SAS Global Forum 2008 Foundations and Fundamentals

Paper 171-2008

Introduction to PROC TABULATE

Thomas J. Winn, Jr.
U.S. Department of Veterans Affairs, Austin, Texas

ABSTRACT
PROC TABULATE is used to build tabular reports containing descriptive statistical information, including
hierarchical relationships among variables. The code that is used to invoke PROC TABULATE is
complicated, and much of it looks quite different from other SAS procedures. Nevertheless, it is well worth
the necessary investment of time and effort for learning the intricacies and subtleties of its syntax. This
paper provides a simplified, step-by-step approach for coding PROC TABULATE.

INTRODUCTION
PROC TABULATE is used to build tabular reports containing descriptive statistical information, including
hierarchical relationships among variables. PROC TABULATE is the SAS® System’s implementation of TPL
(Table Producing Language), which was developed at the U.S. Bureau of Labor Statistics during the 1970s,
for generating tabular reports of descriptive statistics involving employment data.

PROC TABULATE is more powerful for producing tabulations than PROC FREQ, and it is a more flexible
statistical report writer than PROC MEANS. Although PROC TABULATE and PROC REPORT are both
capable of generating similar tabular reports in many situations, each of these procedures has strengths and
weaknesses. PROC TABULATE seems to be better for displaying hierarchical relationships. The syntax
used to invoke PROC TABULATE and PROC REPORT differ from one another, and both are complicated.
However, it is well worth the necessary investment of time and effort for learning the intricacies and
subtleties of coding both procedures. This paper will cover the fundamentals of coding PROC TABULATE.

The examples in this paper make use of the SAS® data files, prdsale (from the SAS 9.1 sashelp data
library), and empldata (an inner join of the empinfo, jobcodes, and salary SAS® data sets, from the SAS 9.1
sample data library).

GETTING STARTED WITH PROC TABULATE

The first thing the would-be PROC TABULATE programmer needs to do, before beginning to write any SAS
code, is to decide what the final report should look like. What is the intended design for the pages, rows,
and columns of the report?

Here are a few examples of uses of PROC TABULATE:

1
SAS Global Forum 2008 Foundations and Fundamentals

2
SAS Global Forum 2008 Foundations and Fundamentals

Unfortunately, a lot of the code that is used to invoke PROC TABULATE looks quite different from the code
that is used for other SAS procedures. Here is the basic syntax for coding PROC TABULATE:

PROC TABULATE <option-list>;

BY <NOTSORTED> <DESCENDING> variable-1
< … <DESCENDING> variable-n>;
CLASS class-variable-list;
CLASSLEV class-variable-list / STYLE=<style-element-name>
<[style-attribute-specification(s)]>;
VAR analysis-variable-list;
TABLE <<page-expression,> row-expression,>
column-expression </ table-option-list>;
FORMAT variable-list-1 format-1
< … variable-list-n format-n>;
FREQ variable;
KEYLABEL keyword-1=‘description-1’
< … keyword-n=“description-n’>;
KEYWORD keyword(s) / STYLE=<style-element-name>
<[style-attribute-specification(s)]>;
LABEL variable-1=‘label-1’ <…variable-n=‘label-n’>;
WEIGHT variable;
TITLE ‘text’;

Here are some options that are frequently used in PROC TABULATE statements:
DATA=SAS-dataset-name
FORMAT=formatname
MISSING
NOSEPS
The DATA= option specifies the SAS dataset to be used.
The FORMAT option specifies a default format for each table cell. The default format is overridden by any
format specified in a subsequent TABLE statement.
The MISSING option requests that missing values be regarded as valid levels for classification variables.
Unless the MISSING option is specified, observations with missing values for class variables will not be
included in the analysis.
The NOSEPS option removes the interior horizontal lines from the printed report.

STEPS FOR CODING PROC TABULATE

The simplest way to approach coding for PROC TABULATE may be expressed in terms of the following five
steps:
(1) Specification of classification variables,
(2) Specification of analysis variables,
(3) Definition of dimensions of the table,
(4) Identification of desired statistics, and
(5) Labeling and formatting the table.

