The correct bibliographic citation for this manual is as follows: Matange, Sanjay, and Dan Heath. 2011.

Statistical Graphics Procedures by Example: Effective Graphs Using SAS®. Cary, NC: SAS Institute Inc.
Statistical Graphics Procedures by Example: Effective Graphs Using SAS®
Copyright © 2011, SAS Institute Inc., Cary, NC, USA
ISBN 978-1-60764-887-1 (electronic book)
ISBN 978-1-60764-762-1
All rights reserved. Produced in the United States of America.
For a hard-copy book: No part of this publication may be reproduced, stored in a retrieval system, or
transmitted, in any form or by any means, electronic, mechanical, photocopying, or otherwise, without the
prior written permission of the publisher, SAS Institute Inc.
For a Web download or e-book: Your use of this publication shall be governed by the terms established by
the vendor at the time you acquire this publication.
The scanning, uploading, and distribution of this book via the Internet or any other means without the
permission of the publisher is illegal and punishable by law. Please purchase only authorized electronic
editions and do not participate in or encourage electronic piracy of copyrighted materials. Your support of
others’ rights is appreciated.
U.S. Government Restricted Rights Notice: Use, duplication, or disclosure of this software and related
documentation by the U.S. government is subject to the Agreement with SAS Institute and the restrictions set
forth in FAR 52.227-19, Commercial Computer Software-Restricted Rights (June 1987).
SAS Institute Inc., SAS Campus Drive, Cary, North Carolina 27513-2414
1st printing, November 2011
®
SAS Publishing provides a complete selection of books and electronic products to help customers use SAS
software to its fullest potential. For more information about our e-books, e-learning products, CDs, and hard-
copy books, visit the SAS Publishing Web site at support.sas.com/publishing or call 1-800-727-3228.
®
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.
Other brand and product names are registered trademarks or trademarks of their respective companies.
Contents
Preface ix
Chapter 1 Introduction
1.1 Principles of Effective Graphics 3
1.2 Automatic Graphs from SAS Procedures 10
1.3 Graph Template Language 11
1.4 Statistical Graphics Procedures 11
1.5 Organization of This Book 12
1.6 Data Sets and Custom Styles 14
1.7 Color and Gray-Scale Graphs 14
1.8 Effective Graphics and the Use of Decorative Skins 15
1.9 SAS 9.2 and SAS 9.3 Features 15
Chapter 2 Statistical Graphics Procedures
2.1 Key Concepts 19
2.2 SGPLOT Procedure 22
2.3 SGPANEL Procedure 24
2.4 Combining Statements 29
2.5 SGSCATTER Procedure 30
2.6 Styles and Their Usage 32
2.7 Template-Based Graphics vs. Device-Based Graphics 33
Chapter 3 Common Graphs
3.1 Introduction 37
3.2 Single-Cell Graphs 38
3.3 Classification Panels 55
3.4 Comparative and Matrix Graphs 57
Chapter 4 Basic Plots
4.1 Introduction 61
4.2 SGPLOT Procedure 62
4.3 Plot Roles and Options 63
4.4 Scatter Plot 64
4.5 Scatter Plots with Data Labels 74
iv Contents

4.6 Series Plot 78

4.7 Step Plot 84
4.8 Band Plot 89
4.9 Needle Plot 94
4.10 Vector Plot 97
4.11 VBarParm and HBarParm Plots (9.3) 100
4.12 Bubble Plot (9.3) 105
4.13 HighLow Plot (9.3) 108
4.14 Reference Lines 112
4.15 Parametric Line Plot (9.3) 115
4.16 Waterfall Chart (9.3) 117
4.17 Combining the Plots 120
Chapter 5 Fit and Confidence Plots
5.1 Introduction 129
5.2 Fit Plot Roles and Options 130
5.3 Regression Plot 131
5.4 Loess Plot 138
5.5 Penalized B-Spline Plot 145
5.6 Ellipse Plot 152
5.7 Combining the Plots 157
Chapter 6 Distribution Plots
6.1 Introduction 161
6.2 Histogram 162
6.3 Density Plot 166
6.4 Vertical Box Plot 169
6.5 Horizontal Box Plot 174
6.6 Combining the Plots 177
Chapter 7 Categorization Plots
7.1 Introduction 181
7.2 Categorization Plot Roles and Common Options 182
7.3 Vertical Bar Charts 183
7.4 Horizontal Bar Charts 191
7.5 Vertical Line Charts 198
7.6 Horizontal Line Charts 205
 Contents v

7.7 Dot Plots 212

7.8 Combining the Plots 218
Chapter 8 Axes, Legends, and Insets
8.1 Introduction 221
8.2 Linear Axis 223
8.3 Log Axis 225
8.4 Time Axis 227
8.5 Discrete Axis 229
8.6 Legends 230
8.7 Insets 232
Chapter 9 Annotation and Attribute Maps (SAS 9.3)
9.1 Annotation 235
9.2 Attribute Maps (9.3) 251
Chapter 10 Classification Panels
10.1 Introduction 257
10.2 SGPANEL Procedure 258
10.3 PANELBY Statement 259
10.4 Classification Panels 261
10.5 Paging of Large Panels 266
Chapter 11 Comparative and Matrix Plots
11.1 Introduction 269
11.2 SGSCATTER Procedure 269
11.3 PLOT Statement 270
11.4 COMPARE Statement 276
11.5 MATRIX Statement 281
Chapter 12 Health and Life Sciences Graphs
12.1 Introduction 287
12.2 Forest Plot 288
12.3 Forest Plot (9.3) 289
12.4 Survival Plot 290
12.5 Adverse Event Timeline 291
12.6 Adverse Event Timeline (9.3) 292
vi Contents

12.7 Maximum LFT Values by Treatment (9.3) 293

12.8 Median of Lipid Profile over Time by Treatment (9.3) 294
12.9 QTc Change from Baseline over Time by Treatment (9.3) 295
12.10 QTc Change Graph with Annotated “At Risk” Values (9.3) 296
12.11 QTc Change from Baseline over Time by Treatment 297
12.12 LFT Safety Panel, Baseline vs. Study 298
12.13 Immunology Profile by Treatment 299
12.14 Most Frequent On-Therapy Adverse Events by Frequency 300
12.15 LFT Patient Profile 301
12.16 Panel of LFT Values 302
12.17 Distribution of Eye Irritation Using PROC SGPANEL (9.3) 303
12.18 Distribution of Eye Irritation Using PROC SGPLOT (9.3) 304
12.19 Vital Signs by Time Point Name 305
12.20 Concomitant Medications 306
12.21 Creating a 2 x 2 Cell Graph Using PROC SGPLOT 307
Chapter 13 Business Graphs
13.1 Introduction 311
13.2 Stock Price and Volume Chart 312
13.3 Financial Trend and Bond Maturity Graph (9.3) 313
13.4 Danger of High P/E Ratios 314
13.5 Oil Consumption Trend by Country 315
13.6 Product Sales and Target Graph (9.3) 316
13.7 Social Network (9.3) 317
Chapter 14 Styles
14.1 Introduction 321
14.2 Using Styles 321
14.3 Style Elements 321
14.4 Using Style Elements 324
14.5 Style Element Usage Precedence 324
Chapter 15 ODS Destination and ODS Graphics Options
15.1 Introduction 327
15.2 ODS Destination Options 327
15.3 ODS Graphics Options 328
 Contents vii

Chapter 16 Tips for Graph Output

16.1 Introduction 333
16.2 Creating Small Graphs for Use in Documents 333
16.3 Creating Large Graphs for Use in Presentations 334
16.4 Combining Graph Size and DPI 336
16.5 Impact of Graph Size on System Resources 336
viii
Preface
As we looked with anticipation to the first release of the Statistical Graphics Procedures in SAS
9.2, it became evident to us that these procedures truly represented a new way of creating
analytical graphs. We started thinking about the best ways to communicate the features and
capabilities of these new procedures. While we discussed different approaches, we found
ourselves receiving a lot of questions from users about the procedures. Often our answers to
them were in the form of examples. As we created more examples in such areas as clinical
safety, pharmaceuticals, health care, and finance, we soon realized that the best way to
communicate the features of these procedures was through such examples.

The question we often got from users was simply, “How do I make this graph?” The best way for
us to answer that question was to show the code needed to create it. If a book could answer that
question, what would it look like? Why not create a book that starts from the end result and work
backward to show exactly what is needed to create the graph? That led to the idea of writing this
book.

The primary audience for this book is the SAS user who wants to visualize raw data or create a
graph from the results of a custom analysis. Often, you already have a mental image of the graph
you want to create; you just need to quickly find the correct syntax. Product documentation and
books on the topic often take a procedure-centric approach and describe the features one at a
time. Figuring out what is needed to create a graph from such resources requires a solid
understanding of the procedure, and it can take a while to obtain the right results. It would be
much easier if we start with the graph you want, and show you the code instead.

This book addresses this situation by using examples to document the procedure options. Users
can look through the large number of graph examples and find the type of graph they want to
create. Each of the graph examples includes the code needed to generate the graph, along with
a brief commentary on key features of the graph.

The reason why this approach works so well for the SG Procedures is because these procedures
take a building-block approach. You start with the basic plot, and simply add the features you
need one at a time. The procedures support a wide array of plot types, so the combinations and
possibilities grow rapidly. For example, if you know how to build a simple series plot, then
creating a plot with three series plots becomes straightforward. You simply add two more series
statements, and the procedure automatically does the work. The same principle also applies to
combinations of disparate, but compatible, plot types, such as a bar chart and a line chart.

The book also describes, by example, other important features of the procedures such as axes,
insets and legends. For example, you can take a sample graph with a linear axis, and change it
to log. You can also create a custom axis displaying only the values you want, add insets and
customize the legends.

For an overview of the organization of this book, see section 1.5.

x Preface

The SG Procedures are designed with the principles of effective graphics built-in to convey the
information with maximum clarity and minimum clutter. By default, these procedures will create
graphs that are free of unnecessary clutter in the graph elements, legends, and axes.

These procedures are designed to create graphs that are suitable for the statistical and analytical
use cases. Such graphs emphasize the maximization of “chart ink” and removal of elements that
are not clearly necessary for the delivery of the information. It will be evident that you have to do
very little to get aesthetically pleasing graphs for these use cases.

Visual aesthetics, however, are in the eye of the beholder. In non-statistical use case, the
expectation of the consumer is for flashy graphs, even at the cost of some effectiveness. One
person’s “chart junk” is another person’s “cool”. These procedures are finding increasing usage
for the creation of graphs for the business domain. They support some options to add “flash” to
these graphs. These options are mainly available for the bar charts and can be used when
necessary.

SG Procedures support full-color graphics. Many styles shipped with SAS are optimized for
creation of color graphs. However, printing color graphs in gray scale can sometimes lead to
undesirable results. This is important, since many technical journals are printed in gray scale.
Since this book is printed in gray scale we have used the appropriate gray scale styles.

The examples and techniques discussed in this book will be relevant and useful for all SAS users.
This is particularly so for statisticians and other analytical users in the pharmaceutical, clinical
trials, health care, financial, and other domains. This book is focused on how to create the
required graph given the data. Techniques for modeling and analysis of the data itself are
beyond the scope of this book.

Acknowledgements
Many people have contributed in many different ways to make this book possible. We would like
to thank Bob Rodriguez, Senior Director in Advanced Analytics at SAS, for supporting the
concept behind this book and for his detailed review of the contents. His insightful suggestions
have significantly improved the quality of the materials and the presentation of the book.

On the contents of the book and accuracy of the information, we received invaluable support from
Melisa Turner and Susan Schwartz, our “eagle eye” team of technical reviewers. Both Melisa
and Susan invested many days reviewing the code and improving the contents. We also thank
our reviewers Lelia McConnell, David Schlotzhauer, Peter Christie, and Rebecca Ottesen for their
valuable suggestions. Our heartfelt thanks go to Susan Slaughter for her review, moral support,
and guidance on this project. The table of Statement Combinations in section 2.4 is modeled
after a similar table from Susan and Lora Delwiche’s recent paper, “Using PROC SGPLOT for
i
Quick High-Quality Graphs.”
 Preface xi

Last and most importantly, we thank our families for the understanding and support they provided
while we spent long evenings and weekends on this project.

Author Pages
Each SAS Press author has an author page, which includes several features that relate to the
author including a biography, book descriptions for coming soon titles and other titles by the
author, contact information, links to sample chapters and example code and data, events and
extras, and more.

You can access our author pages at http://support.sas.com/publishing/authors/matange.html and

http://support.sas.com/publishing/authors/heath.html.

Comments or Questions?

If you have comments or questions about this book, you may contact the authors through SAS as
follows:

Mail:

SAS Institute Inc.

SAS Press
Attn: <Author’s name>
SAS Campus Drive
Cary, NC 27513

E-mail: [email protected]

Fax: (919) 677-4444

Please include the title of the book in your correspondence.

For a complete list of books available through SAS Press, visit support.sas.com/publishing.

SAS Publishing News:

Receive up-to-date information about all new SAS publications via e-mail by subscribing to the
SAS Publishing News monthly eNewsletter. Visit support.sas.com/subscribe.

i
Available at http://support.sas.com/resources/papers/proceedings09/158-2009.pdf.
xii
Chapter 1
Introduction
1.1 Principles of Effective Graphics 3
1.2 Automatic Graphs from SAS Procedures 10
1.3 Graph Template Language 11
1.4 Statistical Graphics Procedures 11
1.5 Organization of This Book 12
1.6 Data Sets and Custom Styles 14
1.7 Color and Gray-Scale Graphs 14
1.8 Effective Graphics and the Use of Decorative Skins 15
1.9 SAS 9.2 and SAS 9.3 Features 15
2 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“Then there is the man who drowned crossing a stream

with an average depth of six inches.”
~W.I.E. Gates
 Chapter 1 Introduction 3

Chapter 1: Introduction
Graphs are an essential part of modern data analysis. From clinical trials to quality control,
effective graphs are integral to the analysis process. Large quantities of data are collected for
clinical drug trials for safety, retail sales, warranty claims, medical lab results, and financial
transactions. Analysis of this data often relies on review of the data in tabular form. Viewing the
data in the form of a graph along with results of the statistical analysis of the data on the same
graph can significantly enhance the understanding of the data and the results.

A key aspect of this process is the ability to create an effective graph that can communicate the
raw data along with the statistical analysis results in a clear and concise form. These graphs can
help the analyst to visualize the trends and patterns in the data and the associations between
variables that are not evident in tabular form. Such insights can guide the direction of further
questions and formulation of additional testing methods and gathering of more focused data.

1.1 Principles of Effective Graphics

Research in the field of visual perception has guided the formulation of the principles for
creation of effective graphs. The Visual Display of Quantitative Information (2001) by Edward
Tufte, Visualizing Data (1993) by William S. Cleveland, and Creating More Effective Graphs
(2004) by Naomi Robbins provide an in-depth coverage of the topic. Some of these
principles are reviewed here.

A graph is considered effective if it

conveys the intended information in a
way that can be understood quickly and
without ambiguity by most consumers.
Figure 1.1 shows an interesting graph of
the levels of usage of text messaging on
cell phones by teenagers. The font sizes
are increased for readability.

In Figure 1.1, the consumer would be

hard pressed to discern the differences
(if any) between individual pie slices
such as the second and third slice Figure 1.1
clockwise from the top. The pie chart
does not facilitate such comparison of magnitude between individual categories.

This graph is further complicated by addition of the cumulative strips along the outside of the
pie chart. One reason why a pie chart is not an effective graph is the difficulty of making
magnitude comparisons when the data is plotted as an angle from a common (or non-
common) base. This applies to the strips and the slices.
4 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 1.2

Comparison of magnitude along a linear scale from a common base is very reliable. The
same data rendered as a bar chart is much easier to decode as shown above in Figure 1.2.
nd rd th
In this graph, the comparisons between the 2 , 3 and 4 bars for the % case are much
easier and more reliable.

The pie chart can be a useful visual for some use cases as shown in Figure 1.3. This graph
shows the portion of sales for Auto as a fraction of the total sales. The pie chart can work
well for visualizing such “part-to-whole” comparisons.

1.1.1 Visual Perception

The concepts of attentive and pre-attentive vision

are summarized by Daniel Carr in Information
Visualization: Perception for Design (2004) by
Colin Ware. Attentive vision requires scrutiny of
the object and an active participation on the part
of the observer.

On the other hand, pre-attentive vision allows

processing to be done prior to conscious
attention. Usage of pre-attentive features helps
in rapid discrimination between various artifacts
of a graph.
Figure 1.3
Differentiation between length of lines is pre-attentive regardless of a common baseline.
 Chapter 1 Introduction 5

In Figure 1.4, it is relatively easy to compare the magnitudes of the responses for all drugs.
The line segments in Figure 1.5 are plotted from different baselines, but still it is possible to
compare the magnitudes of each line segment for the drugs.

Figure 1.4 Figure 1.5

Differentiation of marker shapes and line patterns is pre-attentive. Groups can be easily
differentiated when marker shapes or line patterns are used as grouping indicators as
shown in Figures 1.6 and 1.7.

Figure 1.6 Figure 1.7

Based on their pre-attentive nature (ease of decoding) we conclude the following:

o Distances from a common baseline are useful for magnitude comparisons.

o Marker shape is effective for classification.
o Line pattern is effective for classification.
o Color is also an effective tool for classification, but since this book is published in
gray scale, we have not utilized that feature.
6 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

1.1.2 Accuracy of Magnitude Perception

Stevens' power law proposes a relationship between the magnitude of a physical stimulus
and its perceived intensity or strength. The law proposes that the accuracy of magnitude
perception for visual length is linear. Magnitude of a line twice as long is perceived almost as
twice as long. However, the accuracy of perception of magnitude is reduced for other
representations. For area, it is only about 1.6. That is, an area twice as large only seems
like 1.6 times as large. So, we tend to underestimate areas.

For representation of magnitude, we can conclude the following:

o Linear distance on a common scale and base line is most effective.

o Linear distance on a common but non-aligned scale is effective.
o Areas are only about 70% as effective as linear distance.
o Volumes are only about 60% as effective as linear distance.
o Angular distance from a common base is not very effective.
o Angular distance from non-common base is ineffective.
o Color in general is not an effective representation of magnitude.
o Color intensity can be used as a relative measure of magnitude.
o Color hue is not a good representation of magnitude.
o 3-D plots distort the perception of absolute magnitude.

Figure 1.8 Figure 1.9

Figure 1.8 and Figure 1.9 both display the mean MPG by type of car. Figure 1.8 uses the dot
plot where the response values are plotted as a linear distance from a common baseline. It is
very easy for the eye to decode the relative values of each car type. Grid lines help to line up
the values.

Figure 1.9 displays the same data using a pie chart. Clearly, it is much harder to decode the
relative values since it is difficult to make good magnitude comparisons using angular
distances, especially from different baselines.
 Chapter 1 Introduction 7

1.1.3 Usage of 3-D Graphs

Figure 1.10 and Figure 1.11 both display the mean MPG by type of car. Figure 1.10 uses a
vertical bar chart to display the data. It is very easy for the eye to decode the relative values
for each car type. A format is applied to the response column to reduce the clutter for the
data label.

Figure 1.10 Figure 1.11

Figure 1.11 displays the same data using extruded 3-D bars. This is often referred to as a
2.5-D graph, since the data itself has only two dimensions. The third dimension is artificially
added to make the bars appear like 3-D blocks. Usage of such aesthetic features can
sometimes inhibit the process of decoding the data accurately.

There are several potential pitfalls in the 2.5-D representation of the data. The axis values
are displayed on the left at the “front” face of the bars. The grid lines are drawn along the
side and back face of the graph. So to measure the value for each bar, one has to line up the
correct face (front or back).

Often in such representations, the bars do not occupy the full depth of the walls, thus leaving
room for confusion. Even though the bar values are displayed in the 2.5-D case, some
values, such as for the Truck category, can become partially hidden behind other bars.

In general the following conclusions can be drawn:

o Dot plots, needle plots, and bar charts are good for representations of magnitude.
o Pie charts and area plots are not ideal for representations of magnitude.
o Color intensity is not an effective representation of magnitude.
o 3-D representations are often not effective when the data are 2-D.
o Unobtrusive grid lines can help in the decoding of the data.
8 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

1.1.4 Proximity Increases the Accuracy of Comparisons

When you compare magnitude between categories, closer proximity increases the accuracy
of comparison. For an effective graph, it helps to bring the items that are to be compared as
close to each other as possible.

Figure 1.12 Figure 1.13

Figure 1.12 is suitable for comparison of MPG for sedans and sports cars manufactured in
different regions (origin). The comparison between car types is facilitated by bringing these
categories closer in proximity. In this graph it is harder to compare “Sedan” from USA with
“Sedan” from Asia. Figure 1.13 is more suitable for such a comparison where “Origin” is
used as the category role.

1.1.5 Simplify and Reduce Clutter

Edward Tufte’s principles for the creation of effective graphics include the following
recommendations:

1. Eliminate “chart junk.”

2. Maximize “data ink.”

Often when creating graphics for marketing and sales presentations, there is a desire to
make the graph visually “compelling”. To add this “Wow” factor, visual elements may be
added to the graph to make it more aesthetically appealing. If one is not careful, these
artifacts may introduce distractions or, worse, actually distort the data, making it harder to
decode the data accurately.

Effectiveness of a graph can be enhanced by removing unnecessary artifacts from the graph.
Avoid usage of gradient background and images. Inclusion of embellishments like drop
shadows for data markers can increase the visual appeal of the graph but can reduce the
effectiveness of the graph.
 Chapter 1 Introduction 9

1.1.6 Short-Term Memory

Generally, people find it difficult to absorb and retain a large number of data values at one
time. Short-term memory is limited. Arranging the data in smaller chunks can aid in the
processing of information. Research in this field shows:

o An average person can easily remember 3–5 chunks of information.

o An average person can mentally calculate with 2-digit numbers.
o Excessive eye movement reduces efficiency of decoding a graph.

Figure 1.14 Figure 1.15

Both of the graphs above display revenues by product over time. The data are grouped by
product, each series representing one product. Figure 1.14 uses a traditional legend at the
bottom of the graph to identify each product.

To compare revenues for desks and chairs, you have to move your eyes down to the legend
and then back up to the plot. Figure 1.15 uses direct labeling for each series. This
eliminates eye movement and thus facilitates easier comparisons of the data.

1.1.7 Summary

You can use the above-mentioned guidelines to create graphs that convey information with
maximum effectiveness and minimum distractions. In summary, we suggest that you:

o Use linear distances from a common baseline to represent magnitude.

o Use marker shapes or line patterns for grouping.
o Simplify the appearance of the plot and reduce unnecessary ink.
o Increase proximity of data for better comparisons.
o Reduce eye movement needed to decode the data.

The SG Procedures provide the features you need to implement the above guidelines.
10 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

1.2 Automatic Graphs from SAS Procedures

Now let us see how you can get effective graphs using SAS. Starting with SAS 9.2, many
SAS analytical procedures use the Output Delivery System (ODS) Graphics system to
automatically produce graphs. When the ODS Graphics system is enabled, such graphs are
produced along with the data tables in the right order, and included in the output file such as
a PDF document.

To create the graphs from SAS 9.2 analytical procedures, you only need to switch on the
ODS Graphics system before running the procedure. This is done by including the following
statements in your program. No additional graph code is required.

ods graphics on / < options >;

< procedure statements; >
ods graphics off;

ods html;
ods graphics on;

ods select 'Analysis of

Variance';
ods select 'Fit Plot';

proc reg data=sashelp.class;

model Weight = Height;
quit;

ods graphics off;

ods html close;

Figure 1.16 Figure 1.17

Figure 1.16 shows the usage of ODS Graphics with an example of the REG procedure. In
this example, we have specifically requested the output of the Analysis of Variance table and
the Fit Plot by using the ODS SELECT statements. The resulting HTML output is shown in
Figure 1.17. The Analysis of Variance table and the Fit Plot are produced in the right
sequence in the output HTML file.
 Chapter 1 Introduction 11

It is worth repeating that the Fit Plot is produced automatically, without any graph coding
required on the part of the user. It is useful to note that the procedures have to run
additional processing steps to create these graphs. Some procedures may create a large set
of graphs, which is something to consider when using this option.

With SAS 9.3, running in DMS mode, the default open destination is HTML, and ODS
Graphics is on by default. For line mode, the default open destination is LISTING, and ODS
Graphics is off by default. This is a change from SAS 9.2, where the default open destination
is LISTING, and ODS Graphics is off by default.

1.3 Graph Template Language

The graphs mentioned in section 1.2 are created automatically by the procedure and do not
require any additional programming by user. This is done by using predefined graphics
templates that use the Graph Template Language (GTL). All templates needed to create
these graphs have been supplied with the software. The graph shown in Figure 1.17 is
created by the REG procedure using one of the predefined templates shipped with SAS.

GTL can also be used directly by you, the SAS user, to create your own custom graph
template. Then, you can use the SGRENDER procedure to associate this template with the
appropriate data to produce the resultant graph.

A detailed description of GTL and the SGRENDER procedure is beyond the scope of this
book. However, it should be noted that all graphs created by ODS Graphics system are done
using the GTL syntax at some level. This is also true of the graphs produced by the
Statistical Graphics (SG) procedures, which is the topic of this book.

1.4 Statistical Graphics Procedures

The SG procedures are a set of procedures that work within the ODS Graphics system.
These procedures are designed to provide the familiar procedure syntax for creation of
graphs that are most commonly used in various industries such as health and life sciences,
finance, banking, quality control, and more.

The graphs are created using GTL behind the scene. So, these graphs have the same look
and feel as the automatic graphs created from the SAS analytical procedures. These graphs
are useful for visualization of the raw data or for custom graphs of analysis results.

1. Pre-analysis Data Exploration

When you receive the data from a survey or a study, often you may want to get multiple
graphical views of the raw data before the analysis phase. Viewing the data using simple
scatter plots, histograms, and scatter plot matrices can provide the analyst valuable
insights into the data which can help in the analysis phase of the project.
12 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

2. Analysis of Data
This phase of the project involves the analysis of the data using analytical procedures
and/or your own custom data steps. Automatic graphs can be obtained from individual
procedures as mentioned in section 1.2. In this phase you may also need to create
specialized graphs that are not currently supported by the procedure itself.

3. Post-analysis Data Presentation

After the analysis phase of the project, you may need to present the results of the
analysis in a form that is easily consumed by your audience. This goal is best achieved
by presenting the results in the form of graphs that include the original data along with the
analytical results. If the analysis requires multiple-procedure steps or custom data steps,
it may be necessary to create custom graphs from the results.

For all steps in the process above, you need the ability to create graphs from the raw data or
from the results of your custom analysis. You may also want to use the graphs that are
automatically created for you by individual procedures in the report for the project. In this
case, the SG procedures are the ideal tools for this job for the following reasons:

o Graphs created by the SG procedures are identical in look and feel to the automatic
graphs created by the analytical procedures. Mixing and matching the output from
the SG procedures and the analytical procedures is seamless.

o SG procedure steps can be run along with the analytical procedures and data steps
to produce a sequential output in the open ODS destination.

o SG procedures provide a simple and concise syntax to create many types of graphs,
classification panels, and scatter plot matrices.

o With SAS 9.3, SG procedures also provide the ability to annotate the graph with
Annotate-like functionality using a data set. Additionally, attribute maps can be used
to control the usage of visual attributes like color or marker symbols in the graph.
These topics will be discussed in detail in Chapter 9.

1.5 Organization of This Book

The approach taken for this book is to present the features of the SG procedures via
examples. A textbook on this topic would normally take you through all the features of the
procedures, and you would have to know many aspects of the procedures to be able to
generate a useful graph.

Instead of listing all the options and features of the procedures, we take the reverse
approach. If you have an idea of the graph you want to make, you can just flip through this
book and find the graph closest to what you need. Then, right alongside, you will find the
code necessary to create the graph. From there, you can build on the graph by borrowing
from other examples in the book.
 Chapter 1 Introduction 13

SG procedures utilize a building-block approach to creating a graph. If you see two graphs
that each individually include elements that you want in one graph, it is highly likely that you
can combine the statements in one procedure step and get the combined graph you need.

For example, you can combine a scatter plot, a series plot, and various regression plot
statements from different examples into one procedure step. Some common combinations
are as follows:
o Scatter, Series, Step, Band, Regression, Ellipse, VBarParm, and HBarParm
o Histogram and Density
o Bar Chart, Line Chart, and Dot Plot

This book is organized as follows:

o In Chapter 2, we start with a general description of each procedure. This will show
you the structure of the syntax and the main features with a few examples. From
there on, we focus on examples, starting with single-cell graphs, and then moving on
to more complex cases.
o In Chapter 3, we review graphs that are commonly used in various domains. This
section covers the different graph types you can create. We defer the detailed
discussion of the plot options to subsequent chapters.
o In Chapters 4–7, we cover the main groups of single-cell graphs using the SGPLOT
procedure. Plot statements used in these graphs can be combined within the groups
to create the graph you need. Various supported options are used to demonstrate
the features.
o In Chapter 8, we cover common customizations for axes, legends, and insets.
o In Chapter 9, we cover the topics of annotation and attribute maps. Annotations
allow you to add custom graphical elements to a graph that may or may not be data
driven. Attribute maps provide you the ability to tie the plot attributes like color,
symbols, and line patterns to explicit data values. These powerful features for
detailed customization of graphs are included with SAS 9.3.
o In Chapter 10, we cover classification panels using the SGPANEL procedure. This
topic leverages all you have learned about single-cell graphs to produce graphs that
are classified by multiple class variables.
o In Chapter 11, we cover comparative scatter plots and scatter plot matrices using the
SGSCATTER procedure.
o In Chapter 12, we cover graphs commonly used in the health and life sciences
industry.
o In Chapter 13, we cover some special business graphs. Here you will find detailed
examples that combine features from previous chapters to create the graph.
14 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

o In Chapter 14, we cover the topic of styles. Here you will see the inner workings of
styles and the association between style elements and graph features. We will cover
the basics of creating your own custom style for graphs.
o In Chapter 15, we cover the options on the ODS DESTINATION statement that have
a direct bearing on the rendering of the graphs. We also review the options you can
set on the ODS GRAPHICS statement to control aspects of graph rendering.
o In Chapter 16, we cover how to create graphs appropriate for different use cases.
Often, graphs are created for inclusion in a full slide of a Microsoft PowerPoint
presentation, or in one 3-1/4” column of a Microsoft Word document for a printed
journal. We will provide some tips on how to create graphs that are suitable for such
use cases.

1.6 Data Sets and Custom Styles

Most of the examples in this book use the pre-defined SAS data sets available in the
SASHELP library. These include CARS, HEART, and a few others. Often, to reduce the
number of classifiers, so that a graph will fit in the restricted space, modified data sets are
used that contain a subset of the data from these data sets. In other cases, custom data sets
are needed that are suitable for the example graph.

Custom styles are sometimes used to render some of the graphs in this book. Primarily,
these are necessary to reduce the font sizes to help fit the graphs into the small space
available. The results you see may vary based on the active style for an ODS destination.

1.7 Color and Gray-Scale Graphs

The graphs created by the SG procedures, and by the ODS Graphics system in general, use
the active style of the open destination. Often, these styles are optimized for full color output.
This works well when the graph is also consumed in a color medium.

However, when color graphs are printed in gray scale, there is a significant loss of fidelity in
the representation of distinct categories in the graph. For example, a graph with two series
plots, one for Drug A and one for Drug B can be well represented in color with use of two
distinct colors, say red and blue. These colors are often designed to have equal weight to
avoid unintentional bias.

When such a graph is printed in gray scale, these two series plots may look very similar
unless they have other distinguishing features such as line patterns and marker shapes to
facilitate discrimination between groups. Bar charts can benefit from use of fill patterns to
facilitate such discrimination.

