGE argument has a default value of ‘ALL’, but
if an age value is supplied it will be used in a SELECT IF
command. The GROUP argument has a default value of
JOBCAT (the job category variable name), and the GROUP
value (variable name) will be used to create subgroups within
the report. Since only one value can be supplied for age or
group (this could be generalized to allow for an age range, or a
list of group variables), each is declared as a single token
argument.

!IF..!IFEND: The !IF statement tests the value of the !AGE

argument. If !AGE is not equal to ‘ALL”, then an age value
has been passed in the macro call. In this case, a
TEMPORARY command is added (so the SELECT IF
command will only be in effect for the report). Also, a
SELECT IF command is created and the title string includes

Macro Tricks 8 - 10
SPSS Training
age information. If !AGE equals ‘ALL’, then no TEMPORARY
and SELECT IF commands are created, and a standard title
string is used.

TEMPORARY: The TEMPORARY command will cause any

transformation statements between it and the next procedure
to be temporary changes. That is, any data selection or
modifications will apply to the next procedure, but then
disappear. It is used because we do not want the age selection
to be permanent, but rather apply only to the report.

!CONCAT: The !CONCAT statement builds a SELECT IF

command by concatenating three strings. The first part is the
beginning of a SELECT IF command based on age (‘SELECT
IF (AGE >’). Next comes the age value passed as a macro
argument (!AGE), followed by the end of the command (‘).’). If
AGE 30 were specified on the macro call, the !CONCAT
statement would build the following command: ‘SELECT IF
(AGE > 30).’. This is a simple command and more complex
variations, based on lists of values or names, can be built
using the type of logic we reviewed in the Table example
found in Macro2.sps (Chapter 7).

!LET !TITLE: A macro variable storing the title to be used in the

report is set equal to a string that includes age cutoff
information.

!ELSE..!IFEND: If no age argument was passed, then no

TEMPORARY and SELECT IF commands are created, and
the !TITLE string makes no mention of an age cutoff.

SUMMARIZE: The Case Summaries procedure (SUMMARIZE)

will report on education (EDUC), current salary (SALARY)
and time in current job position (JOBTIME) broken into
groups based on the !GROUP value. Recall that, by default,
!GROUP will be equal to JOBCAT (job category) and, if a
!GROUP argument is used in the macro call, then !GROUP
will contain the variable name supplied. Also, the !TITLE
created earlier appears in the TITLE subcommand. Note that
the !QUOTE function is applied to !TITLE. Titles must be
enclosed in quotes, but if explicit quote marks (‘!TITLE’) were
used , the report title would be “!TITLE” and not “Human
Resources Report..”.

To demonstrate, the macro is called three times. The first instance,

“REPORT1 AGE 50.”, supplies an age cutoff, but no group variable. The
second, “REPORT1 GROUP GENDER.” has no age cutoff, but requests
that GENDER be used as the grouping variable in the report. Both
arguments are used in the last example “REPORT1 AGE 35 GROUP
EDUC.”. Let’s examine the results.

Macro Tricks 8 - 11
SPSS Training
Click Run..All
Scroll to the first Summarize pivot table

Figure 8.8 Report with Age Selection

Note that the Case Summaries pivot table and Title items are hidden
(double-click on item in Outline pane to hide) in these screen shots.

The TEMPORARY and SELECT IF commands appear in the Log and

we see that the report title reflects the age selection. The Case
Summaries report is grouped by job category (recall JOBCAT is the
default value for !GROUP). Notice the total sample size (N summary for
Total) value of 124 employees.

Scroll to the second Summarize pivot table

Macro Tricks 8 - 12
SPSS Training
Figure 8.9 Report with Gender Groups and No Age Selection

Since AGE was not used as an argument, neither a TEMPORARY nor

a SELECT IF command appears in the Log. The pivot table title makes
no reference to age. Also, note that the total sample size is 474, which is
considerably larger than in the previous report, because no data selection
is being done.

Scroll to the third Summarize pivot table

Macro Tricks 8 - 13
SPSS Training
Figure 8.10 Report with Age Selection and Education Groups

An age selection is performed and is reflected in the title of the

report. The subgroup summaries are based on years of education (EDUC)
since EDUC was passed as the GROUP argument.

Variations on this example can provide a flexible reporting system,

especially when multiple pivot tables must be produced within an overall
report.

SUMMARY In this chapter we reviewed several examples (input programs within

macros, single argument lists driving multiple loops, optionally
generating data selection commands) that combine macro features to
produce more flexible and powerful macros.

Macro Tricks 8 - 14
SPSS Training

Exercises
All exercise files for this class are located in the c:\Train\ProgSynMac
folder on your training machine. If you are not working in an SPSS
Training center, the training files can be copied from the floppy disk that
accompanies this course guide. If you are running SPSS Server (click
File..Switch Server to check), then you should copy these files to the
server or a machine that can be accessed (mapped from) the computer
running SPSS Server.

Solutions to the major exercises (not every variation is included) can be

found in SelectedSolutions.sps.

Chapter 3 Complex File Types

Use the ASCII text file ProdNest.dat for this exercise.
As part of a household study, household members were asked to rate
a new product on three measures. Since most households have multiple
members, the data was structured in the following manner. First there is
a household record that contains household ID, number of members and
income category. Next there are as many records as members of the
household, each contains household ID, age, sex and three rating
variables. The goal is to read this as a nested file, so household
information (income) can be related to the individual ratings.
The record layout appears below:

Household record: Type ‘H’

Field Columns
Record Type 1
Household ID 2-5
Income group 8
# Members 11-13

Member record: Type ‘M’

Field Columns
Record Type 1
Household ID 2-5
Age 6-8
Sex 13
Rating 1 18
Rating 2 23
Rating 3 28

a) Read the data in SPSS as a nested file type.

b) Obtain means for age and the rating variables broken down by income
group.

c) Examine the correlations between age and the rating variables.