WHAT ARE CLASSIFICATION AND ANALYSIS VARIABLES?

Classification variables are used to identify categorical groups on which calculations are performed. They
are the variables that make up the rows and columns of the table. They may be either character or numeric.
If numeric, generally only a small number of distinct values are permitted.

Analysis variables are numeric variables that are used to compute statistics that are reported in the body of
the table.

STEP 1 – SPECIFICATION OF CLASSIFICATION VARIABLES

The CLASS statement is used to specify any categorical variables that will be used for grouping purposes in
the analysis. The CLASS statement is required.

PROC TABULATE DATA=tabwkshp.prdsale;

CLASS country division product; . . .

3
SAS Global Forum 2008 Foundations and Fundamentals

STEP 2 – SPECIFICATION OF ANALYSIS VARIABLES

The VAR statement is used to list any variables that will be used for computing the statistics that are to
appear in the body of the table. Frequency counts can be computed without an analysis variable.
However, under most conditions, the VAR statement is required.

PROC TABULATE DATA=tabwkshop.prdsale;

CLASS country division product;
VAR actual; . . .

STEP 3 – DEFINITION OF THE TABLE’S DIMENSIONS

The TABLE statement is used to define both the arrangement of the rows and columns of the table, as well
as the requests for any summary statistics. It is the tricky aspect of using PROC TABULATE. So let’s
consider this complex statement one piece at a time.

In a TABLE statement, the comma is a very important symbol, because it separates the dimensions of the
table.
• If two commas were specified, then the table would have three dimensions, and the order
would be pages, rows, and columns.
• If only one comma was specified, then the table would have two dimensions, and the order
would be rows, columns.
• No comma would be interpreted to mean that the table’s only dimension would be the column
dimension. The table would only have one row.

So, TABLE statements look like the following:

TABLE page-expression, row-expression, column-expression . . .;

TABLE row-expression, column-expression . . .; or
TABLE column-expression . . .;

In this context, an expression can consist of variables, statistics, operators, format specifications, and label
assignments.

Because our immediate concern is only to define the table’s dimensions, we are pleased to discover that
only three operators are needed to specify the page, row, and column headings that identify the structure of
a table.
• An asterisk (*) can be used to cross the classification variables; that is, to arrange them in a
nested manner, according to the order listed (top, middle, and lower).
• A blank space is used to concatenate two classification variables (which will appear in the
table: top-to-bottom for row headings, left-to-right for column headings).
• Parentheses ( ) are used to group the elements of an expression, and to associate an adjacent
operator with each concatenated element inside the parentheses.

Here are a few simple examples:

TABLE var1, var2;

The preceding would generate a two-dimensional table in which the row dimension would be the
values of var1, and the column dimension would be the values of var2.

TABLE var1, var2 var3;

This would result in a two-dimensional table in which the row dimension would be the values of var1,
and the column dimension would be comprised of values resulting from the side-by-side concatenation
of var2 and var3.

TABLE var1, var2*var3;

This statement would generate a two-dimensional table in which the row dimension would be the
values of var1, and the column dimension would be a hierarchical arrangement of the values of var2
and var3, with the values of var2 comprising the top columns, and for each of these values as columns,
the values of var3 as the lower columns.

4
SAS Global Forum 2008 Foundations and Fundamentals

Here are three additional examples:

TABLE var1, var2 * (var3 var4) var5 ;

The preceding would generate a two-dimensional table in which the row dimension would be the
values of var1, and the column dimension would be two side-by-side components, the first of which
would be a hierarchical nesting of the values of var2 over a side-by-side arrangement of the values of
var3 and var4, and the second component would be the values of var5.

TABLE var1, (var2 var3) * var4 ;

This would result in a two-dimensional table in which the row dimension would be the values of var1,
and the column dimension would be two side-by-side hierarchical components, the first of which would
be the values of var2 over the values of var4, and the second would be the values of var3 over the
values of var4.