This book is printed in gray scale, so it is important to create the graphs that will print well in a
gray-scale format. To ensure this, it is best to create the original graph in the gray-scale
format that maximizes the discriminability of the different categories and groups. All of the
 Chapter 1 Introduction 15

graphs included in this book are created using gray-scale styles such as Journal, Journal2 or
Journal3, or styles derived from these styles.

When you run a program from this book, or one of your own, the graph will be rendered using
the active style of the open destination. For SAS 9.2, this is the LISTING destination. For
SAS 9.3 in DMS session, this is the HTML destination. Since both of these destinations use
a default color style, you will get a graph rendered in full color. To get a gray-scale graph,
use one of the styles mentioned above.

The WIDTH or HEIGHT options on the ODS GRAPHICS statement have been used to render
the graphs for this book. However, these options are not shown in the sample code. When
you run the same code without these options, the graphs will render in the default size.

1.8 Effective Graphics and the Use of Decorative Skins

As alluded to earlier, the SG procedures specifically, and ODS Graphics in general, are
designed with the principles of effective graphics in mind. By default, the procedures always
strive to create a graph that delivers the information with maximum clarity and minimum
clutter or distraction.

Though initially designed with the statistical user in mind, these procedures are finding
increasing usage in non-statistical domains. In these use cases, there is often a desire for a
flashier graph, even at the expense of effectiveness.

In the SG procedures, the bar chart statements support an option to apply a decorative skin
to a bar. This does not change the shape of the bar but provides a “flashier” rendering. This
option can be used at the discretion of the user. Some examples in this book use this option
to demonstrate this feature. The intention is primarily to expose these available features.

1.9 SAS 9.2 and SAS 9.3 Features

This book assumes the user has the SAS 9.2 (TS2M0) or higher release. The examples in
this book include the usage of SAS 9.2 (TS2M0) and SAS 9.3 features. Examples that use
the features added in SAS 9.3 are marked with the 9.3 icon, and the specific statement or
option is displayed in bold italics.

In some cases you can remove the new option and still run the code using SAS 9.2.

Note: When running SAS 9.3 in DMS mode, the default open destination is HTML. For non-
DMS mode, the default open destination is LISTING. The ODS GRAPHICS feature is
automatically enabled for the execution of the SG procedures.
16
Chapter 2
Statistical Graphics Procedures
2.1 Key Concepts 19
2.2 SGPLOT Procedure 22
2.3 SGPANEL Procedure 24
2.4 Combining Statements 29
2.5 SGSCATTER Procedure 30
2.6 Styles and Their Usage 32
2.7 Template-Based Graphics vs. Device-Based Graphics 33
18 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“Simple things should be simple.

Complex things should be possible.”
~ Alan Kay
 Chapter 2 SG Procedures 19

Chapter 2: Statistical Graphics Procedures

In this chapter we introduce the key concepts for the SG procedures and the general syntax.

The SG procedures provide a direct procedure interface into the ODS Graphics system. These
procedures create graphs with very little code. The SGPLOT and SGPANEL procedure syntax
allows you to build up complex graphs by use of plot statements and other features. The
SGSCATTER procedures syntax is designed to give single-statement access to a variety of
scatter plot panels and matrices.

2.1 Key Concepts

The graphs created by these procedures can be grouped into three broad categories: single-
cell graph, multi-cell classification panels, and multi-cell comparative scatter plots. These
categories are discussed in the following sections.

2.1.1 Single-Cell Graph (SGPLOT Procedure)

Legend

Data
Area

Plots

Graph

Figure 2.1

Figure 2.1 shows a typical single-cell graph built using the following components:
o one or more titles at the top of the graph.
o one or more footnotes at the bottom of the graph
o one region in the middle displaying the data
o one or more plots in the data area
o one or more legends or insets inside the data area or outside

Note: All statements are referred to as “plot”, regardless of whether it is a series plot,
scatter plot, bar chart, histogram, and so on.
20 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Graph: Refers to the individual output created by the procedure. In most of the common
use cases, each execution of the procedure creates one graph output file. Often these
procedures produce multiple output files (for BY variable usage, or paging of large panels),
each of which is referred to as a “Graph”.

Cell: Each graph can have one or more data areas where the data is plotted as shown in
Figure 2.2. Each of these is referred to as a “Cell”. A cell may or may not include axes.

Plot statements: Each plot statement is responsible for drawing only its own data
representation. It is told by the container where to draw itself and how to scale its data.

Axes: The X and Y axes are shared by all the plots in the graph. The data range for each
axis is determined by the graph and is based on all the plots placed in it. Each graph can
have a second set of axes, called X2 (at the top) and Y2 (on the right). Each plot can
specify which axes it wants to use.

Legends: A graph can have one or more legends, and each can be placed in any part of
the graph. Each legend can specify the information to be displayed in it.

Single-cell graphs are created using the SGPLOT procedure.

2.1.2 Multi-Cell Classification Panels (SGPANEL Procedure)

Figure 2.2 Figure 2.3

Classification panels are very useful to visualize the distribution of data classified by one or
more class variables in one display. Figures 2.2 and 2.3 both display the association
between MPG and Horsepower for vehicles by country of origin and type of car. From both
graphs, one can easily glean some information:

o For both Asian and US cars, “Sedan” outnumbers “Sports” cars.

o Figure 2.2 shows that US sports cars provide higher horsepower choices.
o Figure 2.3 shows that Asian cars offer higher MPG choices.
 Chapter 2 SG Procedures 21

Figure 2.2 displays a classification panel with a Lattice layout, which supports two class
variables, one for row and one for column. Each row and column has a header that
displays the value of the classification variables.

Figure 2.3 displays a classification panel with a Panel layout, which supports multiple class
variables. Each cell in the panel has multiple headers, one for each class variable.

Classification panels are created using the SGPANEL procedure.

2.1.3 Multi-Cell Comparative Scatter Plots (SGCATTER Procedure)

While Classification Panels provide a convenient way to compare the same data across
classifiers, it is often desirable to make side-by-side comparisons of different measures.

Figure 2.4 Figure 2.5

Figure 2.4 displays a comparative graph for Mileage and Price by Weight and Horse
Power. This graph has common axes for comparison of the values. Figure 2.5 shows a
Scatter Plot Matrix for three variables, MPG, Horsepower, and MSRP. Such a matrix can
provide preliminary visual indication of direct or inverse associations between the variables.

Comparative and Matrix graphs are created using the SGSCATTER procedure.

2.1.4 Automatic Features

The SG Procedures examine the procedure syntax and apply built-in heuristics to enhance
the graph automatically. These actions include the following:
o Add a legend when appropriate for multiple overlays and classifiers.
o Create custom legend labels as needed for certain plot types.
o Assign different visual attributes to overlaid plot statements.
o Paginate large classification panels automatically.

If the results of the built-in heuristics are not desirable, they can be turned off.
22 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

2.2 SGPLOT Procedure

SGPLOT is a versatile procedure, providing the ability to create a wide variety of single-cell
graphs by combining the supported statements in creative ways. The SGPLOT procedure
supports over twenty different plot statements, along with statements for customization of
legends, axes, and insets.

The syntax for the SGPLOT procedure is as follows:

1. PROC SGPLOT < DATA= data-set > < options >;

2. plot-statement(s) required-parameters < / options >;

< refline-statement(s) >;

3. < inset-statement(s) >;

< axis-statement(s) >;

< keylegend-statement(s) >;

RUN;

1. The procedure statement supports multiple options. Use of these options will be
demonstrated in the examples shown in later chapters.

2. One or more plot statements are used to represent the data. Each plot statement has its
own set of required data roles and options. These are described in detail in later
chapters. Many plot statements are supported and can be grouped as shown below.
Plot statements from compatible groups can be combined in one procedure step:

a. Basic Plots: These include scatter, series, etc.

b. Fit and Confidence Plots: These include regression and Loess plots.
c. Distribution plots: These include histograms and box plots.
d. Categorization Plots: These include bar charts and dot plots.

3. Supporting statements can be used to customize the graph:

a. Reference Lines
b. Insets
c. Axes
d. Legends

Figure 2.6 shows a typical use case of the SGPLOT procedure. In this example, we have
created a distribution plot for the variable “Horsepower”. Three separate plot statements are
used, one for each data element of the graph. These statements are rendered in the order in
 Chapter 2 SG Procedures 23

which they are specified as illustrated in Figure 2.7. Figure 2.9 shows the actual rendered
graph.

proc sgplot data=sashelp.cars;

histogram horsepower;
density horsepower;
density horsepower / type=kernel;
keylegend / location=inside

position=topright
across=1;
run;

Figure 2.6 Figure 2.7

In Figure 2.6, we start with the HISTOGRAM statement, and then place two DENSITY plots
on it. The KEYLEGEND statement is used to customize the position of the legend.

o The HISTOGRAM statement is placed first in the area bounded by the axes.
o The first DENSITY curve (“Normal” by default) is placed on top of the histogram.
o The second DENSITY curve (Type=Kernel) is placed on top of the Normal curve.
o The legend is actually built by default by the procedure and placed on top. The
KEYLEGEND statement is used to customize its location.

Figure 2.8 and Figure 2.9 display the program and the resulting graph. The XAXIS statement
is added to suppress the x-axis label, as it is unnecessary given the title of the graph.

Proc sgplot data=sashelp.cars;

histogram horsepower;
density horsepower;
density horsepower / type=kernel;
keylegend / location=inside

position=topright
across=1;
xaxis display=(nolabel);
run;

Figure 2.8 Figure 2.9

24 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

As mentioned above, the supported plot statements are grouped in four categories. Plots
within each category can be combined in a procedure step. Plots from the “Basic Plots”
group and “Fit and Confidence Plots” group can also be combined.

INSET, REFLINE, and KEYLEGEND statements can be used as necessary to add these
elements to your graph.

Axes statements can be used to customize the appearance of any of the four axes:

o XAXIS is along the bottom of the graph.

o YAXIS is along the left side of the graph.
o X2AXIS is along the top of the graph.
o Y2AXIS is along the right side of the graph.

2.3 SGPANEL Procedure

The SGPANEL procedure extends the features of the SGPLOT procedure to a classification
panel, using one or more classification variables. The examples shown earlier in Figure 1.12
and Figure 1.13 are classification panels created using the SGPANEL procedure.

The syntax for the SGPANEL procedure is as follows:

1. PROC SGPANEL < DATA= data-set > < options >;

2. PANELBY classvar1 < classvar2 … <classvarN > < / options >;

3. < plot-statement(s) >;

< refline-statement(s) >;

4. < inset-statement(s) >;

< axis-statement(s) >;

< keylegend-statement(s) >;

RUN;

1. The procedure statement supports multiple options that will be demonstrated in the
examples shown in later chapters.

2. The PANELBY statement is required and must be placed before of any of the plot, refline,
inset, axis, or legend statements. This statement is used to set the layout type and other
options that control the overall paneling of the cells.
 Chapter 2 SG Procedures 25

3. One or more plot statements are used to represent data. Each plot statement has its
own set of data roles and options described in detail in later chapters. This procedure
supports most of the same plot statements as the SGPLOT procedure.

4. Supporting statements can be used to customize the graph:

a. Reference Lines
b. Axes
c. Legends

Figure 2.10 shows program statements for the creation of a classification panel displaying the
distribution of mileage by origin and type. The resulting graph is shown in Figure 2.11.

title 'Vehicle Mileage';

proc sgpanel data=sgbook.cars2;
where origin='USA' or
origin='Asia';
panelby origin type / novarname;
histogram mpgc;
density mpgc;
rowaxis display=none grid;
colaxis display=(nolabel);
run;

Figure 2.10 Figure 2.11

The program code in Figure 2.10 has the following noteworthy items:

o The PANELBY statement is required, and must come before any plot, axes,
reference line, or legend statements. This sets the classification variables.
o The two plot statements, HISTOGRAM and DENSITY, define the “Prototype” that is
used to populate each cell of the panel.
o The ROWAXIS and COLAXIS statements are used to customize the row and column
axes.

The PANELBY statement must provide one or more class variables. The procedure
automatically subdivides the graph region into multiple cells based on all crossings of the
unique values of the class variables. In Figure 2.10, we have specified two class variables
“Origin” and “Type”, each having two unique values. Hence, we get a graph with four cells,
as shown in Figure 2.11. Each cell has two headers.

Each cell in the graph is populated with the same set of plot statements as specified in the
procedure syntax. In this case, each cell gets a HISTOGRAM and a DENSITY plot. The
26 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

data for each cell is a subset based on the class value(s) for the cell. In Figure 2.11, the cell
titled “Asia” displays the subset of the data where Origin=’Asia’.

The non-plot statements such as the ROWAXIS and COLAXIS are used to customize the
common axes. A legend is automatically created based on an inspection of the syntax.

All cells of the graphs created by this procedure have common external axes:

o The x-axes for all columns are uniform.

o The y-axes for all rows are uniform.
o Uniform option may be set within rows or columns.

The SGPANEL procedure supports a number of layout types, which will be described in the
following sections.

2.3.1 Layout PANEL

This is the default layout type. This layout supports one or more (N) class variables. A cell is
created for each crossing of the unique values of all the class variables. Figure 2.12 shows a
graph with two class variables, each with 2 levels, resulting in a graph with 4 cells.

Figure 2.12
 Chapter 2 SG Procedures 27

A graph with a panel layout has the following features:

o Each cell has N cell headers, one for each class variable, displaying the value for
each cell.
o By default, only the cells that have data are displayed. Cells without any data are
dropped from the graph. An option can be used to display all cells.
o The procedure automatically decides the number of rows and columns for the grid.
When a graph has many cells, the procedure will automatically break up the graph
into multiple “pages”, to prevent the cells from getting too small.
o Common external row and column axes are used.
o Options are available to allow the user to control the “paging” of the graph.

2.3.2 Layout LATTICE

This layout requires two class variables. The first class variable is treated as the Column
variable, and the second as the Row variable. See Figure 2.13.

o Each unique value of the Column variable creates a column in the grid.
o Each unique value of the Row variable creates a row in the grid.
o Each column gets a column header, by default at the top.
o Each row gets a row header, by default on the right.
o Common external row and column axes are used.
o Every crossing of the unique values of the column and row variables is displayed as
a cell, regardless of whether it has data or not.

Figure 2.13
28 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

2.3.3 Layout COLUMNLATTICE

This is a special case of the LATTICE layout, where only one variable is provided in the list of
class variables. This layout produces a panel of columns in one row. Each cell has a column
header as shown in Figure 2.14.

Figure 2.14

2.3.4 Layout ROWLATTICE

This is a special case of the LATTICE layout, in which only one

variable is provided in the list of class variables. This layout
produces a panel of rows stacked in one column. Each cell has
a row header as shown is Figure 2.15.

2.3.5 Automatic Paging

Classification panels provide a convenient way to visualize data

by multiple class variables. This works very well when the
number of cells in a panel is small, say, less than ten. It is
possible to display a grid of 3 x 3 cells in a graph on a page.

However, as the number of cells grows larger, the individual cell

size shrinks. At some point, each cell becomes too small to
reasonably display the data contents. The SGPANEL
procedure automatically manages the number of cells in a
graph to keep reasonable cell sizes. If needed, the graph is
“paged” into multiple images, each showing a subset of the full
panel.

Options are available for user control of such paging, as will be

shown in some examples later in the book.
Figure 2.15
 Chapter 2 SG Procedures 29

2.4 Combining Statements

Plot statements from the Basic Plots and Fit and Categorization Plots groups can be mixed
and matched together freely. For other plot statements, only specific combinations are
allowed. The table in Figure 2.16 below shows the allowable combinations.

* Ellipse is not supported by PROC SGPANEL.

PBSPLINE
HBARPAR

HIGHLOW
VBARPAR

ELLIPSE *
SCATTER

HISTOGR
DENSITY
LINEPAR
REFLINE
VECTOR
NEEDLE

BUBBLE
SERIES

LOESS

HLINE
VLINE
HBOX
VBOX
BAND

HBAR
VBAR
STEP

REG

DOT
Basic Plots
SCATTER x x x x x x x x x x x x x x x x
SERIES x x x x x x x x x x x x x x x x
STEP x x x x x x x x x x x x x x x x
BAND x x x x x x x x x x x x x x x x
NEEDLE x x x x x x x x x x x x x x x x
VECTOR x x x x x x x x x x x x x x x x
VBARPARM x x x x x x x x x x x
HBARPARM x x x x x x x x x x x
BUBBLE x x x x x x x x x x x x x x x x
HIGHLOW x x x x x x x x x x x x x x x x
REFLINE x x x x x x x x x x x x x x x x x x x x x x x x x
LINEPARM x x x x x x x x x x x x x x x x
LINEPARM x x x x x x x x x x x x x x x x
Fit and Confidence Plots
REG x x x x x x x x x x x x x x
LOESS x x x x x x x x x x x x x x
PBSPLINE x x x x x x x x x x x x x x
ELLIPSE * x x x x x x x x x x x x x x
Distribution Plots
HISTOGRAM x x x
DENSITY x x x
VBOX x x
HBOX x x
Categorization Plots
VBAR x x x
VLINE x x x
HBAR x x x x
HLINE x x x x
DOT x x x x
Figure 2.16
30 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

2.5 SGSCATTER Procedure

The SGSCATTER procedure is optimized for the creation of comparative scatter plots and
scatter plot matrices. PROC SGSCATTER does not use a layered architecture like the
SGPLOT and SGPANEL procedures. Instead, PROC SGSCATTER supports three distinct
plot statements as shown below. The examples shown earlier in Figure 2.4 and Figure 2.5
are created using the SGSCATTER procedure.

The syntax for the SGSCATTER procedure is as follows:

1 PROC SGSCATTER < DATA= data-set > < options >;

2. PLOT plot-requests < / options >;

3. COMPARE X=(variable(s)) Y=(variable(s)) < / options >;

4. MATRIX variable(s) < / options >;

5. RUN;

1. The procedure statement supports multiple options that will be demonstrated in the
examples shown in later chapters.

2. The PLOT statement creates a set of cells arranged in a uniform grid in the graph based
on the plot request. Each cell includes an independent scatter plot with optional fit and
ellipses. Axes can be made uniform across all plots, if needed. Figures 2.17 and 2.18
show an example of the PLOT statement. The two Y axes are not uniform.

title 'Mileage by Horsepower';

proc sgscatter data=sgbook.cars2;
plot (mpgc mpgh) * hp /

markerattrs=(symbol=circlefilled)
transparency=0.85;
run;

Figure 2.17 Figure 2.18

 Chapter 2 SG Procedures 31

3. The COMPARE statement creates a row and column grid of cells in the graph based on
the list of X and Y variables. Each cell includes a scatter plot with optional fit and
ellipses. Each row of the grid has a common external y-axis. Each column of the grid
has a common external x-axis. Options can be used to customize the plots. Figures
2.19 and 2.20 show an example of the COMPARE statement.

title 'Variable Comparisons';

proc sgscatter
data=sgbook.cars2;
compare x=(mpgc mpgh) y=(hp)/
markerattrs=(symbol=
circlefilled)
transparency=0.85 grid;
run;

Figure 2.19 Figure 2.20

4. The MATRIX statement creates a row and column grid of cells based on the list of
variables specified. Each cell includes a scatter plot with optional fit and ellipses. Each
row and column of the grid has external axes. Figures 2.21 and 2.22 show an example
of the MATRIX statement.

title 'Variable
Associations';
proc sgscatter
data=sgbook.cars2;
matrix mpgc weight hp msrp
/
markerattrs=(symbol=
circlefilled)
transparency=0.9;
run;

Figure 2.21 Figure 2.22

32 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

2.6 Styles and Their Usage

Graphs created by the SG procedures use pre-defined visual attributes for various elements
that make up a graph. This information is defined in ODS Styles that have been carefully
designed with aesthetics and effectiveness in mind. This ensures these graphs have an
aesthetically pleasing appearance by default. This also ensures that various elements of the
graph are balanced and distinct from each other. This is achieved through the usage of the
ODS Styles.

A style is a collection of named elements. Each element is a bundle of named visual

attributes such as color, font, marker symbol, and so on. Any output to an ODS destination
derives the necessary visual information from the active style. For example, output tables
derive visual attributes such as fonts and background colors for the headers, size, and color
of titles, etc. from the style.

Graphs also derive the visual attributes for plot colors, marker symbols, line thickness, axis
label fonts, etc. from specific named elements of the style. The association between the
element of the graph and the style element is well defined and described in detail in the ODS
product documentation. Some relevant information is included in Chapter 14.

You can control the visual appearance of the graphs in different ways:

1. Use a pre-defined SAS style.

2. Use a custom style.
3. Use appearance options.

1. Use a pre-defined SAS style: Every ODS output destination has a default style. All
graphs written to this destination use this style by default. You can change the active style
for an ODS output destination by setting the STYLE= option for the destination. All graphs
written to that destination will then use that style.

ods destination style=style-name;

2. Use a custom style: If you like one of the pre-defined SAS styles, but would prefer to
change a few of the appearance settings, you can derive a new style from one of the SAS
styles by using the TEMPLATE procedure. For more information on this topic, see the ODS
product documentation for PROC TEMPLATE.

3. Use appearance options: A style set on the ODS destination affects all features of the
all graph rendered to the destination. You can customize the appearance of specific features
of a graph by setting specific appearance options in the procedure syntax. This overrides the
settings derived from the style only for that one use case.

Usage of these options will become apparent through many examples in this book. More
information on this topic is covered in Chapter 14.
 Chapter 2 SG Procedures 33

2.7 Template-Based Graphics vs. Device-Based Graphics

The SG procedures work within the ODS Graphics system and use GTL to do the work of
rendering the graph output. For this reason, we will refer to these procedures as “template-
®
based graphics”. The other SAS/GRAPH procedures, such as GPLOT and GCHART, will be
referred to as “device-based graphics”.
®
Often, users ask what is different about these procedures from the other SAS/GRAPH
procedures. While we cannot cover every difference here, the key differences in philosophy,
process, and technology are listed below:

1. SG procedures create output directly in industry standard formats, such as PNG, JPG,
PDF, PS, and so on. SG procedures do not create GSEG output.

2. SG procedures do not honor traditional GRAPH global statements such as GOPTIONS,

PATTERN, SYMBOL, AXIS or LEGEND. However, equivalent or similar functionality is
provided through other procedures, ODS Graphics, or ODS destination options.

a. The GOPTIONS statement is not supported (except for using the RESET option to
reset titles and footnotes). Global options that control the output are specified on the
ODS GRAPHICS statement.

b. TITLE and FOOTNOTE statements are supported, with the exception of a few
options.

c. PATTERN and SYMBOL statements are not supported. However, plot types and
attributes are specified directly on the plot statements, which make the application of
those attributes clearer.

d. AXIS statements are not supported. However, PROC SGPLOT supports four axis
statements (XAXIS, YAXIS, X2AXIS, and Y2AXIS) to control the axes around the
cell. PROC SGPANEL supports the ROWAXIS and COLAXIS statements to control
its axes. PROC SGSCATTER does not have axis statements, but there are
techniques that will be described later for controlling some aspects of the axes.

e. LEGEND statements are not supported. PROC SGPLOT and SGPANEL support a
KEYLEGEND statement to control the contents of the legend, its layout, and position.
PROC SGSCATTER has a LEGEND option to give some legend control.
34
Chapter 3
Common Graphs
3.1 Introduction 37
3.2 Single-Cell Graphs 38
3.3 Classification Panels 55
3.4 Comparative and Matrix Graphs 57
36 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“There are 10 types of people in this world:

those who understand binary and those who don't.”
~ Unknown
 Chapter 3 Common Graphs 37

Chapter 3: Common Graphs

In this chapter, we review some graphs commonly used in many domains, from the basic scatter
plot to complex comparative graphs and classification panels. We will defer the detailed
discussion of various plot options to subsequent chapters.

3.1 Introduction
All the graphs have a common structure:

o one or more titles at the top and one or more footnotes at the bottom
o various graphical representations of the data in the middle
o one or more legends, inside or outside the data area
o insets or statistics table inside or outside the data area

Figure 3.1.1

Single-Cell Graphs: In section 3.2, we will use the SGPLOT procedure to create single-cell
graphs. We will put together the needed syntax to create graphs like the one shown above in
Figure 3.1.1.

Classification Panels: In section 3.3, we will use the SGPANEL procedure to create
classification panels. Once you know how to mix and match statements for PROC SGPLOT, all
you need is a PANELBY statement to set up the classification variables.

Comparative and Matrix Graphs: In section 3.4, we will use the SGSCATTER procedure to
create comparative and matrix graphs.

In the rest of this chapter, we provide a broad coverage of the different types of graphs you can
create using the SG Procedures. The goal is to expose to you the different types of graphs,
without going deep into the features of each plot. We will cover that in subsequent chapters.
38 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

3.2 Single-Cell Graphs

Figure 3.2.1: Basic Scatter Plot

This is a basic scatter plot of mileage by

horsepower for sedans and sports cars.

Default option settings are used, and the

appearance settings are derived from the
active style.

Axes are thinned, and “optimal” tick values

are displayed.

title 'MPG by Horsepower for Cars';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city;
run;

Figure 3-4: Grouped

Figure 3.2.2: ScatterPlot
Grouped Scatter Plotwith
with
Grid
a Lines Legend
Custom
This legend
The is a grouped
can be scatter plot of mileage
customized by using by
horsepower by car type. The visual
the KEYLEGEND statement Here we
attributes for each marker are based on the
have
uniqueadded a legend
value of title
the group and placed the
variable.
legend inside the graph area, in the
upper right corner.
The legend The in
is positioned contents of the
the upper right
legend are still
corner of the data area using the
KEYLEGEND statement.

title 'MPG by Horsepower for Cars by Type';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city/ group=type;
keylegend / title='Type: ' location=inside position=topright;
xaxis grid;
yaxis grid;
run;
 Chapter 3 Common Graphs 39

Figure 3.2.3: Fit Plot

This is a fit plot of mileage by horsepower.

The REG statement displays both the original
data and the fit line. The large number of
observations can obscure the fit line and
clutter up the graph.

The thickness of the fit line is set using the

LINEATTRS option.

title 'MPG by Horsepower with Regression Fit';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
reg x=horsepower y=mpg_city / lineattrs=(thickness=2);
keylegend / location=inside position=topright;
xaxis grid;
yaxis grid;
run;

Figure 3.2.4: Fit Plot with Transparent

Markers

In this fit plot, the clutter is reduced by

suppressing the markers on the REG
statement and using a SCATTER statement
to display the markers with transparency.

Using transparency can provide a better feel

for the distribution of the data.

title 'Mileage by Horsepower with Marker Transparency';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city / transparency=0.9
markerattrs=(symbol=circlefilled);
reg x=horsepower y=mpg_city / nomarkers lineattrs=(thickness=3);
keylegend / location=inside position=topright across=1;
xaxis grid;
yaxis grid;
run;
40 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 3.2.5: Fit Plot with Confidence

Limits
nd
This fit plot uses a regression plot with 2
order fit using the DEGREE option on the
REG statement. Confidence limits are
requested using the CLI and CLM.

The default legend is positioned in the top

right corner of the graph.

title 'Mileage by Horsepower';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city / transparency=0.9
markerattrs=(symbol=circlefilled);
reg x=horsepower y=mpg_city / nomarkers lineattrs=(thickness=2)
cli clm degree=2;
keylegend / location=inside position=topright across=1;
xaxis grid; yaxis grid; run;

Figure 3.2.6: Grouped Fit Plot with

Limits

This shows a grouped regression plot with

nd
confidence limits and 2 order fit.
CLMTRANSPARENCY is set on the REG
statement.

To customize the legend, we have provided

a name for the REG plot, and included it in
the KEYLEGEND statement.

title 'Mileage by Horsepower and Type';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city / group=type transparency=0.9
markerattrs=(symbol=circlefilled);
reg x=horsepower y=mpg_city / group=type nomarkers name='a'
lineattrs=(thickness=2) degree=2 clm clmtransparency=0.6;
keylegend 'a' / location=inside position=topright across=1;
xaxis grid; yaxis grid; run;
 Chapter 3 Common Graphs 41

Figure 3.2.7: Histogram

This graph displays a basic distribution of

mileage for all cars using a histogram.

The y-axis grid lines are enabled.

The x-axis label is suppressed.

By default the histogram shows a LINEAR

axis, as shown here. To show a binned
axis, use the SHOWBINS option.

title 'Distribution of Mileage';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
histogram mpg_city;
xaxis display=(nolabel);
yaxis grid;
run;

Figure 3.2.8: Histogram with Density

Plots

This graph shows the distribution of

horsepower for all cars, along with the
Normal and Kernel density functions.

The legend is placed in the upper right

corner of the graph.

title 'Distribution of Horsepower';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
histogram horsepower;
density horsepower;
density horsepower / type=kernel;
keylegend / location=inside position=topright across=1;
xaxis display=(nolabel);
yaxis grid;
run;
42 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 3.2.9: Box Plot – Vertical

This graph displays the distribution of

mileage for all cars by type using a box
plot.

The x-axis label is suppressed since this

information is included in the title already.

title 'Mileage by Type';

proc sgplot data=sashelp.cars;
vbox mpg_city / category=type;
xaxis display=(nolabel);
yaxis grid;
run;

Figure 3.2.10: Grouped Box Plot

This graph displays the distribution of

mileage for vehicles grouped by origin
using a horizontal box plot.

The line attributes are set to make all box

outlines solid, and the legend is moved
into the bottom right corner of the wall.

The x-axis label is suppressed.

title 'Mileage by Type';

proc sgplot data=sashelp.cars; 9.3
where (type='Sedan' or type='Sports' or type='Truck') and
(origin='USA' or origin='Asia');
hbox mpg_city / category=type group=origin groupdisplay=cluster
lineattrs=(pattern=solid);
keylegend / location=inside position=bottomright;
xaxis grid display=(nolabel);
run;
 Chapter 3 Common Graphs 43

Figure 3.2.11: Interval Box Plot

This graph displays the distribution of

horsepower by mileage on an interval axis.

The width of the boxes is based on the

minimum interval between x-axis values.

Note: Box plot for interval data is a SAS 9.3

feature. The TYPE=interval option itself for
XAXIS is available with SAS 9.2.

title 'Horsepower by Mileage';

proc sgplot data=sashelp.cars; 9.3
where origin='USA';
vbox Horsepower / category=mpg_city;
xaxis type=linear;
yaxis grid;
run;

Figure 3.2.12: Series Plot

This graph shows a traditional stock plot of

closing price by month using a series plot.

The x-axis label is suppressed since it is

redundant.

A WHERE clause is used to display the

data for only one value of the stock variable.

title 'Stock Plot for IBM';

proc sgplot data=sashelp.stocks;
where stock = 'IBM';
series x=date y=close;
xaxis grid display=(nolabel);
yaxis grid;
run;
44 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 3.2.13: Grouped Series Plot

This graph shows a plot of closing stock

price by month for three companies.

Patterned lines are hard to follow when the

data fluctuates rapidly. This is a case
where a color style would be more
effective.

ANTIALIASMAX is set to 1000 to prevent

disabling of antialiasing for a large number
of observations.

ods graphics / antialiasmax=1000;

title 'Stock Plot by Company';
proc sgplot data=sashelp.stocks;
series x=date y=close / group=stock;
keylegend / location=outside across=3;
xaxis grid display=(nolabel);
yaxis grid;
run;

Figure 3.2.14: Multi-Response Series

This graph shows the efficacy of

medication over time for three different
treatment groups using multiple series
plots. Each plot is labeled directly, making
a legend unnecessary.