Exercises E - 1
SPSS Training

Chapter 4 Input Programs

Use the ASCII text file ProdRatings.dat for this exercise.
Customers were asked to name the three products (from a list of 10
within a product category) they most often used, and then to rate each of
the three products on five rating scale questions. Thus the data file
contains (in addition to ID and SEX), eighteen variables (three sets of
product plus five ratings). The task is to build a table of means with
products in the rows and the rating scale questions across the columns.
Since each customer picked three of the ten products in any order, a
product could be in any of the three sets of variables. Thus the required
table cannot be produced from the original data structure (one record per
customer).
However, the table can be run if each product a customer rates is on a
separate case. Your task is to accomplish this using the REPEATING
DATA command.

The file (ProgRatings.dat) has the following format:

Variable Position
ID 1-5
SEX 6-10
Product1 11-15
P1Rating1 16-20
P1Rating2 21-25
P1Rating3 26-30
P1Rating4 31-35
P1Rating5 36-40
Product2 41-45
P2Rating1 46-50
P2Rating2 51-55
P2Rating3 56-60
P2Rating4 61-65
P2Rating5 66-70
Product3 71-75
P3Rating1 76-80
P3Rating2 81-85
P3Rating3 86-90
P3Rating4 91-95
P3Rating5 96-100

a) Read the data so that there is one case per product rated by a
customer, with variables ID, SEX, PRODUCT, and RATE1 to RATE5.

b) Use the Means procedure (Analyze..Compare Means..Means) to create

a table of means of the rating scale variables broken down by product.

Exercises E - 2
SPSS Training

Chapter 5 Advanced Data Manipulation

SPSS data files, TranTransact.sav and ProdRatings.save are used in this
exercise.

a) Open the TrainTransact.sav SPSS data file. Open the Syntax file
Rating.sps and execute the commands, which will compute a RATING
variable with values 1 to 100. Open the TransactionAgg.sps Syntax file.
Modify the program so that if a customer takes a course, the rating is
stored under the course name variable in the aggregated file. If a course
was not taken by the customer, course variable’s value in the aggregated
file should be system missing.

This exercise involves the same data structure as the exercise in

Chapter 4, except we begin with an SPSS data file. See the data
description in the Chapter 4 exercises.

a) The SPSS data file, ProdRatings.sav, contains the following variables

ID, SEX, P1, P1R1, P1R2, P1R3, P1R4, P1R5, P2, P2R1…P2R5, P3,
P3R1…P3R5. The task is to create a new file with one case per product
rated by a customer.

b) Use the Means procedure (Analyze..Compare Means..Means) to

produce a table of means for the rating scale variables broken down by
product. Use the Crosstabs procedure to examine the two-way table of
product by sex.

Chapter 6 Introduction to Macros

Use the GSS94.sav SPSS data file for this exercise.

a) Use the SPSS menus to run a Descriptives analysis

(Analyze..Descriptive Statistics..Descriptives) for a single variable.
Return to the dialog and use the Paste pushbutton to obtain the Syntax
for the Descriptives command.

b) Build a macro named TestDes that accepts a single argument (one

variable name) and runs the DESCRIPTIVES command. Build several
variations of the macro: one using a positional argument and one using a
keyword argument. Experiment with additional argument variations:
first have the argument accept a single token, then have it accept an
argument enclosed within colons (i.e. : age : ), then have it accept an
argument ending in a colon (:), finally have it accept the end of command
as a terminator. Test each variation with a single variable name and a
list of several variables (note that the single token variation should fail
when a list of variables is used in the macro call).

c) Modify the final version of the macro created in the previous step, so it
accepts a second argument, named STATS, which will be a list of the
statistics to be displayed in the Descriptives output.

d) For those with extra time: Test the TestDes macro with the arguments
in different orders. If the macro fails due to argument order, modify the
macro so that argument order does not matter.

Exercises E - 3
SPSS Training

Chapter 7 Advanced Macros

Use the GSS94.sav SPSS data file for this exercise.

a) Write a macro that takes one argument: a variable list. It should run a
Frequencies command, using all the variables appearing on the variable
list.

b) Modify the macro to accept a second argument: a chart option. If the

Chart option is set to YES, then the Frequencies command should also
produce bar charts (using the /BARCHART subcommand). Set the default
value of the chart option to NO.

c) Test to see if your macro will work with upper and lower case letters
(YES, Yes, yes). If not, use the !UPCASE function so it works in
regardless of case.

d) Modify the macro so that instead of running a single Frequencies

command for all variables passed through the argument, it runs a
Frequencies command for each variable, followed by an standard pie
chart. Save the Syntax file as FreqMac.sps.

Chapter 8 Macro Tricks

Use the BankNew.sav SPSS data file for this exercise.

a) Open the MacroSelect.sps Syntax file used in the chapter. Modify the
macro so that in place of the age argument it takes a YEAR (4-digit)
argument. If this argument is specified, only those born after the
specified year will be included in the report (hint: use the XDATE.YEAR
SPSS function to obtain the year from the bdate variable).

b) Modify this macro so the YEAR argument will take two values, a
beginning year and an ending year. If this argument is used, only those
born within the range of years given (including the years specified) will
be included in the report. Save the macro as ModMacroSelect.sps.

c) For those with extra time: Modify the macro so that GROUP will accept
a list of variables, and will use them to create additional subdivisions in
the report. (Hint: the BY keyword must separate each grouping variable
from the next in the SUMMARIZE command).

Exercises E - 4