TABLE var1 var2, var3 * var4 ;

This code would generate a two-dimensional table in which the row dimension would be the
concatenated values of var1 and var2, and the column dimension would be a hierarchical nesting of
the values of var3 over the values of var4.

STEP 4 – IDENTIFICATION OF DESIRED STATISTICS

We have been considering incomplete TABLE statements (because up to this point we have only focused
on the dimensions of the tables).
Besides specifying the dimensions, the TABLE statement also identifies which summary statistics should be
produced, and pertaining to which analysis variables. Each statistic is identified by a keyword.
N = the number of observations, the frequency count
MIN = the smallest value
MAX = the largest value
MEAN = the arithmetic mean, or the average value
STD = the standard deviation
VAR = the variance
MEDIAN = the middle (50th percentile) value
SKEWNESS = a measure of the asymmetry of the distribution of values
SUM = the sum of the values
PCTN = the percentage that one frequency is of another frequency
PCTSUM = the percentage that one sum is of another sum
. . . (and other descriptive statistics).

Whenever you cross a variable with a keyword for a statistic, you are identifying the statistic to be applied to
that variable (which tells PROC TABULATE what type of calculation to perform). You can cross
classification variables only with the N or PCTN statistics. By default, if the TABLE statement does not
include an analysis variable or a statistic, then PROC TABULATE automatically crosses the N statistic with
the indicated class variables. Analysis variables can be crossed with any statistic. By default, if the TABLE
statement includes an analysis variable but without crossing it with any statistic, PROC TABULATE
automatically crosses it with SUM.

Here are some examples:

PROC TABULATE DATA=tabwkshp.empldata;

CLASS jobcode location gender;
TABLE jobcode, location*gender;
TABLE jobcode*PCTN, location*gender;
TABLE jobcode, location*gender*PCTN;
TABLE jobcode PCTN, location* gender;

The first TABLE statement would generate a hierarchical breakdown of frequency counts in the data set,
according to values of jobcode (the rows) and the nested values of location and gender (the columns).

5
SAS Global Forum 2008 Foundations and Fundamentals

The second and third TABLE statements would generate a hierarchical breakdown of percentages
represented in each cell, according to values of jobcode and the nested values of location and gender.

The fourth TABLE statement would generate a hierarchical breakdown of frequency counts represented in
each cell, according to values of jobcode and the nested values of location and gender. It would include an
additional row that would represent the percentage of the total population of the data set included in each
column.

6
SAS Global Forum 2008 Foundations and Fundamentals

Here are a few more examples:

PROC TABULATE DATA=tabwkshp.empldata;

CLASS jobcode location gender;
VAR salary;
TABLE jobcode, location*gender*salary;
TABLE jobcode, location*gender*salary*PCTSUM;
TABLE jobcode, location*gender*salary*MEAN;

The first TABLE statement would generate a hierarchical breakdown of the sum of the salary amounts
represented in each cell, according to values of jobcode and the nested values of location and gender.

7
SAS Global Forum 2008 Foundations and Fundamentals

The second TABLE statement would generate a hierarchical breakdown of the percentage of the total of
salary amounts represented in each cell, according to values of jobcode and the nested values of location
and gender.

The third TABLE statement would generate a hierarchical breakdown of the average salary represented in
each cell, according to values of jobcode and the nested values of location and gender.

8
SAS Global Forum 2008 Foundations and Fundamentals

PROC TABULATE has a universal class variable, ALL, which can be used to generate totals for any
specified class variable. Just concatenate the keyword ALL into the row or column expression of a TABLE
statement.