The procedure automatically assigns

different visual attributes to each plot. By
default, the time axis displays the year
value only when needed.

title 'Efficacy Over Time by Treatment Group';

proc sgplot data=sgbook.series noautolegend;
series x=date y=a / lineattrs=(thickness=2) curvelabel='Drug A';
series x=date y=b / lineattrs=(thickness=2) curvelabel='Drug B';
series x=date y=c / lineattrs=(thickness=2) curvelabel='Drug C';
xaxis display=(nolabel);
yaxis min=0 max=15 integer grid label='Efficacy';
run;
 Chapter 3 Common Graphs 45

Figure 3.2.15: Grouped Step Plot

This graph shows a traditional stock plot of

closing price by month for three companies
using a grouped step plot.

This is similar to Figure 3.2.13, with use of a

step plot instead of a series plot.

title 'Stock Plot by Company';

proc sgplot data=sashelp.stocks;
where date > '01Jan99'd and date < '01Jan00'd;
format close dollar5.0;
step x=date y=close / group=stock lineattrs=(thickness=2);
xaxis grid display=(nolabel);
yaxis grid;
keylegend / location=inside position=topright across=1;
run;

Figure 3.2.16: Step Plot with Limits

This plot shows the predicted weight by

height for students in a classroom using a
step plot.

Confidence values are shown as Y error

bars. LINEATTRS option is used to set the
thickness of the step plot line, and
ERRORBARATTRS is used to set the
pattern for the error bars.

title 'Predicted Weight by Height';

proc sgplot data=sashelp.classfit noautolegend;
step x=height y=predict / yerrorlower=lower yerrorupper=upper
lineattrs=(thickness=2) errorbarattrs=(pattern=dash);
xaxis grid;
yaxis grid label='Weight';
run;
46 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 3.2.17: Dot Plot with Limits

This graph displays the mileage for cars by

type using a dot plot. The mean value is
represented by the dot. The 95%
confidence limits are displayed.

The legend is placed inside the data area at

the bottom. To create some space for this,
we have to set the OFFSETMAX=0.2. This
is because the DOT plot sets REVERSE on
the y-axis.

title 'Mileage by Type (Excluding Hybrids)';

proc sgplot data=sashelp.cars;
where type ne 'Hybrid';
dot type / response=mpg_city stat=mean limits=both;
keylegend / location=inside position=bottom;
xaxis grid;
yaxis offsetmax=0.2 display=(nolabel);
run;
Figure 3.2.18: Vector Plot

This graph shows the mileage by

horsepower for sedans made in USA as
related to customer preferences of 20 MPG
and 200 HP using a vector plot.

X and Y origins for the vector plot are set to

constant values, so that all the vectors have
the same starting point.

title 'Sedans Made in USA';

proc sgplot data=sashelp.cars;
where type= 'Sedan' and origin='USA';
vector x=horsepower y=mpg_city / xorigin=200 yorigin=20;
xaxis grid; yaxis grid;
run;
 Chapter 3 Common Graphs 47

Figure 3.2.19: Grouped Vector Plot

This graph shows the particle location and

velocity using a vector plot. Direction
(forward or backwards) is indicated by the
group variable

Here, the X and Y origin roles are assigned

columns from the dataset, so each vector
has different start and end point.

title 'Particle Position & Velocity';

proc sgplot data=sgbook.vectorGroup;
vector x=x y=y / xorigin=xo yorigin=yo group=group lineattrs=(thickness=2);
xaxis grid display=(nolabel);
yaxis grid display=(nolabel);
keylegend / location=inside position=bottomleft across=1;
run;

Figure 3.2.20: Bar Chart

This is a basic bar chart of the mean

mileage by country of origin.

Data labels are displayed using the

DATALABEL option. A format is applied to
the variable to reduce the label precision.

The x-axis label is suppressed as it is

redundant.

title 'Mean Mileage by Origin';

proc sgplot data=sashelp.cars;
format mpg_city 5.2;
vbar origin / response=mpg_city stat=mean datalabel;
xaxis display=(nolabel);
yaxis grid;
run;
48 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 3.2.21: Grouped Bar Chart with

Skins

This is a grouped bar chart of mileage by

vehicle type with country of origin as the
group variable.

GROUPDISPLAY=cluster is set to display

adjacent bars for each group.

title 'Mileage by Country';

proc sgplot data=sashelp.cars; 9.3
where type='Sedan' or type='Sports';
format mpg_city 5.1;
vbar type / response=mpg_city group=origin stat=mean
groupdisplay=cluster datalabel;
xaxis display=(nolabel);
yaxis grid;
run;

Figure 3.2.22: Stacked Bar Chart with

Skins

This is a grouped bar chart of sales by

product with region as the group variable.

The group values are stacked by default.

The x-axis label is suppressed.

title 'Revenues by Product and Region';

proc sgplot data=sashelp.prdsale;
format actual dollar8.0;
vbar product / response=actual group=region;
xaxis display=(nolabel);
yaxis grid;
run;
 Chapter 3 Common Graphs 49

Figure 3.2.23: Bar-Line Overlay

This is a basic bar-line overlay of two

different measures by quarter.

The x-axis label is suppressed. The line

thickness is set using the LINEATTRS
option. Transparency is set to 30%.

title 'Company Finances';

proc sgplot data=sgbook.quarter;
vbar quarter / response=revenues nostatlabel;
vline quarter / response=expenses nostatlabel
lineattrs=(thickness=10) transparency=0.3;
yaxis grid offsetmin=0; xaxis display=(nolabel);
run;

Figure 3.2.24: Bar-Bar Overlay

This is a bar-bar overlay of two different

measures by country. Bar width for the
second bar chart is reduced.

title 'City and Highway Mileage by Country';

proc sgplot data=sashelp.cars;
vbar origin / response=mpg_highway stat=mean nostatlabel;
vbar origin / response=mpg_city stat=mean
nostatlabel barwidth=0.6;
xaxis display=(nolabel);
yaxis grid offsetmin=0 label='Mileage';
run;
50 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 3.2.25: Bar Chart with Multiple

Responses

This graph shows two separate response

variables plotted by country of origin. Here
the two responses are displayed side by
side using DISCRETEOFFSET and
BARWIDTH options. The yaxis label is
used.

This graph uses DATASKIN=Gloss to apply

a decorative finish to the bars.

title 'City and Highway Mileage by Country';

proc sgplot data=sashelp.cars; 9.3
where type='Sedan' or type='Sports';
vbar origin / response=mpg_city stat=mean dataskin=gloss
discreteoffset=-0.22 barwidth=0.4;
vbar origin / response=mpg_highway stat=mean dataskin=gloss
discreteoffset=+0.22 barwidth=0.4;
xaxis display=(nolabel);
yaxis label="Mileage"; run;

Figure 3.2.26: Bar Chart with Patterns

This graph shows a vertical bar chart of

mileage by country of origin with vehicle
type as group variable. The group values
are stacked by default.

This graph uses the Journal2 style. The

bars do not have fill colors, only fill patterns
(9.3).

ods listing style=journal2;

title 'MPG by Country and Type'; 9.3
proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
vbar origin / response=mpg_city stat=mean group=type;
xaxis display=(nolabel);
run;
 Chapter 3 Common Graphs 51

Figure 3.2.27: Grouped Bar Chart with

Skins

This graph shows mileage by origin using a

vertical bar chart with GROUP=type.
GROUPDISPLAY=cluster is used to
display adjacent bars for each group.

This graph uses DATASKIN=Sheen.

title 'City and Highway Mileage by Country';

proc sgplot data=sashelp.cars; 9.3
where type='Sedan' or type='Sports';
vbar origin / response=mpg_city stat=mean group=type
groupdisplay=cluster dataskin=sheen transparency=0;
xaxis display=(nolabel);
run;

Figure 3.2.28: Multiple Bar Chart with

Patterns, Fill Colors, and Skins

This graph shows city and highway mileage

by origin using adjacent bars with fill
patterns, fill colors, and skins.

The Journal3 style uses color shades and

fill patterns. The DISCRETEOFFSET and
BARWIDTH options are used to position
the two responses side by side.

ods listing style=journal3;

title 'City and Highway Mileage by Country'; 9.3
proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
vbar origin / response=mpg_city stat=mean barwidth=0.4
dataskin=pressed discreteoffset=-0.22;
vbar origin / response=mpg_highway stat=mean barwidth=0.4
dataskin=pressed discreteoffset=+0.22;
xaxis display=(nolabel);
yaxis label='Mileage'; run;
52 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 3.2.29: Horizontal Bar Chart

This is a basic horizontal bar chart of

mileage by car type, with data labels.

A horizontal bar chart is more suitable for

larger number of bars and long tick values.

The x-axis and y-axis labels are suppressed

as they are redundant.

title 'Mileage by Type';

proc sgplot data=sashelp.cars;
format mpg_city 5.1;
hbar type / response=mpg_city stat=mean datalabel;
xaxis grid display=(nolabel);
yaxis display=(nolabel);
run;

Figure 3.2.30: Horizontal Bars with Limits

This is a basic horizontal bar chart of

mileage by car type, with limit bars by using
the STAT and LIMIT options.

The y-axis label is disabled by using the

DISPLAY option.

title 'Mileage by Type';

proc sgplot data=sashelp.cars;
hbar type / response=mpg_city stat=mean limits=both;
xaxis grid display=(nolabel);
yaxis display=(nolabel);
run;
 Chapter 3 Common Graphs 53

Figure 3.2.31: HighLow Plot

This is a traditional stock plot of price by

date, showing the high, low, open, and
close prices using a HighLow plot with the
OPEN and CLOSE options.

The default TYPE for the HighLow plot is

‘Line’.

title 'Stock Plot';

proc sgplot data=sashelp.stocks; 9.3
where stock='IBM' and date > '01Jan2004'd;
highlow x=date high=high low=low / open=open close=close
lineattrs=(thickness=2);
xaxis grid display=(nolabel);
yaxis grid display=(nolabel);
run;

Figure 3.2.32: HighLow Plot

This is a plot showing duration of

medications used for a clinical trial. The
duration for each drug is plotted.

In this example, the TYPE for the HighLow

plot is set to ‘Bar’. The HIGHCAP and
LOWCAP options are set to the ‘cap’ data
variable to display bar end caps.

title 'Medications Plot';

proc sgplot data=sgbook.highlow noautolegend; 9.3
highlow y=drug low=low high=high / group=drug
lineattrs=(pattern=solid) highcap=cap lowcap=cap
transparency=0.2 type=bar barwidth=0.4;
yaxis display=(nolabel) grid;
xaxis display=(nolabel) min=0 max=100 grid;
run;
54 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 3.2.33: Bubble Plot

This is a grouped bubble plot, showing the

response by size. Reference lines are
used to demarcate the quadrants.

The max and min response values are

mapped to max and min bubble sizes by
radius.

title 'Response Plot';

proc sgplot data=sgbook.bubble; 9.3
refline 5 / axis=x transparency=0.7;
refline 2 / axis=y transparency=0.7;
bubble x=x y=y size=value / group=group transparency=0.4;
keylegend / title='Drug' location=inside position=topright;
yaxis display=(nolabel) min=0 max=4;
xaxis display=(nolabel) min=0 max=10;
run;

Figure 3.2.34: Bubble Plots

This is a grouped bubble plot, showing the

response by size. Here we are using
unfilled bubbles. The groups are displayed
using line patterns for the outline.

This uses the Journal style and NOFILL

option. BRADIUSMAX is used to set the
maximum size of the bubble in the graph to
50 pixels.

title 'Response Plot';

proc sgplot data=sgbook.bubble; 9.3
bubble x=x y=y size=value / group=group bradiusmax=50 nofill;
keylegend / title='Drug' location=inside position=bottomleft;
yaxis display=(nolabel) min=0 max=4;
xaxis display=(nolabel) min=0 max=10;
run;
 Chapter 3 Common Graphs 55

3.3 Classification Panels

Figure 3.3.1: Bar Chart Panel

This is a classification panel using origin as

the class variable.

Each cell contains a bar chart of mileage by

type. Skins and transparencies can be
combined to achieve different effects.

title 'Mileage by Type and Country';

proc sgpanel data=sashelp.cars ; 9.3
where (type='Sedan' or type='Sports') and
(origin='USA' or origin='Asia');
panelby origin / novarname;
vbar type / response=mpg_city stat=mean dataskin=gloss
transparency=0.6;
rowaxis grid; run;

Figure 3.3.2: Box Plot Row Lattice

This is a classification panel using type as

the row class variable. Each cell contains a
box plot of mileage by origin.

This uses LAYOUT=ROWLATTICE to

create multiple rows in a single column.

The COLAXIS statement is used to display

the grid lines for the column axis.

title 'Mileage by Type and Country';

9.3
proc sgpanel data=sashelp.cars noautolegend;
where (type='Sedan' or type='Sports') and
(origin='USA' or origin='Asia');
panelby type / layout=rowlattice novarname;
hbox mpg_city / category=origin group=origin;
colaxis grid;
run;
56 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 3.3.3: Histogram Lattice

This is a 2x2 lattice displaying the distribution

of cholesterol by sex and blood pressure
status using the HISTOGRAM and DENSITY
statements.

The ROWAXIS and COLAXIS statements

are used to display the grid lines.

title 'Cholesterol Panel';

proc sgpanel data=sashelp.heart noautolegend;
where bp_status='Normal' or bp_status='High';
panelby sex bp_status / layout=lattice novarname;
histogram cholesterol / transparency=0.5;
density cholesterol;
rowaxis grid; colaxis grid;
run;

Figure 3.3.4: Scatter Panel

This graph has three class variables: type,

origin, and drivetrain. Each cell has three
headers, one for each class variable.

The header labels can take up a large

portion of the cell space. Here, we have
reduced the header label font size (in the
style) to allow more space for the plot.

title 'Mileage by Horsepower';

proc sgpanel data=sashelp.cars ;
where (type='Sedan' or type='Sports') and
(origin='USA' or origin='Asia') and
(drivetrain='Front' or drivetrain='Rear');
panelby origin type drivetrain / novarname onepanel columns=4;
scatter x=horsepower y=mpg_city / transparency=0.6;
rowaxis grid; colaxis grid;
run;
 Chapter 3 Common Graphs 57

3.4 Comparative and Matrix Graphs

Figure 3.4.1: Comparative Scatter Plot

This is a grid of scatter plots comparing

systolic and diastolic blood pressure by
cholesterol and weight.

The graph uses common external axes, so

the data is uniform across rows and
columns.

title 'Comparative Graph';

proc sgscatter data=sashelp.heart;
where ageatstart > 58;
compare x=(cholesterol weight) y=(diastolic systolic) /
transparency=0.8 markerattrs=(symbol=circlefilled) grid;
run;

Figure 3.4.2: Scatter Plot Matrix

This is a scatter plot matrix, showing the

association between variables for patients
over the age of 58.

Transparency can be used to see the

distribution of the data. The diagonals
display the distribution of each variable by
itself.

title 'Variable Association';

proc sgscatter data=sashelp.heart;
label weight='Weight' cholesterol='Cholstrl'
diastolic='Diastlc' systolic='Systlc';
where ageatstart > 58;
matrix cholesterol diastolic systolic weight / diagonal=(histogram)
transparency=0.8 markerattrs=(symbol=circlefilled);
run;
58
Chapter 4
Basic Plots
4.1 Introduction 61
4.2 SGPLOT Procedure 62
4.3 Plot Roles and Options 63
4.4 Scatter Plot 64
4.5 Scatter Plots with Data Labels 74
4.6 Series Plot 78
4.7 Step Plot 84
4.8 Band Plot 89
4.9 Needle Plot 94
4.10 Vector Plot 97
4.11 VBarParm and HBarParm Plots (9.3) 100
4.12 Bubble Plot (9.3) 105
4.13 HighLow Plot (9.3) 108
4.14 Reference Lines 112
4.15 Parametric Line Plot (9.3) 115
4.16 Waterfall Chart (9.3) 117
4.17 Combining the Plots 120
60 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“If you can't explain it simply,

you don't understand it well enough.”
~ Albert Einstein
 Chapter 4 Basic Plots 61

Chapter 4: Basic Plots

In this chapter, we examine the basic plots for SGPLOT and SGPANEL procedures. These plots
have an extensive set of options that will be covered in detail in this chapter. Plots within this
group can be mixed and matched with each other in a myriad of ways to build complex graphs as
shown in Figure 3.2.5.

4.1 Introduction
Basic plots used in both the SGPLOT and SGPANEL procedure are the workhorse plots for
visualization of raw data, or of summary statistics that have been computed prior to creation of
the graph. These plots do not process the data in any way, but they plot the raw data values in
the graph. The plots included in this group are:

• SCATTER • SERIES • STEP • BAND

• NEEDLE • VECTOR • VBARPARM • HBARPARM
• BUBBLE • HIGHLOW • REFLINE • LINEPARM
• WATERFALL

Figure 4.1
62 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

4.2 SGPLOT Procedure

The SGPLOT procedure supports the following syntax:

proc sgplot data=datasetname <options>;

Options:
CYCLEATTRS boolean Cycle through style elements for each plot
DATA =sas-data-set Optional data set
DATTRMAP =sas-data-set Data set defining an attribute map
DESCRIPTION =string Description string
NOAUTOLEGEND boolean Do not show the automatic legend
NOCYCLEATTRS boolean Do not cycle style elements for each plot
PAD =value Padding around the outside of the graph
SGANNO =sas-data-set Data set containing the annotations
TMPLOUT =string File name for generated graph template code
UNIFORM =keyword Uniform axis and legends across BY variables

The procedure supports the following statements:

• Basic Plots: BAND, NEEDLE, SCATTER, STEP, SERIES, VECTOR,
BUBBLE, HIGHLOW, HBARPARM, VBARPARM
• Fit Plots: LOESS, PBSPLINE, REG, LINEPARM
• Distribution Plots: DENSITY, ELLIPSE HISTOGRAM, HBOX, VBOX
• Categorical Plots: DOT, HBAR, HLINE, VBAR, VLINE, WATERFALL
• Other: KEYLEGEND, INSET, REFLINE, XAXIS, YAXIS, X2AXIS, Y2AXIS

The procedure has the following syntax:

PROC SGPLOT < DATA= data-set > < options >;

plot-statement(s) required-parameters < / options >;
< refline-statement(s) >;
< inset-statement(s) >;
< keylegend-statement(s) >;
< axis-statement(s) >;
RUN;

As seen in the examples in Chapter 3, one or more plot statements can be provided that will work
together to create the graph. See the table of permissible combinations shown in section 2.4.
 Chapter 4 Basic Plots 63

4.3 Plot Roles and Options

All of the plot statements have the following syntax:

PlotName required-data-roles </ options>;

All plot statements have required and optional parameters necessary to create the plot. These
parameters fall in the following broad categories:

1. required data roles

2. optional data roles
3. options
4. common options

Categories 1 - 3 above include role and option names that are specific to each plot. These role
and option names are used consistently and only as needed. We will discuss these with each
specific plot statement.

Category 4 includes the set of options that are common to all plot statements as shown in the
table below. These options work in the same way across all plot statements. Some plots may
not support one or more of these common options.

Common Options:
ATTRID =string The attribute map ID (See Chapter 9).
CLUSTERWIDTH =value Fraction of the mid-point spacing to be used for
drawing the group cluster.
DISCRETEOFFSET =value Fractional shift within midpoint spacing.
GROUPDISPLAY =keyword OVELAY | CLUSTER. Default value varies.
GROUPORDER =keyword DATA | ASCENDING | DESCENDING
LEGENDLABEL =string The label that appears in the legend to represent
this (non-group) plot.
NAME =string Specifies a name for this statement. Other
statements, such as KEYLEGEND, can refer to a
plot by its name.
TRANSPARENCY =value Specifies transparency for the visual elements.
URL =string-column URL link when used in HTML output.

Common Boolean Options:

NOMISSINGGROUP Missing values are not displayed as a separate group.
X2AXIS Assigns X values to the X2 (top) axis.
Y2AXIS Assigns Y values to the Y2 (right) axis.

These common options are available for each plot statement, and can be used exactly as
described above. To avoid duplication, we will not list the above common options for the
individual plots that are discussed in this chapter.
64 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

4.4 Scatter Plot

The basic scatter plot is a key plot type for visualization of the raw
data. The syntax is:

scatter x=column y=column </ options>;

A marker is displayed at each (x, y) location. Group, Data Labels,

Marker Character, and X and Y Error Bars are supported. Discrete
axes can have cluster grouping and discrete offsets.

Required Data Roles:

X =column Position along the horizontal axis
Y =column Position along the vertical axis

Optional Data Roles:

DATALABEL =column Label String for each observation
FREQ =num-column Frequency count for each observation
GROUP =column Classification variable for each observation
MARKERCHAR =column String to be shown for the observation
URL =column URL for each observation
XERRORLOWER =num-column Lower value for X Error Bar
XERRORUPPER =num-column Upper value for X Error Bar
YERRORLOWER =num-column Lower value for Y Error Bar
YERRORUPPER =num-column Upper value for Y Error Bar

Appearance Options:
ERRORBARATTRS =line-attrs Appearance attributes for error bars
MARKERATTRS =marker-attrs Appearance attributes for markers
MARKERCHARATTRS =text-attrs Appearance attributes for char markers

Boolean Options:
DATALABEL Display Y value as data label

Standard common options are supported

ATTRID, CLUSTERWIDTH, DISCRETEOFFSET, GROUPDISPLAY, GROUPORDER,
LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, URL, X2AXIS,
Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
 Chapter 4 Basic Plots 65

Figure 4.4.1: Basic Scatter Plot

This is a basic scatter plot of mileage by

horsepower. Marker attributes are
nd
customized to use the 2 data element with
filled markers.

Marker transparency is set to 80%. This

reduces clutter and helps to view the
regions of higher density.

title 'MPG by Horsepower for Cars';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city /
markerattrs=graphdata2(symbol=circlefilled) transparency=0.8;
xaxis grid;
yaxis grid;
run;

Figure 4.4.2: Grouped Scatter Plot

When a GROUP variable is set, the

marker attributes of each group are
derived from the GraphData (1-N) style
elements.

An automatic legend is created for group

values. We have used the KEYLEGEND
statement to position the legend inside the
data area in the upper right corner.

title 'MPG by Horsepower for Cars by Type';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city / group=type;
keylegend / title='Type: ' location=inside across=1;
xaxis grid;
yaxis grid;
run;
66 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.4.3: Two Scatter Plot Overlay

This is an overlay of two scatter plots. The

procedure automatically assigns each plot
with different visual attributes.

The legend is automatically generated. The

KEYLEGEND statement is used here to
place the legend in the upper right corner.

title 'Mileage by Horsepower';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city;
scatter x=horsepower y=mpg_highway;
keylegend / title='Type: ' location=inside across=1;
xaxis grid;
yaxis grid label='Mileage'; run;

Figure 4.4.4: Two Scatter Plot with

Custom Attributes

This graph uses overlay of two scatter

plots. The MARKERATTRS option is used
to specify the marker attributes for each
plot.

Legend is positioned in the upper right

corner of the data area by using the
KEYLEGEND statement.

title 'Mileage by Horsepower';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city / transparency=0.7
markerattrs=(symbol=circlefilled);
scatter x=horsepower y=mpg_highway / transparency=0.7
markerattrs=(symbol=trianglefilled);
keylegend / title='Type: ' location=inside across=1;
yaxis grid label='Mileage';
xaxis grid; run;
 Chapter 4 Basic Plots 67

Figure 4.4.5: Scatter with Y Upper Error

This is a scatter plot with Y upper error limit.

The MARKERATTRS option is used to set

the marker attributes. The x- and y-axis
gridlines are enabled.

title 'Uncertain Data';

proc sgplot data=sgbook.limits;
scatter x=x y=y / yerrorupper=yh markerattrs=(symbol=circlefilled);
xaxis grid;
yaxis grid;
run;

Figure 4.4.6: Scatter with Y Lower and

Upper Error

This is a scatter plot with Y upper and

lower error limits.

The MARKERATTRS option is used to set

the marker attributes. The x- and y-axis
gridlines are enabled.

title 'Uncertain Data';

proc sgplot data=sgbook.limits noautolegend;
scatter x=x y=y / yerrorlower=yl yerrorupper=yh
markerattrs=(symbol=circlefilled);
xaxis grid;
yaxis grid;
run;
68 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.4.7: Scatter with X Lower and

Upper Error

This scatter plot displays X lower and upper

error limits.

We have created filled markers with outline

by overlaying two scatter plots. The
NOAUTOLEGEND and NOCYCLEATTRS
options are used to switch off automatic
creation of the legend and cycling of the
attributes.

title 'Uncertain Data';

proc sgplot data=sgbook.limits noautolegend nocycleattrs;
scatter x=x y=y / xerrorlower=xl xerrorupper=xh
markerattrs=(size=11);
scatter x=x y=y / markerattrs=(symbol=circlefilled color=white);
xaxis grid; yaxis grid;
run;

Figure 4.4.8: Scatter with X and Y Limits

This scatter plot displays both lower and

upper limits for X and Y.

We have created filled markers with outline

by overlaying two scatter plots. The
NOAUTOLEGEND and NOCYCLEATTRS
options are used to switch off automatic
creation of the legend and cycling of the
attributes.

title 'Uncertain Data';

proc sgplot data=sgbook.limits noautolegend nocycleattrs;
scatter x=x y=y / xerrorlower=xl xerrorupper=xh
yerrorlower=yl yerrorupper=yh markerattrs=(size=11);
scatter x=x y=y / markerattrs=( symbol=circlefilled color=white);
xaxis grid; yaxis grid;
run;
 Chapter 4 Basic Plots 69

Figure 4.4.9: Grouped Scatter with X and

Y Limits

This shows a grouped scatter plot with X

and Y limits.

To create the gap between the error bars

and the markers, we have used multiple
scatter plot statement to create this effect.
This makes the marker shape more clearly
visible.

title 'Uncertain Data';

proc sgplot data=sgbook.limits nocycleattrs;
scatter x=x y=y / group=group name='a'
xerrorlower=xl xerrorupper=xh yerrorlower=yl yerrorupper=yh;
scatter x=x y=y / group=group markerattrs=(size=13 color=white);
scatter x=x y=y / group=group;
keylegend 'a' / location=inside position=topright;
xaxis grid;
yaxis grid;
run;

Figure 4.4.10: Scatter with X and Y Limits

This shows a grouped scatter plot with X and

Y limits.

This graph uses a custom style that uses

“filled” markers. Error bar attributes are set
to a dash pattern and thicker line. The end
cap size is tied to line thickness.

title 'Uncertain Data';

proc sgplot data=sgbook.limits;
scatter x=x y=y / xerrorlower=xl xerrorupper=xh
yerrorlower=yl yerrorupper=yh group=group
errorBarAttrs=(thickness=2 pattern=dash);
keylegend / location=inside position=topright;
xaxis grid;
yaxis grid;
run;
70 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.4.11: Scatter Plot with Y2 Axis

This is a basic scatter plot with y-axis

displayed on the right side (Y2).

We are using the custom style that uses filled

markers. Transparency is set for a better
view of the density of the observations.

Note the position of the y-axis as indicated by

the arrow.

title 'MPG by Horsepower using Y2 Axis';

proc sgplot data=ssashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city / transparency=0.8 Y2axis
markerattrs=(symbol=circlefilled);
xaxis grid;
y2axis grid;
run;

Figure 4.4.12: Scatter Plot with X2 and Y2

Axes

This is a basic scatter plot with x-axis

displayed at the top (X2) and y-axis
displayed on the right (Y2).

We are using the custom style that uses

filled markers. Markers are transparent.

Note the positions of the axes as indicated

by the arrows.

title 'MPG by Horsepower using X2 and Y2 Axis';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city / transparency=0.8 x2axis Y2axis
markerattrs=(symbol=circlefilled);
x2axis grid;
y2axis grid;
run;
 Chapter 4 Basic Plots 71

Figure 4.4.13: Scatter with Reflected Ticks

on Y2

This is a basic scatter plot with reflected ticks

on the Y2 axis.

We are using a custom style with filled

markers. This graph uses the GraphData3
style element that has square, filled symbols.

The ref ticks are indicated by the arrow.

title 'MPG for Cars with Y2 ticks';

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=horsepower y=mpg_city / markerattrs=(symbol=squarefilled)
transparency=0.8;
xaxis grid;
yaxis grid refticks;
run;

Figure 4.4.14: Scatter with X, Y, and Y2

Axis

Low and high average temperatures by

month are plotted using two scatter plots.
Temperature in Fahrenheit is shown on the
y-axis and in Celsius on the Y2 axis.

The title uses unicode text. XAXIS

OFFSETMIN is used to leave extra space for
the REFLINE label. The LEGENDLABEL
option is used to name each plot in the
legend.

ods escapechar='^';
title "Average Temperatures in ^{unicode '00b0'x}C and ^{unicode '00b0'x}F";
proc sgplot data=sgbook.weather;
refline 32 / label='32' labelloc=inside labelpos=min;
scatter x=month y=high / legendlabel='High';
scatter x=month y=lowc / y2axis legendlabel='Low';
xaxis grid valueattrs=(size=6) offsetmin=0.1;
yaxis grid min=14 max=104;
y2axis min=-10 max=40;
keylegend / location=inside position=topleft across=1;
run;
72 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.4.15: Scatter with Groups on

Discrete Axis

Here is a scatter plot of MPG by origin, with

GROUP=Type. The x-axis variable is
discrete. All markers are overlaid at the (x,
y) location, resulting in a dense plot.

It is not easy to distinguish the different

groups. A color style would have helped a
bit. X-axis offsets are used to move the
markers inwards.

title "MPG by Origin and Type";

proc sgplot data=sashelp.cars;
scatter x=origin y=mpg_city / group=type transparency=0.8;
xaxis display=(nolabel) offsetmin=0.1 offsetmax=0.1;
yaxis grid;
keylegend / location=inside position=topright across=2;
run;

Figure 4.4.16: Scatter with Clustered

Groups on Discrete Axis

In this graph, we have the same data as in

Figure 4.4.15, but we are using
GROUPDISPLAY=Cluster.

When the x-axis variable is discrete, the

GROUPDISPLAY option creates clusters for
each group around the mid-point value, just
like a bar chart.

title "MPG by Origin and Type";

proc sgplot data=sashelp.cars; 9.3
scatter x=origin y=mpg_city / group=type groupdisplay=cluster
transparency=0.8;
xaxis display=(nolabel);
yaxis grid;
keylegend / location=inside position=topright across=2;
run;
 Chapter 4 Basic Plots 73

Figure 4.4.17: Scatter Overlay

This graph is an overlay of two scatter plots

of city and highway mileage by country of
origin. The markers for each are overlaid,
resulting in a cluttered plot.

To leave more room on the outside of the x-

axis, x-axis offsets are used.

See Figure 4.4.18 below.

title "City and Highway Mileage by Type";

proc sgplot data=sashelp.cars;
where type='Sedan' or type='Sports';
scatter x=origin y=mpg_city / transparency=0.8;
scatter x=origin y=mpg_highway / transparency=0.8;
xaxis display=(nolabel) offsetmin=0.1 offsetmax=0.1;
yaxis grid label='Mileage';
keylegend / location=inside position=topright across=1;
run;

Figure 4.4.18: Scatter Overlay with

Discrete Offset

This graph has the same data as for Figure

4.4.17. Here, we have offset each scatter
plot a little bit using the DISCRETEOFFSET
option.

This creates a separation between the

different Y values, making the graph more
readable.

title "City and Highway Mileage by Type";

proc sgplot data=sashelp.cars; 9.3
where type='Sedan' or type='Sports';
scatter x=origin y=mpg_city / transparency=0.8 discreteoffset=-0.1;
scatter x=origin y=mpg_highway / transparency=0.8 discreteoffset=+0.1;
xaxis display=(nolabel) offsetmin=0.2 offsetmax=0.2;
yaxis grid label='Mileage';
keylegend / location=inside position=topright across=1;
run;
74 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

4.5 Scatter Plots with Data Labels

The SCATTER statement supports two ways to display data values as

character strings:

• MARKERCHAR = column
• DATALABEL < = column >

4.5.1 DATALABEL Option

The scatter plot displays a marker at the (x, y) location for each observation. The
SCATTER statement also supports display of data labels with each marker using the
DATALABEL option. In this case, the associated data label is displayed near the marker:

• DATALABEL option displays the Y variable value.