PROC TABULATE DATA=tabwkshp.empldata;

CLASS jobcode location gender;
TABLE jobcode ALL, location*gender ALL;
TABLE jobcode ALL, (location ALL)*gender;

What is the difference between the tables produced by these two TABLE statements? Here are the results
from the first TABLE statement:

9
SAS Global Forum 2008 Foundations and Fundamentals

And here are the results from the second TABLE statement:

CALCULATING PERCENTAGES WITH PROC TABULATE

PROC TABULATE has the capability of determining the percentage of the value in one cell to the value in
another cell, or to the total of a group of cells, through the PCTN and PCTSUM statistics, respectively.
Whenever you invoke PROC TABULATE to calculate percentages you, explicitly or implicitly, define the
value that is to be used as the denominator. If you do not specify a particular denominator then, by default,
PROC TABULATE uses the sum of the values in the N cells for the PCTN denominator, and the sum of the
values in the SUM cells for the PCTSUM denominator.

Brackets < > are used to explicitly specify the denominator that is to be used in the calculation of
percentages.

PROC TABULATE DATA=tabwkshp.empldata;

CLASS jobcode gender;
TABLE jobcode, gender*(N PCTN) ;
TABLE jobcode, gender*(N PCTN<gender>) ;

The first TABLE statement specifies that the display should include the number of instances of values for
gender occurring with each value of jobcode, and the percentages of those numbers to the total across all
combinations of values of jobcode and gender.

10
SAS Global Forum 2008 Foundations and Fundamentals

The second TABLE statement specifies that the display should include the number of occurrences of values
for gender, and the percentage of that number to the total for all values of gender in each jobcode (that is, a
row-percentage).

Here is another example of calculating percentages:

PROC TABULATE DATA=tabwkshp.empldata;

CLASS jobcode gender;
VAR salary;
TABLE jobcode, gender*salary*(SUM PCTSUM<jobcode>) ;

11
SAS Global Forum 2008 Foundations and Fundamentals

This TABLE statement specifies that the display should include breakdowns of the total salary amounts, and
the associated percentages, for each combination of values of jobcode and gender, where the percentages
are calculated in a column-wise manner.

Observe that, to obtain percentages by row, we use the column-expression in the “denominator definition”;
and to obtain percentages by column, we use the row-expression in the “denominator definition.”

And here are two more examples involving percentages, but these examples include the universal class
variable, ALL.

PROC TABULATE DATA=tabwkshp.empldata;

CLASS jobcode gender;
VAR salary;
TABLE jobcode, (gender ALL)*(N PCTN<gender ALL>) ;
TABLE jobcode ALL, gender*salary*(SUM PCTSUM<jobcode ALL>) ;

12
SAS Global Forum 2008 Foundations and Fundamentals

Notice that whenever row- or column-percentages are to be produced for a column- or row-expression that
includes the ALL universal class variable, then ALL also must be included in the “denominator definition.”

STEP 5 – LABELING AND FORMATTING THE TABLE

Now that we know how to define the basic structure of the tables we will generate, we would like to be able
to make the tables more self-explanatory; that is, easier to read and interpret.

As in many other SAS procedures, you can use a LABEL statement to replace variable names with more
descriptive headings for your class variables. There also is a way to specify temporary labels in a TABLE
statement.

Similarly, TITLE and FOOTNOTE statements also can be used to enhance the tabular reports generated by
PROC TABULATE.

To assign labels to procedure-generated statistics and the universal class variable, we use the KEYLABEL
statement.

KEYLABEL N = ‘Count‘
ALL = ‘Total’
PCTN = ‘Percent’;

As in other SAS procedures, formats can be used to substitute labels for values of the classification
variables. Formats also can be used to combine many values of the classification variables into a much
smaller number of values to be printed in the report. We create custom formats by using PROC FORMAT,
and we invoke those formats in PROC TABULATE either through a FORMAT statement, or by crossing
F=format-name. in the TABLE statement with the particular variable.

13
SAS Global Forum 2008 Foundations and Fundamentals

The default for displaying cells with missing numeric values is a period. You can change the way missing
values are displayed by using the MISSTEXT= option to define up to twenty characters of text that will print
in the table cells whenever a particular combination of class variable values is not found in the input data
set.