• DATALABEL=column displays the values from the specified column.

By default, a data label collision avoidance algorithm is used to minimize the collisions
between labels and other markers in the plot. Plots with data labels can get busy very
quickly. When using data labels we recommend the following:

• Use short data labels.

• Restrict the number of markers.
• For moderately large number of markers, restrict the number of data labels. This can
be done by selectively setting data labels to missing for some of the observations.

The data label collision avoidance algorithm attempts to move the labels to avoid collision
with markers or other labels. This can sometimes move a label away from the marker. One
technique to retain context is to use group colors as shown later.

4.5.2 MARKERCHAR Option

Often it is desirable to display a character string at the (x, y) position for each observation
instead of a marker. The MARKERCHAR=column option draws a character string for the
value from the column for each observation at the (x, y) location in the plot area.

Figure 4.5.5 extends the example in Figure 4.4.14 to display the temperature value for each
observation. This makes it easier to see both the trend and the actual value.
 Chapter 4 Basic Plots 75

Figure 4.5.1: Scatter with Data Labels

In this scatter plot we have used the

DATALABEL option to display the data
labels for Asian sedans with MPG > 27.

There are only 16 observations in this

graph. However, due to relatively long
labels, the graph quickly becomes cluttered.
If all Asian sedans were displayed, the
graph would be unreadable.

title 'MPG for Cars with Data Labels';

proc sgplot data=sgbook.cars3;
where type='Sedan' and origin='Asia' and mpgc > 27;
scatter x=hp y=mpg / datalabel=model markerattrs=(symbol=circlefilled);
xaxis grid;
yaxis grid;
run;

Figure 4.5.2: Scatter with Data Labels

In this scatter plot with data labels, we

display only the observations for Asian
sedans with MPG > 32.

When using data labels, it is generally better

to reduce the displayed data to avoid a
cluttered graph.

title 'Cars with MPG > 32';

proc sgplot data=sgbook.cars3 noautolegend;
where type='Sedan' and origin='Asia' and mpgc > 32;
scatter x=hp y=mpg / datalabel=model;
xaxis grid;
yaxis grid;
run;
76 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.5.3: Scatter with Short Labels

This is a basic scatter plot with short data

labels for Asian sedans with MPG > 30.

In this example, we have shortened the data

labels to 10 characters. Here we have 10
observations. Both readability and context
are maintained by using shorter labels.
Using the GROUP variable would help if we
were using a color style.
2
title 'Cars with MPG > 30 and Short Labels';
proc sgplot data=sgbook.cars3 noautolegend;
where type='Sedan' and origin='Asia' and mpgc > 30;
scatter x=hp y=mpg / datalabel=model2 group=model2;
xaxis grid;
yaxis grid;
run;

Figure 4.5.4: Scatter with Reduced

Number of Data Labels

This a scatter plot for all cars, showing data

labels only for high-mileage or high-
horsepower cars.

This example has 428 observations.

However, we have modified the data set to
display labels only for the exceptional cars
by setting the other labels to missing.

title 'Cars with Restricted Labels';

proc sgplot data=sgbook.cars4 noautolegend;
scatter x=hp y=mpg / datalabel=model3 group=model3;
xaxis grid;
yaxis grid;
run;
 Chapter 4 Basic Plots 77

Figure 4.5.5: Scatter with Marker Characters

The MARKERCHAR option can be used to display a
character string for the data in the column at the (x,
y) location in the plot. Here we have displayed the
high and low temperature values for each
observation within the marker.

A Refline for y=32 is displayed.

title "Average Temperatures in ^{unicode '00b0'x}C and ^{unicode '00b0'x}F";

proc sgplot data=sgbook.weather noautolegend;
refline 32 / label='32' labelloc=inside labelpos=min;
scatter x=month y=high / markerattrs=(symbol=circle size=15);
scatter x=month y=high / markerchar=high;
scatter x=month y=lowc / markerattrs=(symbol=circle size=15) y2axis;
scatter x=month y=low / markerchar=low;
xaxis grid offsetmin=0.1;
yaxis grid min=14 max=104;
y2axis min=-10 max=40; run;

Figure 4.5.6: Scatter with Axis Aligned Statistics

Table
The MARKERCHAR option is used to create a table
of statistics that is aligned with a categorical y-axis.

REFLINE statements with DISCRETEOFFSET are

used to create the vertical column borders. X2 axis
is used with DISPLAY option set to simulate column
headers.

title "Average Temperatures for Raleigh";

proc sgplot data=sgbook.weatherTable noautolegend; 9.3
scatter y=month x=x1 / markerchar=high x2axis;
scatter y=month x=x2 / markerchar=low x2axis;
scatter y=month x=x3 / markerchar=highc x2axis;
scatter y=month x=x4 / markerchar=lowc x2axis;
scatter y=month x=x5 / markerchar=precip x2axis;
refline 'High(F)' 'Low(F)' 'High(C)' 'Low(C)' / axis=x2 discreteoffset=0.5;
x2axis display=(noticks nolabel) valueattrs=(size=7);
yaxis display=(noticks) reverse valueattrs=(size=7);
run;
78 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

4.6 Series Plot

The basic series plot is a key plot type for visualization of raw data.
This plot does not summarize the data. The syntax is:

series x=column y=column </ options>;

A series is drawn through each (x, y) location in data order. The

GROUP role is used to create multiple series plots, one for each
unique group value.

Required Data Roles:

X =column Position along the horizontal axis
Y =column Position along the vertical axis

Optional Data Roles:

DATALABEL =column Label String for each observation
GROUP =column Classification for multiple series
URL =column URL for each observation

Appearance Options:
LINEATTRS =line-attrs Appearance attributes for lines
MARKERATTRS =marker-attrs Appearance attributes for markers

Other Options:
CURVELABEL =string Curve label for non-grouped case
CURVELABELLOCATION =keyword Curve label location (INSIDE | OUTSIDE)
CURVELABELPOSITION =keyword Curve label position (TOP, BOTTOM, …)

Boolean Options:
BREAK Break the series plot at missing values
CURVELABEL Display curve labels from group role
DATALABEL Display Y value as data label
MARKERS Display the markers

Standard common options are supported

ATTRID, CLUSTERWIDTH, DISCRETEOFFSET, GROUPDISPLAY, GROUPORDER,
LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, URL, X2AXIS,
Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
 Chapter 4 Basic Plots 79

Figure 4.6.1: Basic Series Plot

This graph shows patient response to a

drug over time.

The x-axis label is suppressed, and YAXIS

MIN is set to zero.

title 'Response over Time';

proc sgplot data=sgbook.seriesGroup;
where drug='Drug A';
series x=date y=val / lineattrs=(thickness=2);
xaxis display=(nolabel);
yaxis min=0 integer grid label='Efficacy';
run;

Figure 4.6.2: Grouped Series Plot with

Legend

This graph shows response over time by

drug. A legend is generated by default.
We have used the KEYLEGEND statement
to suppress the legend title.

Since the number of observations in the

data exceeds the default limit for anti-
aliasing, set ANTIALIASMAX=1100 to
restore anti-aliasing.

ods graphics / reset antialiasmax=1100;

title 'Response over Time by Drug'; 9.3
proc sgplot data=sgbook.seriesGroup;
series x=date y=val / group=drug lineattrs=(thickness=2);
xaxis display=(nolabel);
yaxis min=0 integer grid label='Efficacy';
keylegend / title='' valueattrs=(size=7);
run;
80 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.6.3: Grouped Series Plot with

Curve Labels

In this graph the KEYLEGEND statement is

used to move the legend closer to the data.
This reduces the amount of eye movement
needed to decode which series represents
which drug.

See Figure 4.6.4 for an alternative

arrangement.

title 'Response over Time by Drug';

proc sgplot data=sgbook.seriesGroup;
series x=date y=val / group=drug lineattrs=(thickness=2);
xaxis display=(nolabel);
yaxis min=0 integer grid label='Efficacy';
keylegend / title='' location=inside position=bottomleft across=1;
run;

Figure 4.6.4: Grouped Series Plot with

Markers

In this graph the legend is eliminated by

labeling each series with the name of the
drug. This eliminates eye movement
required to decode the information, thus
increasing the effectiveness of the graph.

The procedure detects the presence of the

CURVELABEL option, and suppresses the
legend automatically.

title 'Response over Time by Drug';

proc sgplot data=sgbook.seriesGroup;
series x=date y=val / group=drug
lineattrs=(thickness=2) curvelabel;
xaxis display=(nolabel);
yaxis min=0 integer grid label='Efficacy';
run;
 Chapter 4 Basic Plots 81

Figure 4.6.5: Grouped Series Plot with

Markers

In this graph we have displayed only some

of the markers. This is done using an
overlaid scatter plot with the FREQ= option.
With this option, a marker is plotted only
when the value of the FREQ variable is > 0.

We did this instead of using the markers

option on the series plot as that would
display a marker at each point on the line.

title 'Response over Time by Drug';

proc sgplot data=sgbook.seriesGroup noautolegend;
series x=date y=val / group=drug
lineattrs=(thickness=2) curvelabel;
scatter x=date y=val / group=drug freq=freq markerattrs=(size=5);
xaxis display=(nolabel);
yaxis min=0 integer grid label='Efficacy';
run;

Figure 4.6.6: Series Plot with Data Labels

In this graph, only some data labels are

displayed using an overlaid scatter plot.
We could display data labels using the
DATALABEL option of the series plot.
However, that would display a label for
every observation, cluttering the plot.

On the scatter plot, we use Y=LabelA which

th
has non-missing values for only every 50
obs, thus reducing the markers and labels.

title 'Response over Time with data labels';

proc sgplot data=sgbook.series noautolegend;
series x=date y=a / lineattrs=(thickness=2) datalabel=labela;
scatter x=date y=labela / markerattrs=(size=5);
xaxis display=(nolabel);
yaxis min=0 max=15 integer grid label='Efficacy';
run;
82 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.6.7: Series Plot with Breaks

In this graph, some values of the response

variable are missing. By default, the series
plot would be drawn with a straight line over
the missing data.

Using the BREAK option causes the series

plot to be drawn with a break at the missing
data.

title 'Response over Time with break';

proc sgplot data=sgbook.seriesBreak;
series x=date y=c / lineattrs=(thickness=2) break;
xaxis display=(nolabel);
yaxis min=0 max=15 integer grid label='Efficacy';
run;

Figure 4.6.8: Series Plot Overlay with

Discrete Offset

This graph overlays three series plots of

sales by quarter on a discrete x-axis. The
DISCRETEOFFSET option is set for each
curve to create a horizontal separation
between the curves.

title 'Sales by Quarter';

proc sgplot data=sgbook.seriesDiscrete; 9.3
series x=qtr y=salesa / discreteoffset=-0.2;
series x=qtr y=salesb / discreteoffset= 0.0;
series x=qtr y=salesc / discreteoffset=+0.2;
xaxis grid display=(nolabel) offsetmin=0.1 offsetmax=0.1;
yaxis grid label='Sales';
keylegend / title='Drug: ';
run;
 Chapter 4 Basic Plots 83

Figure 4.6.9: Series Plot with Group

Overlay

This is a graph with a grouped series plot

overlaid with a scatter plot that has error
bars on a discrete x-axis. The groups are
overlaid on the same midpoint values along
the x-axis.

This can obscure the data. See alternative

arrangement in Figure 4.6.10.

title 'Sales by Quarter';

proc sgplot data=sgbook.seriesDiscreteGroup;
scatter x=qtr y=sales / yerrorupper=high yerrorlower=low
group=drug;
series x=qtr y=sales / group=drug;
xaxis grid display=(nolabel);
yaxis grid;
run;

Figure 4.6.10: Series Plot with Group

Cluster

This graph is the same as in Figure 4.6.9,

with GROUPDISPLAY=Cluster. In this
case, both the scatter and series plots use
the same settings.

This causes the groups to be displayed with

an offset along the x-axis, thus making the
data for each drug more readable.
CLUSTERWIDTH determines the offsets.

title 'Sales by Quarter';

proc sgplot data=sgbook.seriesDiscreteGroup; 9.3
scatter x=qtr y=sales / group=drug yerrorupper=high yerrorlower=low
groupdisplay=cluster clusterwidth=0.6;
series x=qtr y=sales / group=drug lineattrs=(thickness=1)
groupdisplay=cluster clusterwidth=0.6;
xaxis grid display=(nolabel);
yaxis grid;
run;
84 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

4.7 Step Plot

The step plot is similar to the series plot with a step join for each
value. The syntax is:

step x=column y=column </ options>;

A step plot draws a horizontal line through the Y value, with a step
change at the next observation.

Required Data Roles:

X =column Position along the horizontal axis
Y =num-column Position along the vertical axis

Optional Data Roles:

DATALABEL =column Label String for each observation
GROUP =column Classification for multiple series
URL =column URL for each observation
YERRORLOWER =num-column Lower value for Y Error Bar
YERRORUPPER =num-column Upper value for Y Error Bar

Appearance Options:
ERRORBARATTRS =line-attrs Appearance attributes for error bars
LINEATTRS =line-attrs Appearance attributes for lines
MARKERATTRS =marker-attrs Appearance attributes for markers

Other Options:
CURVELABEL =string Curve label for non-grouped case
CURVELABELLOCATION =keyword Curve label location (INSIDE | OUTSIDE)
CURVELABELPOSITION =keyword Curve label position (TOP, BOTTOM, etc…)
JUSTIFY =value Location of data point relative to step

Boolean Options:
BREAK Break the series plot at missing values
CURVELABEL Display curve labels from group role
DATALABEL Display Y value as data label
MARKERS Display the markers

Standard common options are supported

ATTRID, CLUSTERWIDTH, DISCRETEOFFSET, GROUPDISPLAY, GROUPORDER,
LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, URL, X2AXIS, Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
 Chapter 4 Basic Plots 85

Figure 4.7.1: Basic Step Plot

This is a basic step plot of a response over

time for Drug B.

An INSET statement is used to indicate the

data is for one of the drugs (B) in the data.

title 'Response over Time';

proc sgplot data=sgbook.stepGroup;
where drug='Drug B';
step x=date y=val / lineattrs=(thickness=2);
inset "Drug B" / position=topright border;
xaxis display=(nolabel);
yaxis grid label='Response';
run;

Figure 4.7.2: Grouped Step Plot

This is a grouped step plot of a measure

over time for three drugs.

The legend is automatically created for this

group case. The KEYLEGEND statement
is used to position it inside the data area in
a stack.

YAXIS MIN is set to zero.

title 'Response over Time by Drug';

proc sgplot data=sgbook.stepGroup;
step x=date y=val / group=drug lineattrs=(thickness=2);
keylegend / location=inside position=bottomleft across=1;
xaxis display=(nolabel);
yaxis min=0 max=20 integer grid label='Response';
run;
86 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.7.3: Grouped Step Plot with

Markers

This is a grouped step plot of efficacy over

time for three drugs. The LINEATTRS
option is used to set the lines patterns to
solid.

The markers are displayed. The default

justification is left, where the data point is
at the left end of the horizontal step. The
line first goes horizontally to the right from
the data point, then up.

title 'Response over Time by Drug';

proc sgplot data=sgbook.stepGroup;
step x=date y=val / group=drug markers
lineattrs=(pattern=solid thickness=1);
keylegend / location=inside position=bottomleft across=1;
xaxis display=(nolabel);
yaxis min=0 integer grid label='Efficacy';
run;

Figure 4.7.4: Grouped Step Plot with

Center Justified Markers

This is a grouped step plot of efficacy over

time for three drugs.

The markers are displayed. The JUSTIFY

option is set to center. Now the data point
is at the center of the horizontal step.

title 'Response over Time by Drug';

proc sgplot data=sgbook.stepGroup;
step x=date y=val / group=drug markers justify=center
lineattrs=(pattern=solid thickness=1) ;
keylegend / location=inside position=bottomleft across=1;
xaxis display=(nolabel);
yaxis min=0 integer grid label='Efficacy';
run;
 Chapter 4 Basic Plots 87

Figure 4.7.5: Grouped Step Plot with

Upper Limit

This is a step plot of response over time by

drug with upper limit values. Note the step
plot itself supports options for limits, unlike
the series plot, which does not.

The markers are displayed center justified

with reduced size. The error bar pattern is
the same as the pattern for the line.

title 'Response over Time with Limit by Drug';

proc sgplot data=sgbook.stepGroup;
step x=date y=val / yerrorupper=upper group=drug markers
justify=center markerattrs=(size=5);
keylegend / location=inside position=bottomleft across=1;
xaxis display=(nolabel);
yaxis min=0 integer grid label='Response';
run;

Figure 4.7.6: Grouped Step Plot with

Both Limits and Error Bar Attributes

This is a step plot of response over time by

drug with both upper and lower limit values.
Note the step plot itself supports options for
limits, unlike the series plot, which does not.

The markers are displayed center justified

with a reduced size. The line pattern is set
to solid. The color for the error bars is also
set.

title 'Response over Time with Limits by Drug';

proc sgplot data=sgbook.stepGroup;
step x=date y=val / yerrorupper=upper yerrorlower=lower group=drug
markers justify=center lineattrs=(pattern=solid)
markerattrs=(size=5)errorbarattrs=(color=grey);
keylegend / location=inside position=bottomleft across=1;
xaxis display=(nolabel);
yaxis min=0 integer grid label='Response';
run;
88 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.7.7: Overlay Step Plot with

Curve Labels

Here is an overlay of two step plots with

curve labels. CURVELABEL option is
used to label the curves. This provides a
direct association between name and the
curve.

The procedure detects the presence of the

CURVELABEL option and suppresses the
automatic legend.

title 'Response over Time by Drug';

proc sgplot data=sgbook.step;
step x=date y=A / justify=center curvelabel='Drug A';
step x=date y=B / justify=center curvelabel='Drug B';
xaxis display=(nolabel);
yaxis min=0 integer grid label='Response';
run;

Figure 4.7.8 Overlay Step Plot with Break

Here is the same graph as in Figure 4.7.7,

with a break in the data. For Drug A, the
BREAK option is not used. For Drug B, the
BREAK option is used.

The missing data is between 10Jun and

08Oct, as can be seen by the two reference
lines. Drug A plot has one step between
the two values. Drug B plot has a break
between these values.

title 'Response over Time by Drug';

proc sgplot data=sgbook.stepBreak;
step x=date y=A / justify=center curvelabel='Drug A';
step x=date y=B / justify=center curvelabel='Drug B' break;
refline '10jun2009'd '08oct2009'd / axis=x;
xaxis display=(nolabel);
yaxis min=0 integer grid label='Response';
run;
 Chapter 4 Basic Plots 89

4.8 Band Plot

The band plot draws a band between two response values. The syntax
can be one of the following:

band x=column lower=column|value upper= column|value

</ options>;
band y=column lower=column|value upper= column|value
</ options>;

For each value of the X (or Y) variable, a band is drawn between the
lower and upper response values. Response values can come from a column in the data or can
be a constant value.

Required Data Roles:

X or Y =column Position along horizontal or vertical axis
LOWER =num-column | value Lower response
UPPER =num-column | value Upper response

Optional Data Roles:

GROUP =column Classification for multiple series

Appearance Options:
CURVELABELATTRS =text-attrs Appearance of curve label
FILLATTRS =fill-attrs Appearance of fill between values
LINEATTRS =line-attrs Appearance of the lines

Other Options:
CURVELABELLOC =keyword Location of label – INSIDE | OUTSIDE
CURVELABELPOS =keyword Position of label - START | END
CURVELABELLOWER =string Name of lower band curve
CURVELABELUPPER =string Name of upper band curve
MODELNAME =plotname Name of the plot for interpolation info.
NAME =string Name for this statement
TRANSPARENCY =value Transparency for the line and markers
TYPE =value Series or Step

Boolean Options:
FILL | NOFILL Display fill between lower and upper values
OUTLINE | NOOUTLINE Display outline at lower and upper values
NOEXTEND

Standard common options are supported

ATTRID, LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, X2AXIS,
Y2AXIS
90 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.

The band may be filled, with or without an outline. The GROUP role is used to create multiple
bands, one for each unique group value. When used with a series or step plot to display a
confidence or prediction band, the MODELNAME option can be used to create an association
with the fit plot.

Note: The band plot statement is optimal for the display of confidence and prediction bands,
where the X (or Y) values are sorted. The band is drawn in data order, and if the data are not
sorted by X (or Y), then the results may be unpredictable.

Figure 4.8.1: Basic Band Plot

This is a basic band plot along the x-axis,

with upper and lower limits in the y-axis
direction.

title 'Response over Time';

proc sgplot data=sgbook.bandGroup;
where drug='Drug B';
band x=date upper=upper lower=lower;
inset "Drug B" / position=topright border;
xaxis display=(nolabel);
yaxis grid label='Response';
run;
 Chapter 4 Basic Plots 91

Figure 4.8.2: Grouped Band Plot

This is a band plot of upper and lower Y

response by X and drug. The legend is
automatically created.

The x-axis label is suppressed.

title 'Response over Time';

proc sgplot data=sgbook.bandGroup;
band x=date upper=upper lower=lower / group=drug fill;
xaxis display=(nolabel);
yaxis grid label='Response';
run;

Figure 4.8.3: Grouped Band Plot with

Transparency

This is a band plot of upper and lower Y

response by X and drug. The legend is
automatically created.

Use of transparency allows the viewing of

the overlapped regions of the bands.

The x-axis label is suppressed.

title 'Response over Time';

proc sgplot data=sgbook.bandGroup;
band x=date upper=upper lower=lower / group=drug transparency=0.3;
xaxis display=(nolabel);
yaxis grid label='Response';
run;
92 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.8.4: Overlay Band Plots

This is an overlay of two band plots with

independent columns for upper and lower
responses by date for each drug. The
individual bands statements can be assigned
different visual attributes.

The legend is automatically created. The

legend entries are derived from the
LEGENDLABEL settings for each plot.

title 'Response over Time';

proc sgplot data=sgbook.bandBreak;
band x=date upper=upperB lower=lowerB / transparency=0.3
fill outline legendlabel='Drug B';
band x=date upper=upperC lower=lowerC / transparency=0.3
outline legendlabel='Drug C';
xaxis display=(nolabel);
yaxis grid label='Response';
run;

Figure 4.8.5: Overlay Band Plots with

Constant Lower Limit

This is an overlay of two band plots with

independent columns for upper and lower
responses by date for each drug. The first
band has LOWER=0 (constant).

A “fill under line” effect can be created using

a band in this way. LEGENDLABEL options
are used to name the bands in the legend.

title 'Response over Time';

proc sgplot data=sgbook.bandBreak;
band x=date upper=upperB lower=0 / transparency=0.5
legendlabel='Drug B';
band x=date upper=upperC lower=lowerC / transparency=0.5
legendlabel='Drug C';
xaxis display=(nolabel);
yaxis grid label='Response';
run;
 Chapter 4 Basic Plots 93

Figure 4.8.6: Overlay Band Plots with

Curve Labels

This is a graph with two overlaid band plots.

The CURVELABEL options are used to
display the upper and lower labels for one of
the band plots.

CURVELABELPOS option can be set to

start or end. Since ‘end’ is the default
setting, it need not be specified in this
program.

title 'Response over Time';

proc sgplot data=sgbook.bandBreak; 9.3
band x=date upper=upperB lower=lowerB / transparency=0.5
curvelabelupper='High' curvelabellower='Low' curvelabelpos=end
legendlabel='Drug B';
band x=date upper=upperC lower=lowerC / transparency=0.5
legendlabel='Drug C';
xaxis display=(nolabel);
yaxis grid label='Response'; run;

Figure 4.8.7: Overlay Step and Band

Plots

This graph shows a step plot of the

response value along with a band plot of the
limits.

MODELNAME=”Value” is assigned on the

band plot, where “Value” is the name of the
step plot.

The Band plot derives its join

characteristics, such as the step shape and
the justification from the step plot.

title 'Response over Time';

proc sgplot data=sgbook.stepBand;
band x=date upper=upperB lower=lowerB / transparency=0.5
modelname='Value' legendlabel='Limit';
step x=date y=B / justify=center name='Value' legendlabel='Value';
xaxis display=(nolabel);
yaxis grid label='Response';
run;
94 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

4.9 Needle Plot

The needle plot draws a vertical needle for each observation from the
(x,y) coordinate to the x-axis. The syntax is:

needle x=column y=num-column </ options>;

A vertical line (needle) is drawn from each observation to the baseline.

For Interval x-axis, the GROUP role affects only the display attributes of
the needle. Cluster groups are supported for discrete axis.

Required Data Roles:

X =column Position along the horizontal axis
Y =num-column Position along the vertical axis

Optional Data Roles:

DATALABEL =column Label String for each observation
GROUP =column Classification for multiple series
URL =column URL for each observation

Appearance Options:
DATALABELATTRS =text-attrs Appearance attributes for labels
LINEATTRS =line-attrs Appearance attributes for the needles
MARKERATTRS =marker-attrs Appearance attributes for markers

Other Options:
BASELINE =number Y-intercept for the baseline

Boolean Options:
DATALABEL Display Y value as data label
MARKERS Display the markers

Standard common options are supported

ATTRID, DISCRETEOFFSET, GROUPDISPLAY, GROUPORDER, LEGENDLABEL,
NAME, NOMISSINGGROUP, TRANSPARENCY, URL, X2AXIS, Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
 Chapter 4 Basic Plots 95

Figure 4.9.1: Basic Needle Plot

This is a basic needle plot of trading

volume by month.

LINEATTRS option is used to set the

needle thickness to 3px.

title 'Trading Volume over Time';

proc sgplot data=sgbook.needleTrade;
needle x=date y=val / lineattrs=(thickness=3);
xaxis display=(nolabel);
yaxis grid label='(in Millions)';
run;

Figure 4.9.2: Grouped Needle Plot

This is a grouped needle plot of trading

volume by month. The trade is either a
“Buy” or a “Sell”. This is used as the
GROUP role.

An OFFSETMAX=0.1 is set for y-axis to

make room for legend that has been placed
inside.

title 'Trading Volume over Time by Type';

proc sgplot data=sgbook.needleTrade;
needle x=date y=val / lineattrs=(thickness=3) group=trade;
xaxis display=(nolabel);
yaxis grid label='(in Millions)' offsetmax=0.1;
keylegend / location=inside position=topright;
run;
96 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.9.3: Needle Plot with a Baseline

This is a needle plot of trades over time.

BASELINE=15 is set, so needles are drawn
to this baseline.

Markers and data labels are displayed.

Data labels are automatically moved to
avoid collision with other labels

Needle thickness is set to 3px.

title 'Trades over Time';

proc sgplot data=sgbook.needleTrade;
needle x=date y=val / lineattrs=(thickness=3) baseline=15
datalabel=val markers;
xaxis display=(nolabel);
yaxis grid label='Trade';
run;

Figure 4.9.4: Needle Plot with Discrete

Groups

This is a needle plot of trades by month,

grouped by trade type (Buy or Sell).

In this case, the x-axis is discrete, so the

groups can be displayed as clusters. The
CLUSTERWIDTH option is used to reduce
the cluster width to get tighter clustering.

title 'Trades by Month';

proc sgplot data=sgbook.needleDiscrete; 9.3
needle x=month y=val / group=trade lineattrs=(thickness=3)
groupdisplay=cluster clusterwidth=0.4;
xaxis display=(nolabel);
yaxis grid label='Trade';
run;
 Chapter 4 Basic Plots 97

4.10 Vector Plot

The Vector plot draws directional vectors. The syntax is:
vector x=column y=column < / options>;

A vector is drawn to each observation from the origin (0,0) by default.

The origin may be set by using the XORIGIN and YORIGIN roles that
can have data columns or constants.

Required Data Roles:

X =num-column Position of arrow head along the x-axis
Y =num-column Position of arrow head along the y-axis

Optional Data Roles:

DATALABEL =column Label string for each vector
GROUP =column Classification variable
XORIGIN =num-column | value Position of arrow tail along the x-axis
YORIGIN =num-column | value Position of arrow tail along the y-axis

Appearance Options:
DATALABELATTRS =text-attrs Appearance attributes for data labels
LINEATTRS =line-attrs Appearance attributes for lines

Other Options:
ARROWHEADDIRECTION =keyword OUT | IN | BOTH
ARROWHEADSHAPE =keyword OPEN | CLOSED | FILLED | BARBED

Boolean Options:
DATALABEL Display curve labels from group role
NOARROWHEADS Suppress display of arrow heads

Standard common options are supported

ATTRID, LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, X2AXIS,
Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
98 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.10.1: Basic Vector Plot

This is the basic vector plot with default

origin of (0, 0).

A vector is drawn from the origin to the (x,y)

location for each observation.

title 'Random Vectors';

proc sgplot data=sgbook.vector;
vector x=x y=y;
run;

Figure 4.10.2: Vector Plot

This is a vector plot where all the vectors

have a user defined common constant origin
(0.4, 0.4) specified by the XORIGIN and
YORIGIN roles.

The LINEATTRS option is used to set the

line thickness to 2. Note the arrowhead
sizes are proportional to the line thickness.

title 'Random Vectors';

proc sgplot data=sgbook.vector;
vector x=x y=y / xorigin=0.4 yorigin=0.4 lineattrs=(thickness=2);
run;
 Chapter 4 Basic Plots 99

Figure 4.10.3: Vertical Vector Plot

This stock plot is created by using vectors

to draw the monthly closing stock price over
time

ARROWHEADSHAPE option is used to get

barbed arrowheads.

SAS 9.3 supports a new HIGHLOW plot,

which provides another way to create such
a graph.

title 'Stock Plot';

proc sgplot data=sgbook.Stock;
vector x=date y=close / xorigin=date yorigin=open
arrowheadshape=barbed;
xaxis grid;
yaxis grid;
run;

Figure 4.10.4: Grouped Vector Plot

This graph uses a grouped vector plot

where each vector has its own origin.
Increasing and decreasing trend have
different groups. Groups are represented
with different visual attributes.

NOARROWHEADS option is used.

SAS 9.3 supports a new HighLow plot,

which provides another way to create such
a graph.

title 'Stock Plot';

proc sgplot data=sgbook.Stock;
vector x=date y=close / xorigin=date yorigin=open group=trend
noarrowheads lineattrs=(thickness=2);
xaxis grid;
yaxis grid;
run;
100 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

4.11 VBarParm and HBarParm Plots (9.3)

These pre-summarized bar chart statements have the following syntax:
vbarparm column < / response=column> <options>;
hbarparm column < / response=column> <options>;

The VBARPARM and HBARPARM statements create specialized bar

charts in which the response and limit values are displayed as
provided by the user.

Related Style Elements: The GraphDataDefault style element is

used to draw the plot elements. When a GROUP variable is present,
or when multiple plot statements are overlaid, the GraphData1-
GraphData12 elements are used, one for each group or plot
statement.

Required Data Roles:

Category =column Position along the category axis
Response =num-column Response variable

Optional Data Roles:

DATALABEL =column Label String for each observation
GROUP =column Classification for multiple series
URL =column URL for each observation
LIMITLOWER =column Lower limit value
LIMITUPPER =column Upper limit value

Appearance Options:
FILLATTRS =fill-attrs Appearance of bar interior
LIMITATTRS =line-attrs Appearance of limit bars

Other Options:
BARWIDTH = value Width of bar as fraction of tick spacing
DATASKIN = skin-value One of the predefined skin types

Boolean Options:
DATALABEL Display default bar data labels
FILL | NOFILL Display filled / unfilled bars
OUTLINE | NOOUTLINE Display bar outlines or not.
MISSING Accept missing as category value

Standard common options are supported

ATTRID, CLUSTERWIDTH, DISCRETEOFFSET, GROUPDISPLAY, GROUPORDER,
LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, URL
 Chapter 4 Basic Plots 101

Figure 4.11.1: Basic Parametric Bar Chart

This is a basic bar chart of mean MPG (City)

for all cars by origin. The data has been
summarized using the MEANS procedure to
create the data set needed for using this plot
statement.

Since the data was summarized by origin

and type, we have selected _TYPE_=2 for
summary by origin.

title 'Vehicle Mileage by Country';

proc sgplot data=sgbook.carMeans; 9.3
where _type_=2;
vbarparm category=origin response=mean_mpg_city;
run;

Figure 4.11.2: Parametric Bar Chart with

Limits and Data Labels

This is a vertical bar chart of mean MPG

(City) for all cars with limits and labels by
Type.

Because the limits are displayed, the data

label values are displayed below the bars.
The labels can be moved using the
DATALABELPOS option.

title 'Vehicle Mileage by Type';

proc sgplot data=sgbook.carMeans; 9.3
where _type_=1;
vbarparm category=type response=mean_mpg_City / datalabel
limitupper=max_mpg_city limitlower=min_mpg_city
datalabelpos=bottom;
yaxis grid;
run;
102 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.11.3: Grouped Parametric Bar

Chart

This is a grouped bar chart of mileage by

origin and type. GROUPDISPLAY is set to
Cluster for the side-by-side display of the
group values.

This bar chart uses DATASKIN=sheen.

YAXIS OFFSETMAX is set to create some

room for the legend inside the data area.

title 'Vehicle Mileage by Origin and Type';

proc sgplot data=sgbook.carMeans; 9.3
where _type_=3 and (type = 'Sedan' or type='Sports' or type='SUV');
vbarparm category=origin response=mean_mpg_City / group=type
transparency=0.2;
keylegend / location=inside position=topright;
xaxis display=(nolabel); yaxis grid offsetmax=0.2;
run;

Figure 4.11.4: Grouped Parametric Bar

Chart with Data Labels

This grouped bar chart by type is the same

as in Figure 4.11.3. The DATALABEL
option is used to display the bar statistics.

A format is set to reduce the precision on

the data labels. DATASKIN=sheen.

YAXIS OFFSETMAX=0.2 is set to make

room for the legend.

title 'Vehicle Mileage by Origin and Type';

proc sgplot data=sgbook.carMeans; 9.3
where _type_=3 and (type = 'Sedan' or type='Sports' or type='SUV');
format mean_mpg_city 5.1;
vbarparm category=origin response=mean_mpg_City / group=type
dataskin=sheen datalabel;
keylegend / location=inside position=topright;
xaxis display=(nolabel);
yaxis grid offsetmax=0.2; run;
 Chapter 4 Basic Plots 103

Figure 4.11.5: Overlay Parametric Bar

Charts

This graph uses overlay of two VBARPARM

statements, one each for the City and
Highway mileage.

For the common categories, the bars of the

second statement are overlaid directly over
the first, so it is better to set the BARWIDTH
of the second bar to a smaller value.

title 'Vehicle Mileage by Origin';

proc sgplot data=sgbook.carMeans;
where _type_=2;
vbarparm category=origin response=mean_mpg_City;
vbarparm category=origin response=mean_mpg_Highway / barwidth=0.5;
xaxis display=(nolabel);
yaxis grid label='Mileage'; run;

Figure 4.11.6: Overlay Parametric Bar

Charts with Skins and Offsets

Overlay VBARPARM statements, one each

for City and Highway mileage.

Each set of bars is offset from the midpoint

by setting the appropriate value for the
DISCRETEOFFSET option. Bar widths are
reduced appropriately. The values are a
fraction of the category spacing.

ods listing style=journal3;

title 'Vehicle Mileage by Origin'; 9.3
proc sgplot data=sgbook.carMeans;
where _type_=2;
vbarparm category=origin response=mean_mpg_City /
dataskin=pressed discreteoffset=-0.2 barwidth=0.4;
vbarparm category=origin response=mean_mpg_Highway /
dataskin=pressed discreteoffset=0.2 barwidth=0.4;
xaxis display=(nolabel);
yaxis grid;
run;
104 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.11.7: Horizontal Parametric Bar

Chart

This graph uses an HBARPARM plot to

display the mean frequency of adverse
events for a drug.

Horizontal bars are particularly useful when

the category names are long.

title 'Most Frequent Adverse Events';

proc sgplot data=sgbook.AdverseEvents; 9.3
hbarparm category=ae response=mean;
yaxis display=(nolabel);
xaxis grid label='Relative Risk with 95% CL';
run;

Figure 4.11.8: Horizontal Parametric Bar

Chart with Limits

This graph uses an HBARPARM plot to

display the mean frequency of adverse
events for a drug with low and high
confidence limits.

We use the LIMITATTRS option to set the

visual attributes of the limit bars.

title 'Most Frequent Adverse Events';

proc sgplot data=sgbook.AdverseEvents; 9.3
hbarparm category=ae response=mean / limitlower=low limitupper=high
limitattrs=(color=grey pattern=dash);
yaxis display=(nolabel);
xaxis grid label='Relative Risk with 95% CL';
run;
 Chapter 4 Basic Plots 105

4.12 Bubble Plot (9.3)

The BUBBLE statement has the following syntax:
bubble x=column y=column size=num-column </ options>;

Bubble plot draws circular bubbles sized by the Size role at the X and Y
locations in the plot. Radius of the bubbles is scaled linearly by the
SIZE variable.

Required Data Roles:

X column Position along the horizontal axis
Y column Position along the vertical axis
SIZE num-column Size of the bubble

Optional Data Roles:

DATALABEL =column Label String for each observation
GROUP =column Classification for bubble color
URL =column URL for each observation

Appearance Options:
DATALABELATTRS =text-attrs Appearance attributes of the labels
FILLATTRS =fill-attrs Appearance attributes of bubble interior
LINEATTRS =line-attrs Appearance attributes for bubble outline

Other Options:
BRADIUSMAX =value Radius in pixels of the largest bubble
BRADIUSMIN =value Radius in pixels of the smallest bubble

Boolean Options:
FILL | NOFILL Display filled bubbles or not
OUTLINE | NOOUTLINE Display bubble outline or not

Standard common options are supported

ATTRID, LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, URL,
X2AXIS, Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
106 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.12.1: Basic Bubble Plot

This is a basic bubble plot of Age by Height

and Weight. This graph is for female
students in the sashelp.class data set.

An INSET statement is used to identify the

subset of the data.

title 'Student Age by Height and Weight';

proc sgplot data=sashelp.class; 9.3
where sex='F';
bubble x=height y=weight size=age;
inset 'Females only' / position=bottomright;
xaxis grid; yaxis grid;
run;

Figure 4.12.2: Grouped Bubble Plot

This is a grouped bubble plot of horsepower

by City and Highway Mileage for car
statistics summarized by Type, excluding
Hybrids.

The KEYLEGEND statement is used to

position the legend in the bottom right
corner of the data area.

title 'Vehicle Profile by Type';

proc sgplot data=sgbook.carMeans; 9.3
where _type_ = 1 and type <> 'Hybrid';
bubble x=mean_mpg_city y=mean_mpg_highway size=mean_horsepower /
group=type transparency=0.5;
keylegend / location=inside position=bottomright across=1;
xaxis grid; yaxis grid;
run;
 Chapter 4 Basic Plots 107

Figure 4.12.3: Grouped Bubble Plot

This graph is the same as in Figure 4.12.2

with each bubble labeled by the GROUP
variable.

The NOAUTOLEGEND option is used to

remove the automatic legend, since direct
labeling is preferred to having a separate
legend.

title 'Vehicle Profile by Type';

proc sgplot data=sgbook.carMeans noautolegend; 9.3
where _type_ = 1 and type <> 'Hybrid';
bubble x=mean_mpg_city y=mean_mpg_highway size=mean_horsepower /
group=type transparency=0.5 datalabel=type;
xaxis grid; yaxis grid;
run;

Figure 4.12.4: Bubble Plot with Negative

Response Data

This Bubble plot represents negative size

values using a dash pattern. This is
achieved by using the absolute values for
the size role. The GROUP variable is set
based on the sign of the data and used with
the group role.

Bubble radius min and max are set to map

the data proportionately. A scatter plot with
MARKERCHAR is used for the labels.

title 'Negative Bubbles';

proc sgplot data=sgbook.bubble2; 9.3
bubble x=x y=y size=AbsValue / group=group transparency=0.2
bradiusmin=15 bradiusmax=30 nofill name='bubble';
scatter x=x y=y / markerchar=value;
xaxis grid integer; yaxis grid integer;
keylegend 'bubble' / location=inside position=topleft across=1;
run;
108 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

4.13 HighLow Plot (9.3)

The HIGHLOW plot statement has the following syntax:
highlow x=column low=num-column high=num-column </opts>;
highlow y=column low=num-column high=num-column </opts>;

This is a versatile plot type that displays a vertical or horizontal high-low

plot, using either a “bar” or “line” visual. The line version is useful to
create traditional stock plots. The bar version is useful for event
timelines.

Required Data Roles:

X or Y column Position along the vertical or horizontal axis
HIGH num-column End point of the bar or line segment
LOW num-column Start point of the bar or line segment

Optional Data Roles:

CLOSE =num-column Close value for bar or line
GROUP =column Classification for multiple series
HIGHCAP | LOWCAP =column | keyword Caps (NONE, OPENARROW, etc.)
HIGHLABEL | LOWLABEL =column | string Label for the upper or lower end
OPEN =num-column Open value for bar or line
URL =column URL for each observation

Appearance Options:
FILLATTRS =fill-attrs Appearance of bar interior
DATALABELATTRS =text-attrs Appearance of bar labels
LINEATTRS =line-attrs Appearance of line or outline

Other Options:
BARWIDTH =value Width of bar as fraction of tick spacing
GROUPDISPLAY =value Cluster or Overlay
GROUPORDER =keyword DATA | ASCENDING | DESCENDING
INTERVALBARWIDTH =value Width of bar in pixels
TYPE =keyword BAR | LINE

Boolean Options:
FILL | NOFILL Display filled bars or not for Type=Bar
OUTLINE | NOOUTLINE Display outline or not for Type=Bar

Standard common options are supported

ATTRID, CLUSTERWIDTH, DISCRETEOFFSET, GROUPDISPLAY, GROUPORDER,
LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, X2AXIS, Y2AXIS

Related Style Elements: Same as for the VBarParm Plot statement.

 Chapter 4 Basic Plots 109

Figure 4.13.1: HIGHLOW Line Plot

This graph uses the HighLow plot to create

a Stock Plot of prices by month.

The LOW and HIGH roles are used to

display the open and close values. The
default TYPE is line.

title 'Stock Plot';

proc sgplot data=sgbook.stock; 9.3
highlow x=date low=open high=close / lineattrs=(thickness=2);
xaxis grid display=(nolabel);
yaxis grid label='Price';
run;

Figure 4.13.2: HighLow Line Plot with

Open and Close Values

This graph uses the HighLow plot to create

a Stock Plot of prices by month.

The LOW and HIGH roles are used to

display the monthly low and high values.
The OPEN and CLOSE roles are used to
display the open and close values. The
default TYPE is line.

title 'Stock Plot';

proc sgplot data=sgbook.stock; 9.3
highlow x=date low=low high=high / open=open close=close
lineattrs=(thickness=2);
xaxis grid display=(nolabel);
yaxis grid display=(nolabel);
run;
110 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.13.3: HighLowLine Plot with End

Caps

This is a basic Stock Plot showing the open

and close stock prices by Month. The trend
is shown using the end caps.

Note: if the line segment is not long enough,

end caps are not shown.

Lower is the smaller of Open and Close.

Higher is the greater of Open and Close.

title 'Stock Plot';

proc sgplot data=sgbook.stockcap; 9.3
highlow x=date low=lower high=higher / lineattrs=(thickness=2)
highcap=highcap lowcap=lowcap;
xaxis grid display=(nolabel);
yaxis grid display=(nolabel);
run;

Figure 4.13.4: HighLow Bar Chart

This is a HighLow plot of mileage range by

type. TYPE=Bar, and BarWidth = 0.5.

LOWLABEL and HIGHLABEL options are

used to display the low and high labels from
specific columns.

title 'Mileage Range by Type';

proc sgplot data=sgbook.carMeans; 9.3
where _type_ = 1;
highlow x=type low=min_mpg_City high=max_mpg_city / type=bar
barwidth=0.5 lowlabel=min_mpg_city highlabel=max_mpg_city;
xaxis grid display=(nolabel);
yaxis grid;
run;
 Chapter 4 Basic Plots 111

Figure 4.13.5: HighLow Bar Chart

This is an Adverse Event Timeline for a

patient in a drug study. Data comes in
CDISC format. Events are plotted by
aeseq. The LOWLABEL role is used to
display the aedecod names. The y-axis is
discrete.

For two of the events the aeendate is

missing, so the bar segment is not drawn.
This is indicated by the annotated arrows.

title 'Adverse Event Timeline for Patient Id xxxxxxxx';

proc sgplot data=sgbook.AETimeline tmplout='highlow.sas'; 9.3
highlow y=aeseq low=aestdate high=aeendate / type=bar barwidth=0.6
lowlabel=aedecod;
xaxis grid display=(nolabel);
yaxis grid type=discrete display=(noticks novalues nolabel);
run;

Figure 4.13.6: HighLow Bar Chart

This graph is same as in Figure 4.13.5.

For the two events with missing aeendate,

we have inserted the study end date for
them, and marked the obs to display a
HighCap, These events are shown with a
high cap arrow, indicating events continuing
past the study end date.

The bar width includes the cap.

title 'Adverse Event Timeline for Patient Id xxxxxxxx';

proc sgplot data=sgbook.AETimelinecap tmplout='highlow.sas'; 9.3
highlow y=aeseq low=aestdate high=aeendate / type=bar barwidth=0.8
lowlabel=aedecod highcap=highcap;
xaxis grid display=(nolabel);
yaxis grid display=(noticks novalues nolabel);
run;
112 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

4.14 Reference Lines

X and Y reference lines have the following syntax:
refline value <value> </ options>;

One or more vertical or horizontal reference lines can be added to a

graph using the above syntax. Reflines can be aligned to any one of the
four axes X, Y, X2 or Y2. The GraphReference style element is used.

Required Data Roles:

Value-list Position along the x-axis or y-axis

Appearance Options:
LINEATTRS =line-attrs Appearance attributes for the reference lines

Other Options:
AXIS = axis The axis X, X2, Y, Y2 for the values
LABEL = string Labels for the reference line(s)
LABELLOC = string Location of the label (Inside | Outside)
LABELPOS = string Position of the label (Start | End)

Boolean Options:
NOCLIP Refline is included to determine axis range

Common options supported

DISCRETEOFFSET, LEGENDLABEL, NAME, TRANSPARENCY

Related Style Elements: The GraphReference style element is used to draw these lines.
 Chapter 4 Basic Plots 113

Figure 4.14.1: Scatter with Reference Line

In this example, we have added a Y

reference line with default settings. The
associated label is displayed outside the
graph, on the right side.

title 'Vehicle Profile';

proc sgplot data=sgbook.cars2;
where type='Sedan' and (origin ='USA' or origin='Asia');
scatter x=hp y=mpg / transparency=0.5;
refline 24 / label='24';
xaxis grid; yaxis grid;
run;

Figure 4.14.2: Scatter with Reference Line

In this graph the scatter plot response is

plotted on the Y2 axis (right side). By
default, the reference line label would have
been shown on the right side, thus colliding
with the Y2 axis.

Here we have set LABELPOS=MIN, to place

the label on the left side.

title 'Vehicle Profile';

proc sgplot data=sgbook.cars2 ;
where type='Sedan' and (origin ='USA' or origin='Asia');
scatter x=hp y=mpg / group=origin transparency=0.5
markerattrs=(symbol=circlefilled) y2axis;
keylegend / location=inside position=topright across=1;
refline 25 / axis=y2 label='25' labelpos=min;
xaxis grid; yaxis grid;
run;
114 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.14.3: REFLINE on Discrete Axis

This is a bar chart of mean MPG (City) for

all cars by Type. We want to place a
vertical reference line between ‘Sports’ and
‘Truck’ on the x-axis.

We do this by providing the category value,

AXIS=X and DISCRETEOFFSET=-0.5.
This places the reference line midway
between the specified and previous
midpoint value.

title 'Vehicle Profile';

proc sgplot data=sgbook.carMeans ; 9.3
where _type_ = 1;
vbarparm category=type response=mean_mpg_city / dataskin=sheen;
refline 'Truck' / axis=x discreteoffset=-0.5;
yaxis grid;
run;

Figure 4.14.4: Multiple Reference Lines

In this example, we have added multiple Y

reference lines with inside labels positioned
at the lower side of the axis. We have set
XAXIS OFFSETMIN to create some room.

We have added an X reference line at

x=425. Since this value is beyond the data
range, this refline would normally get
clipped. To retain it, we have specified the
NoClip option.

title 'Vehicle Profile';

proc sgplot data=sgbook.cars2;
where type='Sedan' and (origin ='USA' or origin='Asia';
scatter x=hp y=mpg / transparency=0.5;
refline 18 24 / label=('18' '24') labelloc=inside labelpos=min;
refline 425 / axis=x label='425' labelloc=inside noclip;
xaxis offsetmin=0.1 grid; yaxis grid;
run;
 Chapter 4 Basic Plots 115

4.15 Parametric Line Plot (9.3)

This parametric line plot statements has the following syntax:
lineparm x=column y=column slope=num-column </ opts>;

The LINEPARM statement draws an infinitely long line given a point

and slope. The plot cannot be used by itself as it is unbounded.
Graph data extents are set by other data driven plots in the graph.

Required Data Roles:

X column X coordinate of point
Y column Y coordinate of point
SLOPE column Slope of the line. Missing implies a vertical line.

Optional Data Roles:

CURVELABEL =column Label for the line
GROUP =column Classification for multiple series
INDEX =num-column Index for visual attributes

Appearance Options:
CURVELABELATTRS =text-attrs Appearance of the label
CURVELABELLOC =keyword OUTSIDE | INSIDE
CURVELABELPOS =keyword AUTO | MAX | MIN
LINEATTRS =line-attrs Appearance of the line

Other Options:
CLIP =boolean Include data for this plot in the axis range
LEGENDLABEL =string Label that appears in legend
NAME =string Name for this statement

Boolean Options:
NOEXTEND Line is not drawn to the axes
X2AXIS Assigns X values to the X2 (top) axis
Y2AXIS Assigns Y values to the Y2 (right) axis

Standard common options are supported

ATTRID, LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, X2AXIS,
Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
116 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.15.1: Parametric Line

This graph shows city vs. highway mileage

for all cars with a reference line of slope=1
to indicate same performance between city
and Highway.
o
The 45 sloped reference line is drawn using
the LINEPARM statement. A label for the
line is displayed outside the data area using
the CURVELABEL and CURVELABELLOC
options.

title 'City vs Highway Mileage';

proc sgplot data=sgbook.cars2; 9.3
where type='Sedan';
scatter x=mpgc y=mpgh / transparency=0.5 markerattrs=(symbol=circlefilled);
lineparm x=0 y=0 slope=1.0 / curvelabel='Slope=1'
curvelabelloc=outside lineattrs=(pattern=dash);
xaxis grid;
yaxis grid; run;
Figure 4.15.2: Parametric Lines

In this graph, the data for the two reference

lines comes from the data set.

Three new columns “X”, “Y” and “Slope” are

merged into the data set. These columns
are provided to the LINEPARM statement.
The CURVELABEL and CURVELABELPOS
options are used to display the labels.

title 'Mean Vehicle Performance';

proc sgplot data=sgbook.carMeansLine; 9.3
where _type_=1 and type<>'Truck';
bubble x=mean_horsepower y=mean_mpg_city size=_freq_/
transparency=0.5 datalabel=type;
lineparm x=x y=y slope=slope / curvelabel lineattrs=(pattern=dash)
curvelabelpos=min;
xaxis grid offsetmin=0.15;
yaxis grid; run;
 Chapter 4 Basic Plots 117

4.16 Waterfall Chart (9.3)

The WATERFALL statement has the following syntax:
waterfall category=column response=column </ options>;

The waterfall chart draws bar segments positioned relative to the

cumulative sum of the preceding observations. The response value for
each observation is expected to be a delta that is accumulated along the
axis. The initial value for the accumulation can be specified. The final bar value is computed from
the data.

Required Data Roles:

CATEGORY column Position along the horizontal axis
RESPONSE num-column Delta response along the vertical axis

Optional Data Roles:

COLORGROUP =column Classification for bar color
DATALABEL =column Label string for each observation
URL =column URL for each observation

Appearance Options:
DATALABELATTRS =text-attrs Appearance attributes of the labels
FILLATTRS =fill-attrs Appearance attributes of bubble interior
FINALBARATTRS =line-attrs Appearance attributes for bubble outline
INITIALBARATTRS =line-attrs Appearance attributes for bubble outline

Other Options:
BARWIDTH =value Radius in pixels of the largest bubble
DATASKIN =skin-value One of the predefined skin types
FINALBARTICKVALUE =string String for labeling final bar on axis
INITIALBARTICKVALUE =string String for labeling initial bar on axis
INITIALBARVALUE =value Response value for initial bar
STAT =keyword MEAN | SUM

Boolean Options:
FILL | NOFILL Display filled bar or not
OUTLINE | NOOUTLINE Display bar outline or not
MISSING Show missing category values

Standard common options are supported

ATTRID, LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, URL, X2AXIS,
Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a group variable is present, the GraphData1-GraphData12 elements are used, one for
each group.
118 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.16.1: Basic Waterfall Chart

This graph shows the cumulative cash

balance over transactions. Each
observation is drawn relative to the previous
cumulative statistic from zero, which is SUM
in this case.

The final bar value is computed from the

data and displayed at the end.

title "Auto Dealership Balance for March";

proc sgplot data=sgbook.transactions; 9.3
waterfall category=transaction response=amount;
yaxis display=(nolabel);
run;

Figure 4.16.2: Waterfall Chart with Initial

Value

This graph shows the same data as the one

shown in Figure 4.16.1 with the addition of
setting the initial bar value. The cumulative
statistic is computed starting with the initial
value, which is 350 in this case.

The y-axis OFFSETMIN is set to zero and

gridlines are enabled.

title "Auto Dealership Balance for March";

proc sgplot data=sgbook.transactions; 9.3
waterfall category=transaction response=amount
initialbarvalue=350;
yaxis display=(nolabel) grid offsetmin=0;
run;
 Chapter 4 Basic Plots 119

Figure 4.16.3: Grouped Waterfall Chart

This graph shows the same data as the one

shown in Figure 4.16.2 with the addition of a
color group role and data skin. Here each
bar is colored by the transaction type,
Income or Expense.

A legend is displayed in the upper left

corner of the graph.

title "Auto Dealership Balance for March";

proc sgplot data=sgbook.transactions; 9.3
waterfall category=transaction response=amount / colorgroup=Type
initialbarvalue=350 initialbarattrs=(color=white)
dataskin=pressed finalbarattrs=(color=cx2f2f2f);
keylegend / location=inside position=topleft across=1;
yaxis display=(nolabel) grid offsetmin=0;
run;

Figure 4.16.4: Waterfall Chart with Data

Labels

This graph shows the same data as the one

shown in Figure 4.16.3 with the addition of
data labels.

Long tick labels for initial and final bars is

set, and XAXIS FITPOLICY is set to
stagger.

title "Auto Dealership Balance for March";

proc sgplot data=sgbook.transactions; 9.3
waterfall category=transaction response=amount /
initialbartickvalue="Starting balance"
finalbartickvalue="Ending balance" datalabel
initialbarattrs=(color=white) finalbarattrs=(color=black)
colorgroup=Type dataskin=pressed initialbarvalue=350;
keylegend / location=inside position=bottom;
xaxis fitpolicy=stagger;
yaxis display=(nolabel) offsetmin=0 grid; run;
120 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

4.17 Combining the Plots

Basic plots can be combined (overlaid) in unlimited ways as long as the data types on each axis
are consistent. In some of the preceding examples, we have seen various combinations of
scatter, series, step, and band plots. These plot types tend to have numeric interval variables on
both axes, but sometimes may have categorical variables on one or both axes.

The order of the plot statements and the axes determine if certain combinations are allowed. The
axis type (linear, discrete, etc.) is decided by the first plot in the list. Subsequent plots must have
compatible type of data on the same axis for the combination to be acceptable.

Compatibility:
1. If the axis is of type LINEAR, it is compatible with only LINEAR variables.
2. If the axis is of type DISCRETE, then it is compatible with both DISCRETE and LINEAR
variables. Subsequent LINEAR data are treated as DISCRETE.

Axis and Data Compatibility:

1. If the first plot is a SCATTER, with x-axis having linear variables, then the default axis
type is LINEAR. Now, subsequent overlaid plots must have linear variables for the
combination to work. A plot with an x-axis that has a categorical variable will not be
accepted, unless the axis type is explicitly set as DISCRETE.
2. If the first plot has a DISCRETE variable on the y-axis, then the subsequent overlaid plot
can have a Y variable that is either linear or discrete. The linear data will be treated as
DISCRETE.

Common Combinations of Basic Plots:

1. Any plot with one or more plots of the same kind – Multi-Response
2. Scatter with Series and Step
3. Series and Step plots with Band plots
4. Scatter, Series, and Bands
5. Scatter, Series, Step, Band, and BarParm with Reference Lines
6. Scatter, Series, Step, Band, and BarParm with LineParm Lines
7. Bar-Line using VBarParm, and Series
8. Scatter with MarkerChar - for overlaid annotated data labels
9. Scatter and Vector Plots
10. HighLow Type=Line with Series and Bands – Stock Plot
11. HighLow Type=Bar with Scatter – Adverse Event Plot

Note: Basic plot statements can also be combined with:

• Fit and Confidence plots – We will see these examples in Chapter 5.
• Distribution plots – We will see these examples in Chapter 6.
• Categorization plots – We will see these examples in Chapter 7.

Let us look at some examples in the following pages.

 Chapter 4 Basic Plots 121

Figure 4.17.1: Scatter-Scatter Overlay

This is a combination of two scatter plots,

one each for Min and Max Mileage (City) by
mean Horsepower.

A legend is automatically generated, and we

have positioned it in the upper right of the
data area by using the KEYLEGEND
statement.

title 'Mileage Range by Mean Horsepower';

proc sgplot data=sgbook.carMeans;
where _type_=1;
scatter x=mean_horsepower y=min_mpg_city;
scatter x=mean_horsepower y=max_mpg_city;
keylegend / location=inside position=topright across=1;
xaxis grid;
yaxis grid label='MPG';
run;

Figure 4.17.2: Scatter-Series Overlay

This is an overlay of the raw data with a

linear fit plot. The prediction curve has been
computed previously using the REG
procedure.

The standard GRAPHFIT style element is

applied to the fit line.

title 'Class Health Statistics';

proc sgplot data=sashelp.classfit;
scatter x=height y=weight;
series x=height y=predict / lineattrs=GraphFit;
keylegend / location=inside position=bottomright across=1;
xaxis grid;
yaxis grid;
run;
122 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.17.3: Scatter, Series, and Band

This is a standard fit plot with confidence limits.

The prediction curve and the limits have been
computed previously using the REG procedure.

The VALUEATTRS option on the KEYLEGEND

statement is used to create smaller fonts.

title 'Class Health Statistics';

9.3
proc sgplot data=sashelp.classfit;
band x=height upper=upper lower=lower / legendlabel='Confidence' nofill;
band x=height upper=uppermean lower=lowermean / legendlabel='Prediction';
scatter x=height y=weight;
series x=height y=predict / lineattrs=GraphFit legendlabel='Predicted Val';
keylegend / location=inside position=bottomright across=1
valueattrs=(size=6);
xaxis grid;
yaxis grid;
run;

Figure 4.17.4: Step and Scatter Plot

Here is a survival plot by time and treatment. The

statistics for the two treatments are in separate
columns. This allows us to use the scatter plot to
display the at-risk numbers.

A hidden REFLINE is used to label the at-risk

numbers, using its “label” feature, while
suppressing the line itself. Variable PlotD is used
to suppress plotting of count.

title 'Survival by Time and Treatment';

proc sgplot data=sgbook.survival; 9.3
step x=daysD y=survivalD / markers markerattrs=GraphData1(size=4)
name='D' legendlabel='Drug';
step x=daysP y=survivalP / markers markerattrs=GraphData2(size=4)
name='P' legendlabel='Placebo';
scatter x=daysD y=ylabelD / markerchar=leftD markercharattrs=(size=3) freq=plotD;
scatter x=daysP y=ylabelP / markerchar=leftP markercharattrs=(size=3) freq=plotP;
refline -0.1 -0.2 / label=('Drug' 'Placebo') labelloc=outside labelpos=min
lineattrs=(thickness=0);
keylegend 'D' 'P' / location=inside position=topright across=1 valueattrs=(size=5);
xaxis grid label='Days'; yaxis grid values=(0 to 1 by .2) valueshint;
run;
 Chapter 4 Basic Plots 123

Figure 4.17.5: Series and Band

This is a Body Mass Index graph, showing

the regions that are classified as “Normal”,
“Underweight”, etc. using a band plot.

The actual data values for each subject are

plotted over time using a SERIES statement
with markers and data labels. The
CURVELABEL and CURVELABELPOS
options are used to display curve labels.

title 'Three Patient Growth Charts vs BMI';

proc sgplot data=sgbook.bmi;
band x=weight_bmi upper=upper_hgt lower=lower_hgt /
transparency=.7 group=category name="b";
series x=wgt y=hgt / datalabel=year lineattrs=(thickness=1) markers
group=name name="pts" curvelabel curvelabelpos=start;
keylegend "b" / position=bottomright location=inside across=1 title='BMI';
xaxis min=30 max=300 label='Weight (lbs)' grid;
yaxis min=40 max=80 label='Height (in)' grid;
run;

Figure 4.17.6: Scatter and Band

Here is an overlay of two band plots and a

needle plot showing the sales figures over
the year, with lower and upper limits.

The upper limits of each band are labeled,

thus avoiding the need for a legend.

title 'Sales with Limits';

proc sgplot data=sgbook.sales noautolegend;
band x=date upper=high lower=low / fillattrs=graphdata1
fill outline curvelabelupper='High';
band x=date upper=low lower=0 / fillattrs=graphdata5
fill outline curvelabelupper='Low';
needle x=date y=sales / lineattrs=(thickness=2) transparency=0.5;
xaxis grid display=(nolabel);
yaxis grid;
run;
124 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.17.7: Grouped Band

This is an example of a Theme River plot

created using a grouped band plot. The
data for each band covers the full height of
the band. So, we have used
TRANSPARENCY to ensure all bands are
visible.

title 'Theme River Group';

proc sgplot data=sgbook.themeriverGroup;
band x=time upper=upper lower=lower / group=name transparency=0.7
fill outline lineattrs=(pattern=solid);
keylegend / position=bottomright location=inside across=1
valueattrs=(size=4);
xaxis display=(nolabel) grid;
yaxis grid label='Relative Frequency' display=(novalues noticks);
run;

Figure 4.17.8: Band Overlay

This graph is similar to Figure 4.17.7. Here

the data for each band is in separate
columns. So, we have used a four band
overlay to create this graph.