PROC TABULATE DATA=tabwkshp.empldata;

CLASS jobcode location gender;
TABLE jobcode, location*gender / MISSTEXT=‘None’;

Here are a couple of useful TABLE statement options that can be used for customizing the appearance of
tables:
• The RTSPACE= (or RTS=) option defines the total amount of space for the row headings. If there
are several levels of headings for rows, then the space is divided equally among the levels, after
subtracting the spaces that are needed for the vertical lines.
• Whenever a table produced by PROC TABULATE is too wide to fit on a single page, the procedure
automatically splits the table, to span as many separate pages as are necessary for printing. For
short, wide tables, the CONDENSE option could be specified on the TABLE statement, in order to
print as many logical pages as possible on a single page, one below the other.

Some people think that traditional SAS output is ugly. Beginning with Version 7, the SAS System provided
an ability to deliver procedure output in a flexible variety of file types and formats, through the SAS Output
Delivery System (ODS). Under SAS 9.1.3, ODS can be used to generate results as SAS data sets, output
listings, PostScript, HTML, RTF, PDF, PCL, XML, Excel, and other output file types. ODS can be used to
enhance tabular reports, by wrapping the PROC TABULATE code in ODS destinations, by changing fonts,
colors, and other style attributes, and by adding graphics. For further information about ODS and
TABULATE, consult the software documentation provided by SAS Institute.

Here is a final example, which illustrates several of the “Labeling & Formatting” techniques that I have
described:

PROC FORMAT;
VALUE salfmt low-<12000 = ‘Less than $12,000’
12000-<24000 = ‘$12,000 - $23,999’
24000-<48000 = ‘$24,000 - $47,999’
48000-<72000 = ‘$48,000 - $71,999’
72000-<96000 = ‘$72,000 - $95,999’

14
SAS Global Forum 2008 Foundations and Fundamentals

96000-<120000 = ‘$96,000 - $119,999’

120000-high = ‘$120,000 or more’;
RUN;

ODS HTML BODY=‘tables.htm’;

PROC TABULATE DATA=tabwkshp.empldata MISSING FORMAT=9.1;

CLASS title location gender salary;
FORMAT salary salfmt.;
LABEL title = ‘Job Title’;
KEYLABEL PCTN = ‘Percent’ ALL = ‘Total’;
TABLE title*salary ALL, (location*gender ALL)*PCTN
/ RTS=50 MISSTEXT=‘0’;
TITLE ‘Tabular Summary of Employee Information’;
RUN;
ODS HTML CLOSE;
RUN;

CONCLUSION
PROC TABULATE is a very useful and very powerful procedure for constructing tabular reports containing
descriptive statistical information, including hierarchical relationships among variables. It is well worth the
necessary investment of time and effort for learning the intricacies and subtleties of its syntax. Concentrating
on the five steps makes it much easier to learn how to code PROC TABULATE. We have only just
“scratched the surface” of this wonderful procedure! But now you know enough to continue learning about it
on your own. Happy tabulating!

RESOURCES FOR FURTHER READING

Jeffery M. Abolafia & Stephen M. Noga, “The TABULATE Procedure: One Step Beyond the Final Chapter,”
• SESUG ‘97 Proceedings (1997), pp. 293-300; and
• Proceedings of the Twenty-Third Annual SAS Users Group International Conference (1998),
pp. 839-844.

Jonas V. Bilenas, “Making Sense of PROC TABULATE (Updated for SAS9®),” Proceedings of the SAS
Global Forum 2007, Paper #230-2007.

15
SAS Global Forum 2008 Foundations and Fundamentals

Dan Bruns, “The Utter Simplicity of the TABULATE Procedure,”

• Proceedings of the Sixteenth Annual SAS Users Group International Conference (1991), pp.
365-371; and
• Proceedings of the Twentieth Annual SAS Users Group International Conference (1995), pp.
363-368; and
• Proceedings of the Seventh Annual South-Central SAS Users’ Conference (1997), pp. 54-60.

Dan Bruns, “The Utter ‘Simplicity?’ of the TABULATE Procedure,” Proceedings of the Seventeenth Annual
SAS Users Group International Conference (1992), pp. 216-220.