Since we know the data extents of each

band, we can specify the sequence for
drawing, ensuring the narrowest band is
drawn last. The TRANSPARENCY option is
not used in this example.

title 'Theme River Overlay';

proc sgplot data=sgbook.themeriverMulti;
band x=time upper=Janeu lower=Janel / fill outline lineattrs=(pattern=solid);
band x=time upper=BillU lower=BillL / fill outline lineattrs=(pattern=solid);
band x=time upper=MaryU lower=MaryL / fill outline lineattrs=(pattern=solid);
band x=time upper=JohnU lower=JohnL / fill outline lineattrs=(pattern=solid);
keylegend / position=bottomright location=inside across=1 valueattrs=(size=4);
xaxis display=(nolabel) grid;
yaxis grid label='Relative Frequency' display=(novalues noticks);
run;
 Chapter 4 Basic Plots 125

Figure 4.17.9: Pareto Chart

Here is a Pareto Chart, using VBARPARM

and SERIES statements.

BARWIDTH=0.8 and DISCRETEOFFSET=

0.4 are set to ensure the cumulative line is
located at the top right corner of the bars.

Markers are displayed for the series plot,

with data labels for each observation.

title 'Product Sales Pareto Chart';

proc sgplot data=sgbook.pareto; 9.3
vbarparm category=product response=share / barwidth=0.8;
series x=product y=cumulative / markers discreteoffset=0.4
markerattrs=(symbol=circlefilled size=3) datalabel=cumulative;
keylegend / position=topleft location=inside across=1;
xaxis display=(nolabel) grid;
yaxis grid label='Relative Frequency' display=(nolabel);
run;

Figure 4.17.10: Floating Bars and Bands

This is a graph of the monthly average

temperatures for a year in the future (2109),
compared with 30-year average numbers.

The 30-year averages are potted using a

HighLow plot with TYPE=Bar. The high and
low data values are displayed. Values for
2109 are shown using a band plot.

VALUEATTRS is used for small fonts.

title 'Monthly Temperatures for 2109';

proc sgplot data=sgbook.weatherLine; 9.3
highlow x=month high=high low=low / highlabel=high lowlabel=low
type=bar legendlabel='30 Year Average' transparency=0.5;
band x=month upper=thisyearH lower=thisyearL / fill outline
transparency=0.5 fillattrs=graphdata3 legendlabel='This Year';
keylegend / position=bottom location=inside;
xaxis display=(nolabel) grid valueattrs=(size=7);
yaxis grid label='Temperature (F)' valueattrs=(size=7);
run;
126 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 4.17.11: VBARPARM with

Statistics

This is a graph showing product sales with

low and high forecasts. Here we have
displayed some statistics that are not
included in the graph.

Axis offsets are used to create space for the

statistics. We have drawn the stats using
scatter plots with MARKERCHAR
associated with the Y2 axis. The space for
the Y2 data is restricted to the top 15%.
title 'Product Sales with Limits';
proc sgplot data=sgbook.productStat noautolegend; 9.3
vbarparm category=product response=sales / dataskin=gloss
limitupper=high limitlower=low;
scatter x=product y=labelS / markerchar=sales y2axis;
scatter x=product y=labelV / markerchar=vol y2axis;
scatter x=product y=labelM / markerchar=margin y2axis;
xaxis display=(nolabel) grid;
yaxis grid display=(nolabel) offsetmax=0.2 values=(0 to 20000 by 5000) valueshint;
y2axis offsetmin=0.85 display=(nolabel noticks); run;

Figure 4.17.12: HBARPARM with

Statistics

This graph is similar to Figure 4.17.11, using

an HBarParm with limits, and scatter plots to
display the statistics.

Axis offsets are used to create space for the

statistics. We have drawn the stats using
scatter plots with MARKERCHAR
associated with the X2 axis. The space for
the X2 data is restricted to the right 16%.
title 'Product Sales with Limits';
proc sgplot data=sgbook.productStat noautolegend; 9.3
hbarparm category=product response=sales / dataskin=gloss
limitupper=high limitlower=low;
scatter y=product x=labelV / markerchar=vol x2axis;
scatter y=product x=labelM / markerchar=margin x2axis;
xaxis grid display=(nolabel) offsetmin=0 offsetmax=0.20
values=(0 to 20000 by 5000) valueshint;
x2axis offsetmin=0.84 offsetmax=0.05 display=(noticks nolabel) valueattrs=(size=7);
yaxis display=(nolabel); run;
Chapter 5
Fit and Confidence Plots
5.1 Introduction 129
5.2 Fit Plot Roles and Options 130
5.3 Regression Plot 131
5.4 Loess Plot 138
5.5 Penalized B-Spline Plot 145
5.6 Ellipse Plot 152
5.7 Combining the Plots 157
128 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“Do, or do not. There is no 'try'.”

~ Yoda
 Chapter 5 Fit and Confidence Plots 129

Chapter 5: Fit and Confidence Plots

In this chapter, we examine the fit and confidence plots for SGPLOT and SGPANEL procedures.
The basic syntax for each plot and related options will be covered in detail in this chapter. Plots
within this group can be mixed and matched with each other and with many X-Y plots from the
“Basic Plots” group.

5.1 Introduction
The SGPLOT and SGPANEL procedures support multiple fit and confidence plots as listed
below. These plots normally also display the raw data as a scatter plot, and also support display
of confidence bands. These plots can be combined with many basic X-Y plots, most commonly
with the scatter plot.

REG LOESS
PBSPLINE ELLIPSE
130 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

5.2 Fit Plot Roles and Options

All of the fit plot statements have the following syntax:
FitPlotName required-data-roles </ options>;

All plot statements have required and optional parameters necessary to create the visual. These
parameters fall into the following broad categories:

1. required data roles

2. optional data roles
3. options
4. common options

Here is a list of common options for fits:

Common Options:
ATTRID =string The attribute map ID used for the group
variable.
CURVELABELATTRS =line-attrs Appearance attributes for the curve label.
CURVELABELLOC =keyword Curve label location - INSIDE | OUTSIDE
CURVELABELPOS =keyword Curve label position - AUTO | MAX | MIN
DATALABEL <=column> Display a scatter label for the Y variable or
from an optional variable.
DATALABELATTRS =text-attrs Appearance attributes for DataLabel.
GROUP =column Classification variable for fit calculations.
LEGENDLABEL =string The label that appears in the legend to
represent this (non-group) plot.
LINEATTRS = line-attrs Appearance attributes for the fit line.
MARKERATTRS = marker-attrs Appearance attributes for the markers.
MAXPOINTS =positive-integer Maximum number of prediction points.
NAME =string Specifies a name for this statement. Other
statements, such as a KEYLEGEND, can
refer to a plot by its name.
TRANSPARENCY =value Specifies the transparency for the visual
elements.

Common Boolean Options:

NOMARKERS Do not show the scatter markers with the fit line.
X2AXIS Assigns X values to the X2 (top) axis.
Y2AXIS Assigns Y values to the Y2 (right) axis.

These common options are available for each plot statement, and can be used exactly as
described above. To avoid duplication, we will not list the above common options for the
individual plots that are discussed in this chapter.
 Chapter 5 Fit and Confidence Plots 131

5.3 Regression Plot

The Regression plot shows a linear or nonlinear fit of the raw data.
The syntax is:

reg x=column y=column </ options>;

A marker is displayed at each (x, y) location. A fit line is computed

and displayed. Optionally, different confidence limits can also be
displayed.

Required Data Roles:

X =num-column Position along the horizontal axis
Y =num-column Position along the vertical axis

Optional Data Roles:

FREQ =num-column Frequency of the current observation
WEIGHT =num-column Priority weights for a regression fit

Appearance Options:
ALPHA =positive-number Confidence level
CLIATTRS =cli-attrs Appearance attributes for CLI
CLMATTRS =clm-attrs Appearance attributes for CLM
DEGREE =positive-integer The degree of the polynomial

Other Options:
CLI =string Display the confidence limits for individual predicted
values using the string for the legend label.
CLM =string Display confidence limits for mean predicted values
using the string for the legend label.

Boolean Options:
CLI Display confidence limits for individual predicted values.
CLM Display confidence limits for mean predicted values.
NOLEGCLI Exclude the CLI from the legend.
NOLEGCLM Exclude the CLM from the legend.
NOLEGFIT Exclude the fit line from the legend.

Standard common options are supported by fits

ATTRID, CURVELABEL, CURVELABELLOC, CURVELABELPOS, DATALABEL,
DATALABELATTRS, GROUP, LEGENDLABEL, LINEATTRS, MARKERATTRS,
MAXPOINTS, NAME, NOMARKERS, TRANSPARENCY, X2AXIS, Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
132 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 5.3.1: Basic Linear Fit

This plot contains a basic linear regression

fit. By default, the REG statement produces
both the fit line and the scatter points. This
behavior can be controlled by the
NOMARKERS option.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan' and origin='USA';
reg x=horsepower y=mpg_city;
xaxis grid;
yaxis grid;
run;

Figure 5.3.2: CLM Band

In addition to the linear fit and scatter points,

this plot displays a computed confidence
limit for the mean (CLM).

By default, the alpha value for the

computation is .05 (95% confidence). This
value can be adjusted using the ALPHA
option.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan' and origin='USA';
reg x=horsepower y=mpg_city / clm;
xaxis grid;
yaxis grid;
run;
 Chapter 5 Fit and Confidence Plots 133

Figure 5.3.3: CLI Band

In addition to the linear fit, scatter points,

and CLM band, this plot displays the
prediction limits for individual values (CLI).

By default, the alpha value for the

computation is .05 (95% confidence). This
value can be adjusted using the ALPHA
option, which affects both CLM and CLI.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan' and origin='USA';
reg x=horsepower y=mpg_city / clm cli;
xaxis grid;
yaxis grid;
run;

Figure 5.3.4: Automatic Legend

This plot displays a legend containing

information about the fit. This legend
appears by default and can be disabled by
using the NOAUTOLEGEND option or by
specifying a custom legend.

The NOLEGFIT option prevents the fit line

information from being displayed in the
legend.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars;
where type='Sedan' and origin='USA';
reg x=horsepower y=mpg_city / clm cli nolegfit;
xaxis grid;
yaxis grid;
run;
134 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 5.3.5: Axis Assignment

In this plot, the regression fit is assigned to

both the X and Y secondary axes.

Notice that the grid line must now be

specified on the X2AXIS and Y2AXIS
statements.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan' and origin='USA';
reg x=horsepower y=mpg_city / clm cli x2axis y2axis;
x2axis grid;
y2axis grid;
run;

Figure 5.3.6: Attribute Control

Many of the visual attributes of the REG

statement can be controlled with options.

In this example, the CLI line is set to a

“solid” pattern and the scatter marker
shapes are set to “plus”.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend; 9.3
where type='Sedan' and origin='USA';
reg x=horsepower y=mpg_city / clm cli markerattrs=(symbol=plus)
cliattrs=(clilineattrs=(pattern=solid)) ;
xaxis grid;
yaxis grid;
run;
 Chapter 5 Fit and Confidence Plots 135

Figure 5.3.7: Scatter Data Labels

By default, the DATALABEL option displays

the non-missing Y values of the scatter
points. You can assign a column to the
option to display alternate values.

In this example, the Y values of only the

outlying points were copied to a column
called “outlier”. The missing values in that
column are not displayed.

title "The Impact of Horsepower on MPG";

proc sgplot data=sgbook.carlabel noautolegend;
where type='Sedan' and origin='USA';
reg x=horsepower y=mpg_city / clm cli datalabel=outlier;
xaxis grid;
yaxis grid;
run;

Figure 5.3.8: Polynomial Regression Fit

The DEGREE option can be used to change

the polynomial for the regression fit.

In this example, a quadratic fit is made

through the scatter points by setting the
DEGREE option to 2.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan' and origin='USA';
reg x=horsepower y=mpg_city / degree=2;
xaxis grid;
yaxis grid;
run;
136 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 5.3.9: Grouped Fits

When a GROUP column is specified on a

REG statement, both the scatter points and
the fit lines are grouped.

Two separate statements are used to

control the transparency of the markers.
The NOMARKER option on the REG
statement is used to disable the default
scatter points.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars;
where type='Sedan';
scatter x=horsepower y=mpg_city / group=origin transparency=0.85
name="scat";
reg x=horsepower y=mpg_city / group=origin nomarkers name="reg";
keylegend "scat" / title="Origin";
keylegend "reg" / location=inside across=1;
xaxis grid;
yaxis grid; run;

Figure 5.3.10: Single Fit with a Group

To create a single fit through all of your

grouped scatter data, use a SCATTER
statement to plot the grouped data and a
REG statement without a group column and
the NOMARKER option.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars;
where type='Sedan';
scatter x=horsepower y=mpg_city / group=origin transparency=0.85
name="scat";
reg x=horsepower y=mpg_city / nomarkers degree=3 name="reg";
keylegend "scat" / location=inside across=1 title="Origin";
xaxis grid;
yaxis grid;
run;
 Chapter 5 Fit and Confidence Plots 137

Figure 5.3.11: Overlaid Fits

You can overlay fits for comparison by using

the NOMARKER option on all REG
statements but one, or you can use a
separate SCATTER statement for more
control over the scatter plot.

Notice that the REG statements are

assigned to a custom legend. The
LEGENDLABEL option overrides the default
legend label for each fit.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars;
where type='Sedan';
scatter x=horsepower y=mpg_city / transparency=0.85;
reg x=horsepower y=mpg_city / nomarkers degree=2 name="reg1"
legendlabel="degree=2";
reg x=horsepower y=mpg_city / nomarkers degree=3 name="reg2"
legendlabel="degree=3";
keylegend "reg1" "reg2" / location=inside across=1;
xaxis grid; yaxis grid; run;

Figure 5.3.12: Curve Labels

Instead of using a legend to identify each fit,

you can use the CURVELABEL option to
directly label the fit lines. This approach
works well when the lines are not too
crowded.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan';
scatter x=horsepower y=mpg_city / transparency=0.85;
reg x=horsepower y=mpg_city / nomarkers degree=2 curvelabel="degree=2";
reg x=horsepower y=mpg_city / nomarkers degree=3 curvelabel="degree=3";
xaxis grid;
yaxis grid;
run;
138 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

5.4 Loess Plot

The Loess plot shows a smoothed fit of the raw data with a Loess fit.
The syntax is:

loess x=column y=column </ options>;

A marker is displayed at each (x, y) location. A fit line is computed

and displayed. Optionally, confidence limits of the means can be displayed.

Required Data Roles:

X =num-column Position along the horizontal axis.
Y =num-column Position along the vertical axis.

Optional Data Roles:

WEIGHT =num-column Priority weights for a regression fit.

Appearance Options:
ALPHA =positive-number Confidence level to compute.
CLMATTRS =clm-attrs Appearance attributes for CLM.
DEGREE =positive-integer The degree of the polynomial.
INTERPOLATION =keyword The degree of the interpolating polynomials
used for blending local polynomial fits at the
kd tree vertices – LINEAR | CUBIC.
REWEIGHT =postive-integer The number of iterative reweighting steps.
SMOOTH =postive-integer Specifies a smoothing parameter value.

Other Options:
CLM =string Display confidence limits for mean predicted values
using the string for the legend label.

Boolean Options:
CLM Display confidence limits for mean predicted values.
NOLEGCLM Exclude the CLM from the legend.
NOLEGFIT Exclude the fit line from the legend.

Standard common options are supported by fits

ATTRID, CURVELABEL, CURVELABELLOC, CURVELABELPOS, DATALABEL,
DATALABELATTRS, GROUP, LEGENDLABEL, LINEATTRS, MARKERATTRS,
MAXPOINTS, NAME, NOMARKERS, TRANSPARENCY, X2AXIS, Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
 Chapter 5 Fit and Confidence Plots 139

Figure 5.4.1: Basic Loess Fit

This plot contains a basic loess fit. By

default, the LOESS statement produces
both the fit line and the scatter points. This
behavior can be controlled by the
NOMARKERS option.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan' and origin='USA';
loess x=horsepower y=mpg_city;
xaxis grid;
yaxis grid;
run;

Figure 5.4.2: Fit Plot with Smooth Curves

The SMOOTH option enables you to specify

a smoothing parameter for the loess fit.

Compare the default fit of this data in Figure

5.4.1 with this output.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan' and origin='USA';
loess x=horsepower y=mpg_city / smooth=0.4;
xaxis grid;
yaxis grid;
run;
140 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 5.4.3: CLM Band

In addition to the linear fit and scatter points,

this plot displays the confidence limit of the
mean (CLM).

By default, the alpha value for the

computation is .05 (95% confidence). This
value can be adjusted using the ALPHA
option.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan' and origin='USA';
loess x=horsepower y=mpg_city / smooth=0.4 clm;
xaxis grid;
yaxis grid;
run;

Figure 5.4.4: Automatic Legend

This plot displays a legend containing

information about the fit. This legend
appears by default and can be disabled by
using the NOAUTOLEGEND option or by
specifying a custom legend.

The NOLEGFIT option prevents the fit line

information from being displayed in the
legend.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars;
where type='Sedan' and origin='USA';
loess x=horsepower y=mpg_city / smooth=0.4 clm nolegfit;
xaxis grid;
yaxis grid;
run;
 Chapter 5 Fit and Confidence Plots 141

Figure 5.4.5: Axis Assignment

In this plot, the loess fit is assigned to both

the X and Y secondary axes.

Notice that the grid line must now be

specified on the X2AXIS and Y2AXIS
statements.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan' and origin='USA';
loess x=horsepower y=mpg_city / smooth=0.4 clm x2axis y2axis;
x2axis grid;
y2axis grid;
run;

Figure 5.4.6: Attribute Control

Many of the visual attributes of the LOESS

statement can be controlled with options.

In this example, the CLM band is set to a

“#dddddd” gray color, and the scatter
marker shapes are set to “plus”.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend; 9.3
where type='Sedan' and origin='USA';
loess x=horsepower y=mpg_city / clm markerattrs=(symbol=plus)
clmattrs=(clmfillattrs=(color=graydd)) smooth=0.4;
xaxis grid;
yaxis grid;
run;
142 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 5.4.7: Scatter Data Labels

By default, the DATALABEL option displays

the non-missing Y values of the scatter
points. You can assign a column to the
option to display alternate values.

In this example, the Y values of the min/max

points were copied to a column called
“minmax”. The missing values in that
column are not displayed.

title "The Impact of Horsepower on MPG";

proc sgplot data=sgbook.carlabel2 noautolegend;
where type='Sedan' and origin='USA';
loess x=horsepower y=mpg_city / clm smooth=0.4 datalabel=minmax;
xaxis grid;
yaxis grid;
run;

Figure 5.4.8: Cubic Loess Fit

The INTERPOLATION option can be used

to change the degree of the interpolating
polynomials. In this example, a cubic fit is
made through the scatter points by setting
the INTERPOLATION option to CUBIC.

Compare this output to the default LINEAR

interpolation output in Figure 5.4.1.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan' and origin='USA';
loess x=horsepower y=mpg_city / interpolation=cubic;
xaxis grid;
yaxis grid;
run;
 Chapter 5 Fit and Confidence Plots 143

Figure 5.4.9: Grouped Fits

When a GROUP column is specified on a

LOESS statement, both the scatter points
and the fit lines are grouped.

Two separate statements are used to

control the transparency of the markers.
The NOMARKERS option on the LOESS
statement is used to disable the default
scatter points.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars;
where type='Sedan' and drivetrain ne "All";
scatter x=horsepower y=mpg_city / group=DriveTrain transparency=0.8;
loess x=horsepower y=mpg_city / group=DriveTrain smooth=0.4
nomarkers name="loess";
keylegend "loess" / location=inside across=1 title="Drivetrain";
xaxis grid;
yaxis grid;
run;

Figure 5.4.10: Single Fit with a Group

To create a single fit through all of your

grouped scatter data, use a SCATTER
statement to plot the grouped data and a
LOESS statement without a group column
and the NOMARKERS option.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars;
where type='Sedan' and drivetrain ne "All";
scatter x=horsepower y=mpg_city / group=DriveTrain
transparency=0.8 name="scat";
loess x=horsepower y=mpg_city / smooth=0.4 nomarkers;
keylegend "scat" / location=inside across=1;
xaxis grid;
yaxis grid; run;
144 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 5.4.11: Overlaid Fits

You can overlay fits for comparison by using

the NOMARKERS option on all LOESS
statements but one, or you can use a
separate SCATTER statement for more
control over the scatter plot.

Notice that the LOESS statements are

assigned to a custom legend. The default
legend label reflects the fit type used and
the smoothing parameter specified.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars;
where horsepower <= 300;
scatter x=horsepower y=mpg_city / transparency=0.8;
loess x=horsepower y=mpg_city / nomarkers smooth=0.8 name="loess1";
loess x=horsepower y=mpg_city / nomarkers smooth=0.3 name="loess2";
keylegend "loess1" "loess2" / location=inside across=1;
xaxis grid;
yaxis grid; run;

Figure 5.4.12: Curve Labels

Instead of using a legend to identify each fit,

you can use the CURVELABEL option to
directly label the fit lines. This approach
works well when the lines are not too
crowded.

Here the labels are moved to the MIN side

of the fits and outside of the data area. The
plots are assigned to the secondary y-axis
to give space to the fit labels on the left.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars noautolegend;
where horsepower <= 300;
scatter x=horsepower y=mpg_city / transparency=0.8 y2axis;
loess x=horsepower y=mpg_city / nomarkers smooth=0.8
curvelabel="Smooth=0.8" curvelabelpos=min curvelabelloc=outside y2axis;
loess x=horsepower y=mpg_city / nomarkers smooth=0.3
curvelabel="Smooth=0.3" curvelabelpos=min curvelabelloc=outside y2axis;
xaxis grid;
y2axis grid; run;
 Chapter 5 Fit and Confidence Plots 145

5.5 Penalized B-Spline Plot

The penalized B-spline plot shows a smoothed fit of the raw data
created with a penalized B-spline:

pbspline x=column y=column </ options>;

A marker is displayed at each (x, y) location. A fit line is computed and displayed. Optionally,
confidence limits and/or confidence limits of the means can be displayed.

Required Data Roles:

X =num-column Position along the horizontal axis
Y =num-column Position along the vertical axis

Optional Data Roles:

FREQ =num-column Frequency of the current observation
WEIGHT =num-column Priority weights for a regression fit

Appearance Options:
ALPHA =positive-number Confidence level to compute
CLIATTRS =cli-attrs Appearance attributes for CLI
CLMATTRS =clm-attrs Appearance attributes for CLM
DEGREE =positive-integer The degree of the polynomial
NKNOTS =positive-number Number of evenly-spaced internal knots

Other Options:
CLI =string Display confidence limits for individual predicted
values using the string for the legend label.
CLM =string Display confidence limits for mean predicted values
using the string for the legend label.
CLMTRANSPARENCY =value Transparency for the CLM band.

Boolean Options:
CLI Display confidence limits for individual predicted values.
CLM Display confidence limits for mean predicted values.
NOLEGCLI, NOLEGCLM Exclude the CLI or CLM from the legend.
NOLEGFIT Exclude the fit line from the legend.

Standard common options are supported by fits

ATTRID, CURVELABEL, CURVELABELLOC, CURVELABELPOS, DATALABEL,
DATALABELATTRS, GROUP, LEGENDLABEL, LINEATTRS, MARKERATTRS,
MAXPOINTS, NAME, NOMARKERS, X2AXIS, Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple statements are overlaid, the GraphData1-
GraphData12 elements are used, one for each group or statement.
146 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 5.5.1: Basic Parametric B-Spline

Fit

This plot contains a basic penalized B-

spline fit. By default, the PBSPLINE
statement produces both the fit line and the
scatter points. This behavior can be
controlled by the NOMARKERS option.

title "Heart Study Cholesterol Trend";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
pbspline x=weight y=cholesterol;
xaxis grid;
yaxis grid;
run;

Figure 5.5.2: CLM Band

In addition to the PBSpline fit and scatter

points, this plot displays a computed
confidence limit of the mean (CLM).

By default, the alpha value for the

computation is .05 (95% confidence). This
value can be adjusted using the ALPHA
option.

title "Heart Study Cholesterol Trend";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
pbspline x=weight y=cholesterol / clm;
xaxis grid;
yaxis grid;
run;
 Chapter 5 Fit and Confidence Plots 147

Figure 5.5.3: CLI Band

In addition to the PBSpline fit, scatter points,

and CLM band, this plot displays the
confidence limits (CLI) for individual
predicted values for each observation.

By default, the alpha value for the

computation is .05 (95% confidence). This
value can be adjusted using the ALPHA
option, which affects both CLM and CLI.

title "Heart Study Cholesterol Trend";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
pbspline x=weight y=cholesterol / clm cli;
xaxis grid;
yaxis grid;
run;

Figure 5.5.4: Automatic Legend

This plot displays a legend containing

information about the fit. This legend
appears by default and can be disabled by
using the NOAUTOLEGEND option or by
specifying a custom legend.

The NOLEGFIT option prevents the fit line

information from being displayed in the
legend.

title "Heart Study Cholesterol Trend";

proc sgplot data=sashelp.heart;
where ageatstart <= 30;
pbspline x=weight y=cholesterol / clm cli nolegfit;
xaxis grid;
yaxis grid;
run;
148 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 5.5.5: Axis Assignment

In this plot, the PBSPLINE fit is assigned to

both the X and Y secondary axes.

Notice that the grid lines must now be

specified on the X2AXIS and Y2AXIS
statements.

title "Heart Study Cholesterol Trend";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
pbspline x=weight y=cholesterol / clm cli x2axis y2axis;
x2axis grid;
y2axis grid;
run;

Figure 5.5.6: Attribute Control

Many of the visual attributes of the

PBSPLINE statement can be controlled with
options.

In this example, the CLI line is set to a

“solid” pattern and the scatter marker
shapes are set to “plus”.

title "Heart Study Cholesterol Trend";

proc sgplot data=sashelp.heart noautolegend; 9.3
where ageatstart <= 30;
pbspline x=weight y=cholesterol / clm cli
cliattrs=(clilineattrs=(pattern=solid))
markerattrs=(symbol=plus);
xaxis grid;
yaxis grid;
run;
 Chapter 5 Fit and Confidence Plots 149

Figure 5.5.7: Scatter Data Labels

By default, the DATALABEL option displays

the non-missing Y values of the scatter
points. You can assign a column to the
option to display alternate values.

In this example, the Y values of the outlying

points were copied to a column called
“outlier”. The missing values in that column
are not displayed.

title "Heart Study Cholesterol Trend";

proc sgplot data=sgbook.heartLabel noautolegend;
where ageatstart <= 30;
pbspline x=weight y=cholesterol / clm cli datalabel=outlier;
xaxis grid;
yaxis grid;
run;

Figure 5.5.8: The DEGREE Option

The DEGREE option can be used to change

the degree of the B-spline.

In this example, the DEGREE option is set

to 1. Compare the result with the default
result in Figure 5.5.1, where the degree is 3.

title "Heart Study Cholesterol Trend";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
pbspline x=weight y=cholesterol / degree=1;
xaxis grid;
yaxis grid;
run;
150 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 5.5.9: Grouped Fits

When a GROUP column is specified on a

PBSPLINE statement, both the scatter
points and the fit lines are grouped.

Two separate statements are used to

control the transparency of the markers.
The NOMARKER option on the PBSPLINE
statement is used to disable the default
scatter points.

title "Heart Study Cholesterol Trend";

proc sgplot data=sashelp.heart;
where ageatstart <= 30;
scatter x=weight y=cholesterol / group=sex transparency=0.5 name="scat";
pbspline x=weight y=cholesterol / group=sex nomarkers name="fit";
keylegend "fit";
keylegend "scat" / location=inside;
xaxis grid;
yaxis grid;
run;

Figure 5.5.10: Single Fit with a Group

To create a single fit through all of your

grouped scatter data, use a SCATTER
statement to plot the grouped data and a
PBSPLINE statement without a group
column and the NOMARKER option.

title "Heart Study Cholesterol Trend";

proc sgplot data=sashelp.heart;
where ageatstart <= 30;
scatter x=weight y=cholesterol / group=sex transparency=0.5 name="scat";
pbspline x=weight y=cholesterol / nomarkers;
keylegend "scat";
xaxis grid;
yaxis grid;
run;
 Chapter 5 Fit and Confidence Plots 151

Figure 5.5.11: Overlaid Fits

You can overlay fits for comparison by using

the NOMARKERS option on all PBSPLINE
statements but one, or you can use a
separate SCATTER statement for more
control over the scatter plot.

Notice that the PBSPLINE statements are

assigned to a custom legend. The
LEGENDLABEL option overrides the default
legend label for each fit.

title "Heart Study Cholesterol Trend";

proc sgplot data=sashelp.heart;
where ageatstart <= 30;
scatter x=weight y=cholesterol / transparency=0.5;
pbspline x=weight y=cholesterol / nomarkers degree=3 name="three"
legendlabel="Degree=3";
pbspline x=weight y=cholesterol / nomarkers degree=1 name="one"
legendlabel="Degree=1";
keylegend "three" "one";
xaxis grid;
yaxis grid;
run;

Figure 5.5.12: Curve Labels

Instead of using a legend to identify each fit,

you can use the CURVELABEL option to
directly label the fit lines.

This approach works well when the lines are

not too crowded.

title "Heart Study Cholesterol Trend";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
scatter x=weight y=cholesterol / transparency=0.5;
pbspline x=weight y=cholesterol / nomarkers degree=3 curvelabel="Degree=3";
pbspline x=weight y=cholesterol / nomarkers degree=1 curvelabel="Degree=1";
xaxis grid;
yaxis grid;
run;
152 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

5.6 Ellipse Plot

The Ellipse plot shows a confidence ellipse of the raw data. The
syntax is:

ellipse x=column y=column </ options>;

A marker is displayed at each (x, y) location. An ellipse is computed

and displayed.

Required Data Roles:

X =num-column Position along the horizontal axis
Y =num-column Position along the vertical axis

Optional Data Roles:

FREQ =num-column Frequency of the current observation

Appearance Options:
ALPHA =positive-number Confidence level to compute
FILLATTRS =fill-attrs Attributes for the ellipse fill
LINEATTRS =line-attrs Attributes for the ellipse outline
TYPE =keyword Type of the confidence ellipse
PREDICTED | MEAN

Boolean Options:
CLIP Exclude the ellipse data from the axis data range calculation
FILL / NOFILL Enable/Disable display of a fill color in the ellipse
OUTLINE/NOOUTLINE Enable/Disable display of an outline around the ellipse

Standard common options are supported by fits

LEGENDLABEL, NAME, TRANSPARENCY, X2AXIS, Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
 Chapter 5 Fit and Confidence Plots 153

Figure 5.6.1: Basic Ellipse

By default, the ELLIPSE statement

produces a prediction ellipse with an ALPHA
value of .05.

title "Heart Study Cholesterol Analysis";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
ellipse x=weight y=cholesterol;
scatter x=weight y=cholesterol;
xaxis grid;
yaxis grid;
run;

Figure 5.6.2: The Filled Ellipse

When the FILL option is specified, the

ellipse is filled with color and the outline is
not displayed.

The outline can be re-enabled by specifying

the OUTLINE option.

title "Heart Study Cholesterol Analysis";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
ellipse x=weight y=cholesterol / fill;
scatter x=weight y=cholesterol;
xaxis grid;
yaxis grid;
run;
154 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 5.6.3: Ellipse Clipping

The CLIP option prevents the ellipse data

from contributing to the axis range, causing
the ellipse to clip if it extends beyond the
other data.

title "Heart Study Cholesterol Analysis";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
ellipse x=weight y=cholesterol / clip;
scatter x=weight y=cholesterol;
xaxis grid;
yaxis grid;
run;

Figure 5.6.4: Ellipse with Confidence

Level

The ALPHA option sets the confidence level

for the computation. By default, the alpha
value is .05.

Compare this output with Figure 5.6.1 to see

the effect of the alpha change.

title "Heart Study Cholesterol Analysis";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
ellipse x=weight y=cholesterol / alpha=.35;
scatter x=weight y=cholesterol;
xaxis grid;
yaxis grid;
run;
 Chapter 5 Fit and Confidence Plots 155

Figure 5.6.5: Ellipse Attributes

When the fill or outline features are visible,

the attributes can be controlled from the
ELLIPSE statement.

title "Heart Study Cholesterol Analysis";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
ellipse x=weight y=cholesterol / fill outline lineattrs=(pattern=dash)
fillattrs=(color=cxeeeeee);
scatter x=weight y=cholesterol;
xaxis grid;
yaxis grid;
run;

Figure 5.6.6: Ellipse Type

By default, the ellipse type is PREDICTED,

which creates a prediction ellipse for a new
observation.

In this example, the TYPE is set to MEAN,

which specifies a confidence ellipse of the
population mean.

title "Heart Study Cholesterol Analysis";

proc sgplot data=sashelp.heart noautolegend;
where ageatstart <= 30;
ellipse x=weight y=cholesterol / type=mean fill alpha=.001;
scatter x=weight y=cholesterol;
xaxis grid;
yaxis grid;
run;
156 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 5.6.7: Overlaid Ellipses

In this example, three ellipses with different

alpha values are overlaid and associated
with a legend. The legend reflects the
information about the ellipses.

In addition, the cholesterol values for the

outlying points are copied to a column called
“outlier” and displayed using the
DATALABEL option.

title "Heart Study Cholesterol Analysis";

proc sgplot data=sgbook.heartlabel2 noautolegend;
where ageatstart <= 30;
ellipse x=weight y=cholesterol / alpha=.05 name="e1";
ellipse x=weight y=cholesterol / alpha=.25 name="e2";
ellipse x=weight y=cholesterol / alpha=.50 name="e3";
scatter x=weight y=cholesterol / datalabel=outlier;
keylegend "e1" "e2" "e3";
xaxis grid;
yaxis grid; run;

Figure 5.6.8: Overlaid Filled Ellipses

This example is similar to Figure 5.6.7,

except that the ellipses are filled, creating a
backdrop for the scatterplot.

The ellipse representation in the legend also

changes with the presence of the FILL
option.

Ellipses are ordered in the program so that

the smaller ellipses are drawn on top.

title "Heart Study Cholesterol Analysis";

proc sgplot data= sgbook.heartlabel2 noautolegend;
where ageatstart <= 30;
ellipse x=weight y=cholesterol / fill alpha=.05 name="e1";
ellipse x=weight y=cholesterol / fill alpha=.25 name="e2";
ellipse x=weight y=cholesterol / fill alpha=.50 name="e3";
scatter x=weight y=cholesterol / datalabel=outlier;
keylegend "e1" "e2" "e3";
xaxis grid;
yaxis grid; run;
 Chapter 5 Fit and Confidence Plots 157

5.7 Combining the Plots

Figure 5.7.1: Comparing Fits

Different types of fits with various

parameters can be overlaid on the raw plot
data for comparison.

title "The Impact of Horsepower on MPG";

proc sgplot data=sashelp.cars;
where type='Sedan' and origin='USA';
reg x=horsepower y=mpg_city;
loess x=horsepower y=mpg_city / nomarkers smooth=0.4;
xaxis grid;
yaxis grid; run;

Figure 5.7.2: Comparing Fits with CLM

To prevent the CLM bands from overlapping

the fit lines, the statements are ordered to
draw the CLM bands first, then the fit lines.

The first two plot statements suppress the fit

lines by specifying a zero-thickness line
attribute. The legend is assigned to the
statements that draw the fit lines.

proc sgplot data=sashelp.cars;

where type='Sedan' and origin='USA';
reg x=horsepower y=mpg_city / clm nomarkers lineattrs=(thickness=0);
loess x=horsepower y=mpg_city / clm nomarkers smooth=0.4
lineattrs=(thickness=0);
reg x=horsepower y=mpg_city / nomarkers name="reg";
loess x=horsepower y=mpg_city / name="loess" smooth=0.4
lineattrs=(pattern=solid);
keylegend "reg" "loess";
xaxis grid; yaxis grid; run;
158
Chapter 6
Distribution Plots
6.1 Introduction 161
6.2 Histogram 162
6.3 Density Plot 166
6.4 Vertical Box Plot 169
6.5 Horizontal Box Plot 174
6.6 Combining the Plots 177
160 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“Information is not knowledge.”

~ Albert Einstein
 Chapter 6 Distribution Plots 161

Chapter 6: Distribution Plots

In this chapter, we examine the distribution plots that are supported by the SGPLOT and
SGPANEL procedures. The basic syntax for each plot and the options will be covered in detail in
this chapter. Plots within this group can be mixed and matched with each other and with many X-
Y plots from the basic plots group.

6.1 Introduction
The SGPLOT and SGPANEL procedures support multiple distribution plots as listed below.
These plots normally do not display the raw data. These plots can be combined only in specific
ways with other plots in this category or the basic plots.

HISTOGRAM DENSITY
VBOX HBOX
162 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

6.2 Histogram
The histogram shows the distribution of a numeric variable. The
syntax is:

histogram column </ options>;

The observations are binned into N number of bins determined by the

statement. The counts in the bin are shown as percentage by default.
Alternatively, frequency counts can be displayed.

Required Data Roles:

X num-column Position along the horizontal axis

Optional Data Roles:

FREQ =num-column Frequency count for each observation

Appearance Options:
FILLATTRS =fill-attrs Appearance attributes for bins

Other Options:
BINSTART = value X coordinate of the first bin.
BINWIDTH =value Width of the bin
BOUNDARY =keyword Location of boundary values – UPPER |
LOWER
NBINS =value Number of bins
SCALE =keyword PERCENT | COUNT | PROPORTION for Y axis
scaling

Boolean Options:
FILL | NOFILL Whether or not the bins are filled
OUTLINE | NOOUTLINE Whether or not the bins have outlines
SHOWBINS Tickmark on the X axis is shown at the midpoint of the bin

Standard common options are supported

LEGENDLABEL, NAME, TRANSPARENCY, X2AXIS, Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
 Chapter 6 Distribution Plots 163

Figure 6.2.1: Basic Histogram

This is a basic histogram of horsepower for

only the sedans in the data set.

Default scale for y-axis is percent. The x-

axis label is suppressed.

By default, the x-axis is linear.

title 'Distribution of Horsepower';

proc sgplot data=sashelp.cars;
where type='Sedan';
histogram horsepower;
yaxis grid;
xaxis display=(nolabel);
run;

Figure 6.2.2: Histogram with Frequency

This is a histogram of horsepower for the

cars. The FREQ role is used to apply the
frequency for each observation. The data
set itself only has observations where
type=’Sedan’.

In this graph the y-axis is “Count”. The x-

axis label is suppressed.

By default, the x-axis is linear.

title 'Distribution of Horsepower';

proc sgplot data=sgbook.carsFreq;
histogram horsepower / freq=freq scale=count;
yaxis grid;
xaxis display=(nolabel);
run;
164 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 6.2.3: Histogram with NBINS

For this histogram, the NBINS=25. This is

a hint for the plot; the actual bins may be
different.

Default scale for y-axis is percent. The x-

axis label is suppressed. Transparency is
set to 40%.

By default, the x-axis is linear.

title 'Distribution of Horsepower';

proc sgplot data=sashelp.cars; 9.3
histogram horsepower / nbins=25 transparency=0.4;
yaxis grid;
xaxis display=(nolabel);
run;

Figure 6.2.4: Histogram with BINSTART

and BINWIDTH

For this histogram the BINSTART=0 and

the BINWIDTH=20. Bins are created
starting from zero with a width of 20 until
the max data value is crossed.

Default scale for y-axis is percent. The x-

axis label is suppressed. Transparency is
set to 40%.

title 'Distribution of Horsepower';

proc sgplot data=sashelp.cars; 9.3
histogram horsepower / binstart=0 binwidth=20 transparency=0.4;
yaxis grid;
xaxis display=(nolabel);
run;
 Chapter 6 Distribution Plots 165

Figure 6.2.5: Multiple Histograms

This graph shows multiple histograms, one

for city and one for highway mileage.

For one of the plots we have set NOFILL to

distinguish the two, and see through to the
first. Color and transparency can also be
used to distinguish the two histograms.

A default legend is automatically created.

title 'Distribution of Mileage';

proc sgplot data=sashelp.cars;
histogram mpg_City / transparency=0.5;
histogram mpg_highway / nofill;
yaxis grid;
xaxis display=(nolabel);
run;

Figure 6.2.6: Multiple Histograms

This graph shows multiple histograms, one

each for diastolic and systolic blood
pressure. BINSTART and BINWIDTH are
set on both histograms for similar binning.

Reference lines show the normal ranges.

Transparency is used to distinguish
between the histograms.

title 'Distribution of Blood Pressure';

proc sgplot data=sashelp.heart; 9.3
histogram diastolic / binstart=40 binwidth=10 transparency=0.7;
histogram systolic / binstart=40 binwidth=10 transparency=0.5;
refline 80 100 120 140 / axis=X label=('80' '100' '120' '140')
labelpos=max lineattrs=(pattern=dash);
keylegend / location=inside position=topright noborder;
xaxis display=(nolabel);
yaxis grid; run;
166 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

6.3 Density Plot

The density plot shows the fitted distribution of a numeric variable.
The syntax is:

density column </ options>;

The distribution of the observations is shown as a curve. The plot

supports normal density and a kernel density estimate.

Required Data Roles:

num-column Position along the horizontal axis

Optional Data Roles:

FREQ =num-column Frequency count for each observation

Appearance Options:
LINEATTRS =line-attrs Appearance attributes for line

Other Options:
TYPE =keyword (opts) NORMAL (options) | KENNEL (options)
SCALE =keyword PERCENT | COUNT | PROPORTION for y-axis
scaling

Type=Normal Options:
MU =value Mean value to be used
SIGMA =value Standard deviation to be used

Type=Kernel Options:
C =value Band Width
WEIGHT =keyword NORMAL | QUADRATIC | TRIANGULAR

Standard common options are supported

LEGENDLABEL, NAME, TRANSPARENCY, X2AXIS, Y2AXIS

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
 Chapter 6 Distribution Plots 167

Figure 6.3.1: Normal Density Curve

This is a graph of the normal density

distribution for horsepower for all cars in the
data set.

title 'Normal Density for Horsepower';

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan';
density horsepower;
yaxis grid;
xaxis display=(nolabel);
run;

Figure 6.3.2: Density Curve Options

This graph is of the same data as in Figure

6.3.1, with SCALE=Percent, and using a
dash line pattern.

title 'Normal Density for Horsepower';

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan';
density horsepower / scale=percent lineattrs=(pattern=dash);
yaxis grid;
xaxis display=(nolabel);
run;
168 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 6.3.3: Normal Density Curve

The normal density function can be

customized by specifying the Mu and Sigma
parameters.

In this graph SIGMA=30, as shown in the

inset.

ods escapechar '^';

title 'Normal Density for Horsepower';
proc sgplot data=sashelp.cars noautolegend;
density horsepower / scale=percent type=Normal(sigma=30);
inset "^{unicode SIGMA} = 30" / position=topright;
yaxis grid;
xaxis display=(nolabel);
run;

Figure 6.3.4: Kernel Density Estimate

This graph shows the kernel density

estimate for horsepower.

TYPE=Kernel supports options for C and

weight.

title 'Kernel Density for Horsepower';

proc sgplot data=sashelp.cars noautolegend;
density horsepower / scale=percent type=Kernel;
yaxis grid;
xaxis display=(nolabel);
run;
 Chapter 6 Distribution Plots 169

6.4 Vertical Box Plot

The vertical box (VBOX) plot shows the distribution of the analysis
variable by category. The syntax is:

vbox num-column </ options>;

A box is displayed for each unique category value. The default display
includes the Q1-Q3 interval, mean, median, whiskers, and outliers.

Required Data Roles:

num-column Analysis variable.

Optional Data Roles:

CATEGORY =column For category values along x-axis
FREQ =num-column Frequency count for each observation
GROUP =column Group variable for each observation

Appearance Options:
CONNECTATTRS =line-attrs Appearance for connect line
DATALABELATTRS =text-attrs Appearance for data labels
FILLATTRS =fill-attrs Appearance of box interior
LINEATTRS =line-attrs Appearance of box outlines
MEANATTRS OUTLIERATTRS =marker-attrs Appearance of mean and outliers
MEDIANATTRS =line-attrs Appearance of median
WHISKERATTRS =line-attrs Appearance of whiskers

Other Options:
BOXWIDTH =value Width of the box as a fraction of spacing
CAPSHAPE =keyword SERIF | LINE | BRACKET
CONNECT =keyword The statistic to connect – MEAN, MEDIAN, etc.

Boolean Options:
DATALABEL Label the outliers
EXTREME Whiskers are drawn to the extreme values
FILL | NOFILL Specifies whether the boxes are filled or not
LABELFAR Display labels for far outliers
MISSING NOCAPS NOMEAN Various options
NOMEDIAN NOOUTLIERS NOTCHES Various options

Standard common options are supported

CLUSTERWIDTH, DISCRETEOFFSET, GROUPDISPLAY, GROUPORDER,
LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, X2AXIS, Y2AXIS

Related Style Elements: Same as for Scatter, Series, VBar, and HBar.
170 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 6.4.1: Basic Box Plot

This is a basic box plot of mileage by type.

Default display includes the Q1-Q3 interval,
mean marker, median line, whiskers, and
outliers.

title 'Distribution of Mileage by Type';

proc sgplot data=sashelp.cars noautolegend;
where type='Sedan' or type='Sports' or type='SUV';
vbox mpg_city / category=type;
yaxis grid;
xaxis display=(nolabel);
run;

Figure 6.4.2: Grouped Box Plot

This is a grouped box plot of mileage by

type and drivetrain. For consistency with
other graphs like scatter, series, etc., the
default GROUPDISPLAY is Overlay.

Set GROUPDISPLAY=Cluster to get side-

by-side display of group values.

A legend is displayed by default.

title 'Distribution of Mileage by Type';

proc sgplot data=sashelp.cars; 9.3
where type='Sedan' or type='Sports' or type='SUV';
vbox mpg_city / category=type group=drivetrain
groupdisplay=cluster;
yaxis grid;
xaxis display=(nolabel);
run;
 Chapter 6 Distribution Plots 171

Figure 6.4.3: Grouped Unfilled Box Plot

This is a grouped box plot, where we have

set the line patterns for the box to solid, and
set NOFILL.

The legend displays the line patterns for the

whiskers to distinguish the boxes.

title 'Distribution of Mileage by Type';

proc sgplot data=sashelp.cars; 9.3
where type='Sedan' or type='Sports' or type='SUV';
vbox mpg_city / category=type group=drivetrain groupdisplay=cluster
nofill lineattrs=(pattern=solid);
keylegend / location=inside position=topright across=1;
yaxis grid;
xaxis display=(nolabel);
run;

Figure 6.4.4: Box Plot Overlay

This graph shows the distribution of mileage

for city and highway as an overlay of two
separate variables. The two box plots use
DISCRETEOFFSET and BOXWIDTH to
display the boxes side by side.

With SAS 9.3, box plots with common

category axis can be overlaid. This is a
common use case in clinical trials for
multiple treatments.

title 'Distribution of City and Highway Mileage';

proc sgplot data=sashelp.cars; 9.3
where type='Sedan' or type='Sports' or type='SUV';
vbox mpg_city / category=type discreteoffset=-0.12 boxwidth=0.2;
vbox mpg_highway / category=type boxwidth=0.2 discreteoffset= 0.12;
keylegend / location=inside position=topright across=1;
yaxis grid label='Mileage';
xaxis display=(nolabel);
run;
172 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 6.4.5: Box Plot on Linear Axis

This is a graph of a VBOX where the x-axis

data is set to linear (SAS 9.3) and so is not
categorical. Format is MONNAME3.

This is a common use case in the health

and life sciences industry, where the data is
distributed unevenly on the time axis.

XAXIS TYPE=linear. The box spacing is

determined by the smallest interval in data.

title 'Lab Results over Time';

proc sgplot data=sgbook.intervalBoxMulti; 9.3
vbox DrugA / category=date;
xaxis grid type=linear display=(nolabel) tickvalueformat=monname3.
values=('01Jan2009'd to '01Dec2009'd by month);
yaxis grid display=(nolabel);
run;

Figure 6.4.6: Box Overlay for Linear Data

This is an overlay of two VBOX plots on a

date axis with uneven values.

Here we want to view the lab results over

time and by treatment type. A VBOX
overlay must have the same X variable, so
we have to jitter the date values
appropriately. See the code for generating
this data set.

XAXIS TYPE=linear.

title 'Lab Results over Time and Treatment';

9.3
proc sgplot data=sgbook.intervalBoxMulti;
vbox DrugA / category=date2 meanattrs=(size=5) outlierattrs=(size=5);
vbox DrugB / category=date2 meanattrs=(size=5) outlierattrs=(size=5);
xaxis grid type=linear display=(nolabel) offsetmin=0.05 offsetmax=0.05
values=('01Jan2009'd to '01Dec2009'd by month)
tickvalueformat=monname3.;
yaxis grid display=(nolabel);
run;
 Chapter 6 Distribution Plots 173

Figure 6.4.7: Box Plot with Labels

This graph shows a VBOX plot of mileage

by vehicle type, with CONNECT=Mean.

DATALABEL option displays the data labels

for the outliers.

title 'Mileage by Type';

proc sgplot data=sashelp.cars;
vbox mpg_city / category=type connect=mean datalabel;
xaxis grid display=(nolabel);
yaxis grid display=(nolabel);
run;

Figure 6.4.8: Box Plot with Notches

This is a graph showing boxes with notches

using the NOTCHES option.

OFFSETMAX=0.2 is set to provide space

for the notches in this case.

title 'Mileage by Type';

proc sgplot data=sashelp.cars;
vbox mpg_city / category=type notches;
xaxis grid;
yaxis grid display=(nolabel) offsetmax=0.2;
run;
174 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

6.5 Horizontal Box Plot

The horizontal box (HBOX) plot shows the distribution of the analysis
variable by category. The syntax is:

hbox column </ options>;

A box is displayed for each unique category value. The default

display includes the Q1-Q3 interval, mean, median, whiskers, and
outliers.

Required Data Roles:

num-column Analysis variable

Optional Data Roles:

CATEGORY =column For category values along x-axis
FREQ =num-column Frequency count for each observation
GROUP =column Group variable for each observation

Appearance Options:
CONNECTATTRS =line-attrs Appearance for connect line
DATALABELATTRS =text-attrs Appearance for data labels
FILLATTRS =attrs Appearance of box interior
LINEATTRS =attrs Appearance of box outlines
MEANATTRS OUTLIERATTRS =marker-attrs Appearance of mean and outliers
MEDIANATTRS =line-attrs Appearance of median
WHISKERATTRS =line-attrs Appearance of whiskers

Other Options:
BOXWIDTH =value Width of the box as a fraction of spacing
CAPSHAPE =keyword SERIF | LINE | BRACKET
CONNECT =keyword The statistic to connect – MEAN, MEDIAN, etc.

Boolean Options:
DATALABEL Label the outliers
EXTREME Whiskers are drawn to the extreme values
FILL | NOFILL Specifies whether the boxes are filled or not
LABELFAR Display labels for far outliers
MISSING NOCAPS NOMEAN Various options
NOMEDIAN NOOUTLIERS NOTCHES Various options

Standard common options are supported

CLUSTERWIDTH, DISCRETEOFFSET, GROUPDISPLAY, GROUPORDER,
LEGENDLABEL, NAME, NOMISSINGGROUP, TRANSPARENCY, X2AXIS, Y2AXIS

Related Style Elements: Same as for VBOX.

 Chapter 6 Distribution Plots 175

Figure 6.5.1: Basic Horizontal Box Plot

This graph uses the basic HBOX to display

the distribution of cholesterol by cause of
death.

A horizontal orientation for the box plot is

particularly useful for larger number of
category values having long strings.

title 'Cholesterol by Cause of Death';

proc sgplot data=sashelp.heart;
hbox cholesterol / category=deathcause;
xaxis grid display=(nolabel);
yaxis grid;
run;

Figure 6.5.2: Horizontal Box Plot with

Notches

This graph uses the HBOX with NOTCHES

to display the distribution of cholesterol by
cause of death.

A horizontal orientation for the box plot is

particularly useful where the category
values have long character strings.

title 'Cholesterol by Cause of Death';

proc sgplot data=sashelp.heart;
hbox cholesterol / category=deathcause notches;
xaxis grid display=(nolabel);
yaxis grid;
run;
176 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 6.5.3: Grouped Horizontal Box

Plot

This graph shows the distribution of

cholesterol by cause of death with
GROUP=Sex.

LINEATTRS option sets the outline for all

boxes to solid.

title 'Cholesterol by Cause of Death and Sex';

proc sgplot data=sashelp.heart; 9.3
where deathcause <> 'Other' and deathcause <> 'Unknown';
hbox cholesterol / category=deathcause group=sex
lineattrs=(pattern=solid);
xaxis grid display=(nolabel);
yaxis grid;
run;
Figure 6.5.4: Grouped Horizontal Box
with CLUSTERWIDTH

This graph shows the distribution of

cholesterol by cause of death with
GROUP=Sex.

LINEATTRS option sets the outline for all

boxes to solid. CLUSTERWIDTH=0.5
brings the group values closer.

title 'Cholesterol by Cause of Death and Sex';

proc sgplot data=sashelp.heart; 9.3
where deathcause <> 'Other' and deathcause <> 'Unknown';
hbox cholesterol / category=deathcause group=sex
lineattrs=(pattern=solid) clusterwidth=0.5;
xaxis grid display=(nolabel);
yaxis grid;
run;
 Chapter 6 Distribution Plots 177

6.6 Combining the Plots

Figure 6.6.1: Histogram with Normal
Density Plot

This graph displays the distribution of the

mileage for all cars, along with the normal
density curve.

title 'Distribution of Mileage';

proc sgplot data=sashelp.cars;
histogram mpg_city;
density mpg_city;
keylegend / location=inside position=topright;
xaxis display=(nolabel);
yaxis grid; run;

Figure 6.6.2: Histogram with Normal and

Kernel Density Plots

This graph displays the distribution of the

mileage for all cars. The normal density
curve and the kernel density estimate are
included.

title 'Distribution of Mileage';

proc sgplot data=sashelp.cars;
histogram mpg_city;
density mpg_city;
density mpg_city / type=kernel;
keylegend / location=inside position=topright across=1;
xaxis display=(nolabel);
yaxis grid; run;
178
Chapter 7
Categorization Plots
7.1 Introduction 181
7.2 Categorization Plot Roles and Common Options 182
7.3 Vertical Bar Charts 183
7.4 Horizontal Bar Charts 191
7.5 Vertical Line Charts 198
7.6 Horizontal Line Charts 205
7.7 Dot Plots 212
7.8 Combining the Plots 218
180 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“There are no facts, only interpretations.”

~ Friedrich Nietzsche
 Chapter 7 Categorization Plots 181

Chapter 7: Categorization Plots

In this chapter, we examine the categorization plots for SGPLOT and SGPANEL procedures. The
unique feature of this class of plots is that the data is summarized by using a specified or default
statistic. The basic syntax for each plot and the options will be covered in detail in this chapter.
Plots within this group can be overlaid only with another plot in this group of the same orientation.

7.1 Introduction

The SGPLOT and SGPANEL procedures support the categorization plots listed below. Each of
these plots support sum, mean, and frequency statistics. In addition, these plots support limit
calculations for mean statistics. The limit statistics include the confidence limit of the mean (CLM),
standard error (STDERR), and standard deviation (STDDEV). The confidence of the CLM statistic
can be adjusted using the ALPHA option, and the number of STDERRs or STDDEVs can be
specified using the NUMSTD option.

These plots cannot be used to draw pre-calculated limits. To draw custom limits on a bar chart,
you should use a VBARPARM or HBARPARM plot (see Figure 4.11.1). For line charts, use a
SERIES plot overlaid with a SCATTER plot (see Figure 4.6.10). For DOT plot, simply use a
scatter plot (see Figure 4.4.7).

VBAR HBAR DOT

VLINE HLINE
182 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

7.2 Categorization Plot Roles and Common Options

All of the categorization plot statements have the following syntax:

CategorizationPlotName required-data-roles </ options>;

All plot statements have required and optional parameters necessary to create the visual. Here is
a list of common options for categorization plots:

Common Options:
ALPHA =value Confidence level of CLM statistic
ATTRID =string The attribute map ID used for the group
variable
CATEGORYORDER =keyword RESPASC | RESPDESC
CLUSTERWIDTH =number Cluster width as a ratio of midpoint spacing
DATALABEL <=column> Display a scatter label for the Y variable or
from an optional variable
DATALABELATTRS =text-attrs Appearance attributes for DataLabel
DISCRETEOFFSET =number Amount to offset all data primitives from the
category midpoints
GROUPORDER =keyword DATA | ASCENDING | DESCENDING
LEGENDLABEL =string The label that appears in the legend to
represent this (non-group) plot
LIMITATTRS =line-attrs Appearance attributes for the limits
LIMITS =keyword BOTH | UPPER | LOWER
LIMITSTAT =keyword CLM|STDDEV | STDERR
NAME =string Specifies a name for this statement. Other
statements, such as a KEYLEGEND, can
refer to a plot by its name.
NUMSTD =positive-integer Number of standard units to compute
STAT =keyword FREQ | MEAN | SUM
TRANSPARENCY =value Specifies the transparency for the visual
elements
URL =string-column Contains URLs for drilldown
WEIGHT =num-column Used to weight observations

Common Boolean Options:

MISSING Show missing category values
NOSTATLABEL Suppress statistic on the axis label
 Chapter 7 Categorization Plots 183

7.3 Vertical Bar Charts

The VBAR plot shows the summarization of the response variable by
category or the frequency count of the category. The syntax is:

vbar column </ options>;

By default, the data is summarized for each unique category value;

however, user-defined formats can be used to create categorical
ranges for summarization. One bar is drawn for each resulting category. If a group variable is
specified, this bar can be subdivided by each group value, or the bar can be drawn as a cluster of
smaller group bars (see the GROUPDISPLAY option).

Required Data Roles:

CATEGORY column Category variable

Optional Data Roles:

FREQ =num-column Frequency count for each observation
GROUP =column Group variable for each observation
RESPONSE =num-column Response values for the chart

Appearance Options:
FILLATTRS =fill-attrs Appearance attributes for Fill

Other Options:
BARWIDTH =value Width of bar a fraction of tick spacing
DATALABELPOS =keyword DATA | TOP | BOTTOM
DATASKIN =keyword Quasi-3-D effect for the bars (GLOSS, etc.)
GROUPDISPLAY =keyword STACK | CLUSTER

Boolean Options:
FILL | NOFILL Controls bar fill
OUTLINE | NOOUTLINE Controls bar outline

Standard common options are supported

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
184 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.3.1: Bar Chart with Fill Attributes

and Data Skins

This basic bar chart has been enhanced by

specifying a particular color for the bar fill. In
addition, the DATASKIN option is used to
give the bars a modern, three-dimensional
appearance while maintaining the visual
accuracy of a two-dimensional bar chart.
See the documentation for the different
skins available.

title "Product Sales";

proc sgplot data=sgbook.product_sales; 9.3
vbar product / response=actual fillattrs=(color=grayee)
dataskin=pressed;
run;

Figure 7.3.2: Bar Chart with No Fill

This example uses the NOFILL option to

create a “minimal ink” appearance that
reproduces well in black and white
publications.

There is also a NOOUTLINE option to

remove the bar outlines when you only want
the fill color.

title "Product Sales";

proc sgplot data=sgbook.product_sales;
vbar product / response=actual nofill;
run;
 Chapter 7 Categorization Plots 185

Figure 7.3.3: Bar Chart with Data Labels

This example displays the bar response

values above the bars by using the
DATALABEL option. This option can be
specified with a variable to display
information other than the bar response
values.

The DATALABELATTRS can be used to

override the default text attributes of the
labels.

title "Product Sales";

proc sgplot data=sgbook.product_sales; 9.3
vbar product / response=actual datalabel
datalabelattrs=(weight=bold style=italic);
run;

Figure 7.3.4: Bar Chart with Reference

Line

Reference lines can be added to the bar

chart by specifying a REFLINE statement.
The order of the statements is significant.

If the REFLINE statement is specified before

the VBAR statement, the line is drawn
behind the bar chart. REFLINE statements
specified after the VBAR statement are
drawn in front of the bar chart as shown.

title "Product Sales";

proc sgplot data=sgbook.product_sales;
vbar product / response=actual;
refline 80;
run;
186 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.3.5: Bar Chart with Confidence

Limits

This example shows a bar chart showing

the confidence limit of the mean (CLM). The
standard error and standard deviation
statistics are also available.

The ALPHA option can be used to specify

the confidence for the CLM calculation. The
default is 0.05. The default legend includes
the limit statistic information.

title "Average Cholesterol by Age Group";

proc sgplot data=sashelp.heart;
format AgeAtStart agefmt.;
yaxis label="Avg. Cholesterol";
vbar AgeAtStart / response=cholesterol stat=mean limitstat=clm
alpha=0.05;
run;

Figure 7.3.6: Bar Chart with an Upper

Limit

You can use the LIMITS option to control

which side of the limits are displayed.

This example uses the STDERR statistic

with a NUMSTD multiplier to display two
standard errors instead of one.

title "Average Cholesterol by Age Group";

proc sgplot data=sashelp.heart;
format AgeAtStart agefmt.;
yaxis label="Avg. Cholesterol";
vbar AgeAtStart / response=cholesterol stat=mean limits=upper
limitstat=stderr numstd=2;
run;
 Chapter 7 Categorization Plots 187

Figure 7.3.7: Bar Chart with Limits and

Label Positioning

This example shows the display of the bar

and limits data.

The position of the data labels can be

controlled by using the DATALABELPOS
option.

title "Average Cholesterol by Age Group";

proc sgplot data=sashelp.heart; 9.3
format AgeAtStart agefmt.;
yaxis label="Avg. Cholesterol";
vbar AgeAtStart / response=cholesterol stat=mean limitstat=clm
datalabel datalabelpos=bottom;
run;
Figure 7.3.8: Bar Chart with Response
Sorting

By default, the bars are sorted by the

unformatted category value.

In this example, the CATEGORYORDER

option is used to sort the bars by
descending response value.

title "Product Sales";

proc sgplot data=sgbook.product_sales; 9.3
vbar product / response=actual categoryorder=respdesc;
run;
188 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.3.9: Bar Chart with Stacked

Groups

The GROUP option is use to specify an

additional level of classification in a bar
chart. By default, the group values are
represented as stacked bar segments. The
data labels can be displayed above or below
the chart.

The grouped bar attributes can be modified

by using an attribute map (see Chapter 9) or
by modifying an ODS style.

title "Product Sales";

proc sgplot data=sgbook.product_sales; 9.3
yaxis label="Sales (in millions)";
vbar product / response=actual group=region datalabel
datalabelpos=top;
run;

Figure 7.3.10: Bar Chart with Adjacent

Groups

Instead of stacking the group values, the

GROUPDISPLAY= CLUSTER option can be
used to display the bars side by side.

title "Product Sales";

proc sgplot data=sgbook.product_sales; 9.3
yaxis label="Sales (in millions)";
vbar product / response=actual group=region groupdisplay=cluster;
run;
 Chapter 7 Categorization Plots 189

Figure 7.3.11: Bar Chart with Groups and

Pattern Fills

The JOURNAL2 style is designed to be a

black-and-white style. When this style is
used with bar charts, fill patterns (SAS 9.3)
are used in situations that need unique
colors.