Dan Bruns, “Advanced Features of PROC TABULATE -- or -- The Utter Simplicity of the TABULATE
Procedure - The Sequel,” Proceedings of the Twenty-First Annual SAS Users Group International
Conference (1996), pp. 242-247.

Dan Bruns, “The Utter ‘Simplicity?’ of the TABULATE Procedure -- The Final Chapter,”
• Proceedings of the Twenty-Second Annual SAS Users Group International Conference (1997),
pp. 251-256 ; and
• Proceedings of the Seventh Annual South-Central SAS Users’ Conference (1997), pp. 61-66;
and
• Proceedings of the Twenty-Ninth Annual SAS Users Group International Conference (2004),
Paper #241-29.

Dan Bruns, “The Power and Simplicity of the TABULATE Procedure,”

• Proceedings of the Twenty-Fourth Annual SAS Users Group International Conference (1999),
Paper #152; and
• Proceedings of the Twenty-Fifth Annual SAS Users Group International Conference (2000),
Paper #152-25; and
• Proceedings of the Twenty-Sixth Annual SAS Users Group International Conference (2001),
Paper #148.

Dan Bruns, “The Simplicity and Power of the TABULATE Procedure,”

Proceedings of the Twenty-Eighth Annual SAS Users Group International Conference (2003),
Paper #197.

Dan Bruns & Ray Pass, “Battle of the Titans: REPORT vs. TABULATE,” Proceedings of the Twenty-Seventh
Annual SAS Users Group International Conference (2002), Paper #133-27.

Dan Bruns & Ray Pass, “To REPORT or to TABULATE?: That is the Question!,” Proceedings of the
Twenty-Ninth Annual SAS Users Group International Conference (2004), Paper #122-29.

Diane Louise Rhodes, “Speaking Klingon: A Translator’s Guide to PROC TABULATE,” Proceedings of the
Thirtieth Annual SAS Users Group International Conference (2005), Paper #258-30.

SAS Institute Inc., “The TABULATE Procedure,” in SAS® 9.1.3 Help and Documentation, an HTML
application that is installed as a component of the SAS System and is accessible from the SAS
main menu by clicking “Help.”

SAS Institute Inc., “The TABULATE Procedure,” Chapter 52 of Base SAS® 9.1.3 Procedures Guide (2006),
available online at
http://support.sas.com/documentation/onlinedoc/91pdf/sasdoc_913/
base_proc_8977_new.pdf .

SAS Institute Inc., SAS Guide to TABULATE Processing, Second Edition (Cary, NC: SAS Institute Inc.,
1990).

Bob Virgile, “The Right Approach to Learning PROC TABULATE,” SESUG ‘97 Proceedings (1997), pp. 189-
195.

Tom Winn, “Introduction to Using PROC TABULATE,”

• Proceedings of the Eighth Annual South-Central SAS Users’ Conference (1998), pp. 316-326,

16
SAS Global Forum 2008 Foundations and Fundamentals

• Proceedings of the New Mexico SAS Users Conference (2002),

• Proceedings of the Arkansas SAS Day (2005), pp. 14-33,
• Proceedings of the 2006 Wisconsin-Illinois SAS Users Conference .

Tom Winn, “Advanced Features of PROC TABULATE,”

• Proceedings of the Twenty-Fourth Annual SAS Users Group International Conference (1999),
Paper #153,
• Proceedings of the Ninth Annual South-Central SAS Users’ Conference (1999), pp. 243-248.

Tom Winn, “A Hands-On Introduction to PROC TABULATE,” Proceedings of the SCSUG SAS Educational
Forum 2007.

CONTACT INFORMATION
Thomas Winn, Ph.D.
U.S. Department of Veterans Affairs
DSS Support Office – Austin OI&T
1615 Woodward Street, 776 / 19F-4
Austin, TX 78772-0001

Telephone: 512 / 326-6180

E-Mail: [email protected]

SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute
Inc. in the USA and other countries. ®indicates USA registration.

17