The MONOCHROMEPRINTER style is also

a black-and-white style, but uses serif fonts
instead of sans-serif fonts.

ods listing style=journal2;

title "Product Sales"; 9.3
proc sgplot data=sgbook.product_sales;
yaxis label="Sales (in millions)";
vbar product / response=actual group=region groupdisplay=cluster;
run;

Figure 7.3.12: Grouped Bar Chart using

PROC SGPANEL

In the previous grouping examples, the

group values are clustered around a
category value. In this example, the
category values are clustered around a
group value.

The NOBORDER option is used to remove

the border between cells. The LAYOUT
=COLUMNLATTICE option displays all cells
in one row.

title "Product Sales";

proc sgpanel data=sgbook.product_sales;
panelby region / layout=columnlattice noborder novarname onepanel;
vbar product / response=actual;
run;
190 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.3.13: Bullet Chart

This bullet chart is drawn using two overlaid

bar charts. The first chart is drawn using the
default bar width (0.85). The overlaid bar
chart is drawn using a bar width of 0.3.

This example can be enhanced by adding

transparency to the charts, enabling you to
see the top of the first bars through the
second bars.

title "Product Sales";

proc sgplot data=sgbook.product_sales;
yaxis label="Sales (in millions)";
vbar product / response=actual;
vbar product / response=predict barwidth=0.3;
run;

Figure 7.3.14: Overlaid Bar Charts with

Discrete Offsets

Overlaid bars can be placed side by side by

using the DISCRETEOFFSET options. Bar
widths have to be adjusted as shown.

A negative discrete offset moves the bars to

the left; a positive value moves the bars to
the right. The axis REVERSE option
reverses this behavior.

title "Product Sales";

proc sgplot data=sgbook.product_sales; 9.3
yaxis label="Sales (in millions)";
vbar product / response=actual barwidth=0.4 discreteoffset=-0.2;
vbar product / response=predict barwidth=0.4 discreteoffset=0.2;
run;
 Chapter 7 Categorization Plots 191

7.4 Horizontal Bar Charts

The HBAR plot shows the summarization of the response variable by
category or the frequency count of the category. The syntax is:

hbar column </ options>;

By default, the data is summarized for each unique category value;

however, user-defined formats can be used to create categorical
ranges for summarization. One bar is drawn for each resulting category. If a group variable is
specified, this bar can be subdivided by each group value, or the bar can be drawn as a cluster of
smaller group bars (see the GROUPDISPLAY option).

Required Data Roles:

CATEGORY column Category variable.

Optional Data Roles:

FREQ =num-column Frequency count for each observation
GROUP =column Group variable for each observation
RESPONSE =num-column Response values for the chart

Appearance Options:
FILLATTRS =fill-attrs Appearance attributes for fill

Other Options:
BARWIDTH =value Width of bar a fraction of tick spacing
DATASKIN =keyword Quasi-3D effect for the bars (GLOSS, etc.)
GROUPDISPLAY =keyword STACK | CLUSTER

Boolean Options:
FILL | NOFILL Controls bar fill
OUTLINE | NOOUTLINE Controls bar outline

Standard common options are supported

Related Style Elements: The GraphDataDefault style element is used to draw the plot elements.
When a GROUP variable is present, or when multiple plot statements are overlaid, the
GraphData1-GraphData12 elements are used, one for each group or plot statement.
192 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.4.1: Bar Chart with Fill

Attributes and Data Skins

This basic bar chart has been enhanced by

specifying a particular color for the bar fill.
In addition, the DATASKIN option is used to
give the bars a modern, three-dimensional
appearance while maintaining the visual
accuracy of a two-dimensional bar chart.
See the documentation for the different
skins available.

title "Product Sales";

proc sgplot data=sgbook.product_sales; 9.3
hbar product / response=actual fillattrs=(color=grayee)
dataskin=pressed;
run;

Figure 7.4.2: Bar Chart with No Fill

This example uses the NOFILL option to

create a “minimal ink” appearance that
reproduces well in black and white
publications.

There is also a NOOUTLINE option to

remove the bar outlines when you only want
the fill color.

title "Product Sales";

proc sgplot data=sgbook.product_sales;
hbar product / response=actual nofill;
run;
 Chapter 7 Categorization Plots 193

Figure 7.4.3: Bar Chart with Data Labels

This example displays the bar response

values above the bars by using the
DATALABEL option. This can be specified
with a variable to display information other
than the bar response values.

The DATALABELATTRS is used to override

the default text attributes of the labels.
CATEGORYORDER is used to sort the bars
by descending response.

title "Product Sales";

proc sgplot data=sgbook.product_sales; 9.3
hbar product / response=actual datalabel categoryorder=respdesc
datalabelattrs=(weight=bold style=italic);
run;

Figure 7.4.4: Bar Chart with a Reference

Line

Reference lines can be added to the bar

chart by specifying a REFLINE statement.
All plot statements are drawn in the order in
which they are specified.

Since the REFLINE statement is specified

after the HBAR statement, it is drawn in
front of the bar chart.

title "Product Sales";

proc sgplot data=sgbook.product_sales;
hbar product / response=actual;
refline 80 / axis=X;
run;
194 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.4.5: Bar Chart with Confidence

Limits

This example shows a bar chart showing

the confidence limit of the mean (CLM). The
standard error and standard deviation
statistics are also available.

The ALPHA option can be used to specify

the confidence for the CLM calculation. The
default is 0.05. The default legend includes
the limit statistic information.

title "Average Cholesterol by Age Group";

proc sgplot data=sashelp.heart;
format AgeAtStart agefmt.;
xaxis label="Avg. Cholesterol";
hbar AgeAtStart / response=cholesterol stat=mean limitstat=clm
alpha=0.05;
run;

Figure 7.4.6: Bar Chart with Upper Limit

You can use the LIMITS option to control

which side of the limits are displayed.

This example uses the STDERR statistic

with a NUMSTD multiplier to display two
standard errors instead of one.

title "Average Cholesterol by Age Group";

proc sgplot data=sashelp.heart;
format AgeAtStart agefmt.;
xaxis label="Avg. Cholesterol";
hbar AgeAtStart / response=cholesterol stat=mean limits=upper
limitstat=stderr numstd=2;
run;
 Chapter 7 Categorization Plots 195

Figure 7.4.7: Bar Chart with Groups

The GROUP option is use to specify an

additional level of classification in a bar
chart. By default, the group values are
represented as stacked bar segments.

The bar segments are automatically

assigned unique attributes for each group
value. These attributes can be modified by
using an attribute map (see Chapter 9) or by
modifying an ODS style.

title "Product Sales";

proc sgplot data=sgbook.product_sales;
xaxis label="Sales (in millions)";
hbar product / response=actual group=region;
run;

Figure 7.4.8: Bar Chart with Adjacent

Groups

Instead of stacking the group values, the

GROUPDISPLAY option can be used to
CLUSTER the bars.

title "Product Sales";

proc sgplot data=sgbook.product_sales; 9.3
xaxis label="Sales (in millions)";
hbar product / response=actual group=region groupdisplay=cluster;
run;
196 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.4.9: Bar Chart with Groups and

Pattern Fills

The JOURNAL2 style is designed to be a

black-and-white style. When this style is
used with bar charts, fill patterns (SAS 9.3)
are used in situations that need unique
colors.

The MONOCHROMEPRINTER style is also

a black-and-white style, but uses serif fonts
instead of sans-serif fonts.

title "Product Sales";

proc sgplot data=sgbook.product_sales; 9.3
xaxis label="Sales (in millions)";
hbar product / response=actual group=region groupdisplay=cluster;
run;

Figure 7.4.10: Grouped Bar Chart using

PROC SGPANEL

In the previous grouping examples, the

group values are clustered around a
category value. In this example, the
category values are clustered around a
group value.

The NOBORDER option is used to remove

the border between panel cells, and the
LAYOUT=ROWLATTICE option ensures
cells are in one column.

title "Product Sales";

proc sgpanel data=sgbook.product_sales;
panelby region / layout=rowlattice noborder novarname onepanel;
hbar product / response=actual;
run;
 Chapter 7 Categorization Plots 197

Figure 7.4.11: Bullet Chart

This bullet chart is drawn using two overlaid

bar charts. The first chart is drawn using the
default bar width (0.85). The overlaid bar
chart is drawn using a bar width of 0.3.

This example can be enhanced by adding

transparency to the charts, enabling you to
see the top of the first bars through the
second bars.

title "Product Sales";

proc sgplot data=sgbook.product_sales;
xaxis label="Sales (in millions)";
hbar product / response=actual;
hbar product / response=predict barwidth=0.3;
run;

Figure 7.4.12: Overlaid Bar Charts with

Discrete Offsets

Overlaid bars can be drawn side by side by

using the DISCRETEOFFSET option. Bar
widths have to be adjusted as shown.

A negative discrete offset moves the bars

up; a positive value moves the bars down.
The axis REVERSE option reverses this
behavior. The bars can be overlapped.

title "Product Sales";

proc sgplot data=sgbook.product_sales; 9.3
xaxis label="Sales (in millions)";
hbar product / response=actual barwidth=0.4 discreteoffset=-0.2;
hbar product / response=predict barwidth=0.4 discreteoffset=0.2;
run;
198 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

7.5 Vertical Line Charts

The VLINE plot shows the summarization of the response variable by
category or the frequency count of the category. The syntax is:

vline column </ options>;

By default, the data is summarized for each unique category value;

however, user-defined formats can be used to create categorical
ranges for summarization. One line is drawn connecting all resulting
categories. If a group variable is specified, one line is drawn for each group value, either overlaid
on the tick value or offset from the tick value to create a cluster effect (see the GROUPDISPLAY
option).

Required Data Roles:

CATEGORY column Category variable

Optional Data Roles:

FREQ =num-column Frequency count for each observation
GROUP =column Group variable for each observation
RESPONSE =column Response values for the chart

Appearance Options:
CURVELABELATTRS =text-attrs Appearance attributes for the line labels
LINEATTRS =line-attrs Appearance attributes for the line
MARKERATTRS =marker-attrs Appearance attributes for the markers

Other Options:
CURVELABEL <=”string”> Label(s) for the line(s)
CURVELABELLOC =keyword INSIDE | OUTSIDE
CURVELABELPOS =keyword AUTO | MIN | MAX | START | END
DATALABELPOS =keyword DATA | TOP | BOTTOM
GROUPDISPLAY =keyword OVERLAY | CLUSTER

Boolean Options:
BREAK Break the line at missing response values
MARKERS Display markers on the data points
X2AXIS Assign the plot to the secondary x-axis
Y2AXIS Assign the plot to the secondary y-axis

Standard common options are supported

Related Style Elements: Same as VBAR and HBAR.

 Chapter 7 Categorization Plots 199

Figure 7.5.1: Basic Line Chart

This example shows a basic line chart using

a MEAN statistic. The visual attributes of the
line can be controlled using the LINEATTRS
option.

The axis LABEL option is used to override

the column label from the data set.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
yaxis label="Workers (in thousands)";
vline date / response=electric stat=mean
lineattrs=(color=gray thickness=2);
run;

Figure 7.5.2: Line Chart with Markers

The MARKERS option displays markers on

the data points along the line.

The MARKERATTRS option is used to

control the attributes of the markers
independently from the line attributes.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
yaxis label="Workers (in thousands)";
vline date / response=electric stat=mean markers
lineattrs=(color=gray thickness=2)
markerattrs=(color=gray symbol=circlefilled);
run;
200 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.5.3: Line Chart with Break

By default, line charts connect all of the non-

missing response values. However, you
might want to show the missing values by
breaking the line using the BREAK option.

In this example, 1979 has all missing values.

The line continues, starting at the next non-
missing value.

title "Avg. Number of Electrical Workers";

proc sgplot data=sgbook.workers_missing;
format date year.;
yaxis label="Workers (in thousands)";
vline date / response=electric break stat=mean markers
lineattrs=(thickness=2) markerattrs=(symbol=circlefilled);
run;

Figure 7.5.4: Line Chart with Data Labels

This example displays the line response

values along the line by using the
DATALABEL option. This option can be
specified with a variable to display
information other than the line response
values.

The DATALABELATTRS can be used to

override the default text attributes of the
labels.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers; 9.3
format date year.;
yaxis label="Workers (in thousands)";
vline date / response=electric stat=mean markers datalabel
lineattrs=(thickness=2) markerattrs=(symbol=circlefilled)
datalabelattrs=(weight=bold style=italic);
run;
 Chapter 7 Categorization Plots 201

Figure 7.5.5: Line Chart with a Reference

Line

Reference lines can be added to the line

chart by specifying a REFLINE statement.
The order of the statements is significant.

Since the REFLINE statement is specified

after the VLINE statement, the reference
line is drawn in front of the line chart.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
yaxis label="Workers (in thousands)";
vline date / response=electric stat=mean lineattrs=(thickness=2)
markers markerattrs=(symbol=circlefilled);
refline 300;
run;

Figure 7.5.6: Line Chart with Confidence

Limits

This example shows a line chart showing

the confidence limit of the mean (CLM). The
standard error and standard deviation
statistics are also available.

The ALPHA option can be used to specify

the confidence for the CLM calculation. The
default is 0.05.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers noautolegend;
format date year.;
yaxis label="Workers (in thousands)";
vline date / response=electric stat=mean limitstat=clm
lineattrs=(thickness=2);
run;
202 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.5.7: Line Chart with Upper

Limits

You can use the LIMITS option to control

which side of the limits are displayed.

This example uses the STDERR statistic

with a NUMSTD multiplier to display two
standard errors instead of one.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers noautolegend;
format date year.;
yaxis label="Workers (in thousands)";
vline date / response=electric stat=mean limits=upper
limitstat=stderr numstd=2 lineattrs=(thickness=2);
run;

Figure 7.5.8: Line Chart with CLM and

Data Label Position

The DATALABELPOS option not only

moves the data labels above or below the
plot area, it will also display the limits (if
applicable).

The LABEL statement is used to control the

default column label for ELECTRIC (needed
for the data label table), but this label is
further overridden by the LABEL option on
the YAXIS statement.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers; 9.3
format date year.;
label electric="Workers (in thousands)";
yaxis label="Workers";
vline date / response=electric markers stat=mean limitstat=clm
markerattrs=(symbol=circlefilled) datalabel datalabelpos=bottom;
run;
 Chapter 7 Categorization Plots 203

Figure 7.5.9: Grouped Line Chart

The GROUP option is use to specify an

additional level of classification in a line
chart. The plot lines and markers are
automatically assigned unique attributes for
each group value. These attributes can be
modified by using an attribute map (see
Chapter 9) or by modifying an ODS style.
Data labels can be displayed above or
below the chart, as well as along the line.

title "Avg. Number of Electrical Workers";

9.3
proc sgplot data=sgbook.workers_group; /* Transpose of sashelp.workers */
label _name_="Type of Work";
format date year.;
yaxis label="Workers";
vline date / response=col1 markers stat=mean
group=_name_ datalabel datalabelpos=bottom;
run;

Figure 7.5.10: Line Chart with Curve

Labels

In this grouped plot example, the

CURVELABEL option uses the group values
to label the lines. When a line chart has no
group variable, you can specify a curve label
by specifying CURVELABEL=”<your
string>”.

Curve labels can be positioned inside or

outside the data area and at the beginning
or end of the plot line.

title "Avg. Number of Electrical Workers";

proc sgplot data=sgbook.workers_group;
format date year.;
yaxis label="Workers";
vline date / response=col1 markers stat=mean
group=_name_ curvelabel curvelabelloc=outside;
run;
204 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.5.11: Overlaid Line Charts

This example plots two response variables

by overlaying a line chart for each variable.

Each statement has its own set of visual

attributes and can be assigned to different
axes. Also, the legend label can be changed
for each chart by using the LEGENDLABEL
option.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
yaxis label="Workers (in thousands)";
vline date / response=electric stat=mean markers markerattrs=(color=black
symbol=circle) lineattrs=(color=black thickness=2);
vline date / response=masonry stat=mean markers markerattrs=(color=gray
symbol=circle) lineattrs=(color=gray pattern=solid thickness=2);
run;

Figure 7.5.12: Overlaid Line Charts with

Discrete Offset

Normally, response values of overlaid lines

are drawn at the tick positions along the
category axis, which can cause overlap. To
move the values off of the tick positions, use
the DISCRETEOFFSET option.

Negative values down to -0.5 move the

points to the left. Positive values up to 0.5
move the points to the right.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers noautolegend; 9.3
format date year.;
yaxis label="Workers (in thousands)"; xaxis offsetmin=0.05 offsetmax=0.05;
vline date / response=electric stat=mean markers limitstat=clm discreteoffset=-0.1
markerattrs=(color=black symbol=circle) lineattrs=(color=black thickness=2);
vline date / response=masonry stat=mean markers discreteoffset=0.1 limitstat=clm
markerattrs=(color=gray symbol=circle) lineattrs=(color=gray pattern=solid
thickness=2);
run;
 Chapter 7 Categorization Plots 205

7.6 Horizontal Line Charts

The HLINE plot shows the summarization of the response variable by
category or the frequency count of the category. The syntax is:

hline column </ options>;

By default, the data is summarized for each unique category value;

however, user-defined formats can be used to create categorical
ranges for summarization. One line is drawn connecting all resulting categories. If a group
variable is specified, one line is drawn for each group value, either overlaid on the tick value or
offset from the tick value to create a cluster effect (see the GROUPDISPLAY option).

Required Data Roles:

CATEGORY column Category variable

Optional Data Roles:

FREQ =num-column Frequency count for each observation
GROUP =column Group variable for each observation
RESPONSE =column Response values for the chart

Appearance Options:
CURVELABELATTRS =text-attrs Appearance attributes for the line labels
LINEATTRS =line-attrs Appearance attributes for the line
MARKERATTRS =marker-attrs Appearance attributes for the markers

Other Options:
CURVELABEL <=”string”> Label(s) for the line(s)
CURVELABELLOC =keyword INSIDE | OUTSIDE
CURVELABELPOS =keyword AUTO | MIN | MAX | START | END
GROUPDISPLAY =keyword OVERLAY | CLUSTER

Boolean Options:
BREAK Break the line at missing response values
MARKERS Display markers on the data points
X2AXIS Assign the plot to the secondary x-axis
Y2AXIS Assign the plot to the secondary y-axis

Standard common options are supported

Related Style Elements: Same as VBAR and HBAR.

206 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.6.1: Basic Line Chart

This example shows a basic line chart using

a MEAN statistic. The visual attributes of the
line can be controlled using the LINEATTRS
option.

The axis LABEL option is used to override

the column label from the data set.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
yaxis grid;
xaxis label="Workers (in thousands)";
hline date / response=electric stat=mean
lineattrs=(color=gray thickness=2);
run;

Figure 7.6.2: Line Chart with Markers

The MARKERS option displays markers on

the data points along the line.

The MARKERATTRS option is used to

control the attributes of the markers
independently from the line attributes.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
yaxis grid;
xaxis label="Workers (in thousands)";
hline date / response=electric stat=mean lineattrs=(color=gray thickness=2)
markerattrs=(color=gray symbol=circlefilled) markers;
run;
 Chapter 7 Categorization Plots 207

Figure 7.6.3: Line Chart with Break

By default, line charts connect all of the

non-missing response values. However, you
might want to show the missing values by
breaking the line using the BREAK option.

In this example, 1979 has all missing

values. The line continues, starting at the
next non-missing value.

title "Avg. Number of Electrical Workers";

proc sgplot data=sgbook.workers_missing;
format date year.;
yaxis grid;
xaxis label="Workers (in thousands)";
hline date / response=electric break stat=mean lineattrs=(thickness=2)
markers markerattrs=(symbol=circlefilled);
run;

Figure 7.6.4: Line Chart with Data Labels

This example displays the line response

values along the line by using the
DATALABEL option. This option can be
specified with a variable to display
information other than the line response
values.

The DATALABELATTRS can be used to

override the default text attributes of the
labels.

title "Avg. Number of Electrical Workers";

9.3
proc sgplot data=sashelp.workers;
format date year.;
yaxis grid;
xaxis label="Workers (in thousands)";
hline date / response=electric stat=mean markers lineattrs=(thickness=2)
markerattrs=(symbol=circlefilled) datalabel
datalabelattrs=(weight=bold style=italic);
run;
208 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.6.5: Line Chart with a Reference

Line

Reference lines can be added to the line

chart by specifying a REFLINE statement.
The order of the statements is significant.

Since the REFLINE statement is specified

after the HLINE statement, the reference
line is drawn in front of the line chart.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
xaxis label="Workers (in thousands)";
hline date / response=electric stat=mean lineattrs=(thickness=2) markers
markerattrs=(symbol=circlefilled);
refline 300 / axis=X;
run;

Figure 7.6.6: Line Chart with Confidence

Limits

This example shows a line chart showing

the confidence limit of the mean (CLM). The
standard error and standard deviation
statistics are also available.

The ALPHA option can be used to specify

the confidence for the CLM calculation. The
default is 0.05.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers noautolegend;
format date year.;
xaxis label="Workers (in thousands)" grid;
hline date / response=electric stat=mean limitstat=clm
lineattrs=(thickness=2);
run;
 Chapter 7 Categorization Plots 209

Figure 7.6.7: Line Chart with Upper Limits

You can use the LIMITS option to control

which side of the limits are displayed.

This example uses the STDERR statistic

with a NUMSTD multiplier to display two
standard errors instead of one.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers noautolegend;
format date year.;
xaxis label="Workers (in thousands)" grid;
hline date / response=electric stat=mean limits=upper
limitstat=stderr numstd=2 lineattrs=(thickness=2);
run;

Figure 7.6.8: Line Chart with Response

Sorting

Use the CATEGORYORDER option to order

your category values based on ascending or
descending response values.

In this example, the products with the

highest sales are sorted to the top of the
axis.

title "Product Sales for 1993";

proc sgplot data=sashelp.prdsale (where=(year=1993));
yaxis grid;
hline product / response=actual categoryorder=respdesc markers
markerattrs=(symbol=circlefilled);
run;
210 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.6.9: Grouped Line Chart

The GROUP option is used to specify an

additional level of classification in a line
chart. The plot lines and markers are
automatically assigned unique attributes for
each group value.

These attributes can be modified by using

an attribute map (see Chapter 9) or by
modifying an ODS style.

title "Avg. Number of Electrical Workers";

proc sgplot data=sgbook.workers_group;
format date year.;
label _name_="Type of Work";
xaxis label="Workers";
hline date / response=col1 markers group=_name_;
run;

Figure 7.6.10: Line Chart with Curve

Labels

This grouped plot example uses the

CURVELABEL option to label the lines.

When a line chart has no group variable,

you can specify a curve label by specifying
CURVELABEL=”<string>”. Curve labels can
be positioned inside or outside the data area
and at the beginning or end of the plot line.

title "Avg. Number of Electrical Workers";

proc sgplot data=sgbook.workers_group;
format date year.;
xaxis label="Workers";
hline date / response=col1 markers stat=mean
group=_name_ curvelabel;
run;
 Chapter 7 Categorization Plots 211

Figure 7.6.11: Overlaid Line Charts

This example plots two response variables

by overlaying a line chart for each variable.

Each statement has its own set of visual

attributes, and can be assigned to different
axes. Also, the legend label can be changed
for each chart by using the LEGENDLABEL
option.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
xaxis label="Workers (in thousands)";
hline date / response=electric stat=mean markers markerattrs=(color=black
symbol=circle) lineattrs=(color=black thickness=2);
hline date / response=masonry stat=mean markers markerattrs=(color=gray
symbol=circle) lineattrs=(color=gray pattern=solid thickness=2);
run;

Figure 7.6.12: Overlaid Line Charts with

Discrete Offset

Normally, response values of overlaid lines

are drawn at the tick positions along the
category axis, which can cause overlap.

To move the values off of the tick positions,

use the DISCRETEOFFSET option.
Negative values down to -0.5 move the
points down. Positive values up to 0.5 move
the points up.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers; 9.3
format date year.;
xaxis label="Workers (in thousands)"; xaxis offsetmin=0.05 offsetmax=0.05;
hline date / response=electric stat=mean markers limitstat=clm discreteoffset=-0.1
markerattrs=(color=black symbol=circle) lineattrs=(color=black thickness=2);
hline date / response=masonry stat=mean markers discreteoffset=0.1 limitstat=clm
markerattrs=(color=gray symbol=circle) lineattrs=(color=gray pattern=solid
thickness=2);
run;
212 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

7.7 Dot Plots

The dot plot shows the summarization of the response variable by
category or the frequency count of the category. The syntax is:

dot column </ options>;

By default, the data is summarized for each unique category value;

however, user-defined formats can be used to create categorical
ranges for summarization. One marker is drawn for resulting category. If a group variable is
specified, one marker is drawn for each group value, either overlaid on the tick value or offset
from the tick value to create a cluster effect (see the GROUPDISPLAY option).

Required Data Roles:

CATEGORY column Category variable

Optional Data Roles:

FREQ =num-column Frequency count for each observation
GROUP =column Group variable for each observation
RESPONSE =column Response values for the chart

Appearance Options:
MARKERATTRS =marker-attrs Appearance attributes for the markers

Other Options:
GROUPDISPLAY =keyword OVERLAY | CLUSTER

Boolean Options:
X2AXIS Assign the plot to the secondary x-axis
Y2AXIS Assign the plot to the secondary y-axis

Standard common options are supported

Related Style Elements: Same as VBAR and HBAR.

 Chapter 7 Categorization Plots 213

Figure 7.7.1: Basic Dot Plot

Dot plots are highly effective plots for

comparing a large number of category
values.

This example shows a basic dot plot using a

MEAN statistic. The axis LABEL option is
used to override the column label from the
data set.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
xaxis label="Workers (in thousands)";
dot date / response=electric stat=mean;
run;

Figure 7.7.2: Dot Plot with Marker

Attributes

The MARKERATTRS option is used to

control the attributes of the markers in the
plot.

In this example, the marker shape and the

marker size are modified.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
xaxis label="Workers (in thousands)";
dot date / response=electric stat=mean
markerattrs=(symbol=diamondfilled size=11px);
run;
214 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.7.3: Dot Plot with Data Labels

This example displays the plot response

values by using the DATALABEL option.
This option can be specified with a variable
to display information other than the plot
response values.

The DATALABELATTRS can be used to

override the default text attributes of the
labels.

title "Avg. Number of Electrical Workers";

proc sgplot data=sgbook.workers_missing; 9.3
format date year.;
xaxis label="Workers (in thousands)";
dot date / response=electric stat=mean datalabel
datalabelattrs=(weight=bold style=italic);
run;

Figure 7.7.4: Dot Plot with a Reference

Line

Reference lines can be added to the dot plot

by specifying a REFLINE statement.

The order of the statements is significant. If

the REFLINE statement is specified before
the DOT statement, the line is drawn behind
the dot plot. REFLINE statements specified
after the DOT statement are drawn in front
of the dot plot.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
xaxis label="Workers (in thousands)";
dot date / response=electric stat=mean;
refline 300 / axis=X;
run;
 Chapter 7 Categorization Plots 215

Figure 7.7.5: Dot Plot with Confidence

Limits

This example shows a dot plot showing the

confidence limit of the mean (CLM). The
standard error and standard deviation
statistics are also available.

The ALPHA option can be used to specify

the confidence for the CLM calculation. The
default is 0.05.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers noautolegend;
format date year.;
xaxis label="Workers (in thousands)" grid;
dot date / response=electric stat=mean limitstat=clm;
run;

Figure 7.7.6: Dot Plot with Upper Limits

You can use the LIMITS option to control

which side of the limits are displayed.

This example uses the STDERR statistic

with a NUMSTD multiplier to display two
standard errors instead of one.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers noautolegend;
format date year.;
xaxis label="Workers (in thousands)" grid;
dot date / response=electric stat=mean limits=upper
limitstat=stderr numstd=2;
run;
216 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 7.7.7: Dot Plot with Response

Sorting

Use the CATEGORYORDER option to order

your category values based on ascending or
descending response values.

In this example, the products with the

highest sales are sorted to the top of the
axis.

title "Product Sales for 1993";

proc sgplot data=sashelp.prdsale (where=(year=1993)); 9.3
dot product / response=actual categoryorder=respdesc;
run;

Figure 7.7.8: Grouped Dot Plot

The GROUP option is used to specify an

additional level of classification in a dot plot.
The plot markers are automatically assigned
unique attributes for each group value.

These marker attributes can be modified by

using an attribute map (see Chapter 9) or by
modifying an ODS style.

title "Avg. Number of Electrical Workers";

proc sgplot data=sgbook.workers_group;
format date year.;
label _name_="Type of Work";
xaxis label="Workers";
dot date / response=col1 group=_name_ stat=mean;
run;
 Chapter 7 Categorization Plots 217

Figure 7.7.9: Overlaid Dot Plots

This example plots two response variables

by overlaying a dot plot for each variable.
Each statement has its own set visual
attributes and can be assigned to different
axes.

The legend label can be changed for each

plot by using the LEGENDLABEL option.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers;
format date year.;
xaxis label="Workers (in thousands)";
dot date / response=electric stat=mean;
dot date / response=masonry stat=mean;
run;

Figure 7.7.10: Dot Plot with Discrete

Offsets

Normally, response values of overlaid dot

plots are drawn at the tick positions along
the category axis, which can cause overlap.

To move the values off of the tick positions,

use the DISCRETEOFFSET option.
Negative values down to -0.5 move the
points down. Positive values up to 0.5 move
them up.

title "Avg. Number of Electrical Workers";

proc sgplot data=sashelp.workers; 9.3
format date year.;
xaxis label="Workers (in thousands)";
yaxis offsetmin=0.075 offsetmax=0.075;
dot date / response=electric stat=mean limitstat=clm
discreteoffset=-0.15 markerattrs=(symbol=circle);
dot date / response=masonry stat=mean discreteoffset=0.15
limitstat=clm markerattrs=(symbol=circlefilled);
run;
218 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

7.8 Combining the Plots

Figure 7.8.1: Bar-Line Overlay

In this example, a line chart is overlaid on a

bar chart. The line chart is assigned to the
Y2 axis to display the weight independently
from the cholesterol range.

The data labels are displayed for each chart

and positioned above and below the chart.
The Y2 axis minimum is explicitly set to zero
to synchronize with the y-axis range.

title "Average Cholesterol and Weight by Age Group";

proc sgplot data=sashelp.heart noautolegend; 9.3
format AgeAtStart agefmt.;
yaxis label="Cholesterol";
y2axis label="Weight" min=0;
vbar AgeAtStart / response=cholesterol stat=mean datalabel
datalabelpos=bottom;
vline AgeAtStart / response=weight stat=mean y2axis markers
datalabel datalabelpos=top;
run;

Figure 7.8.2: HLINE-Dot Overlay

In this example, the HLINE draws the actual

values in descending order. The line creates
a visual separation for evaluating the
predicted values.

Predicted values to the left of the line mean

that sales exceeded the predicted value.
Predicted values to the right of the line
mean that sales fell short of the predicted
value.

title "Product Sales for 1993";

proc sgplot data=sashelp.prdsale (where=(year=1993));
hline product / response=actual categoryorder=respdesc markers;
dot product / response=predict;
run;
Chapter 8
Axes, Legends, and Insets
8.1 Introduction 221
8.2 Linear Axis 223
8.3 Log Axis 225
8.4 Time Axis 227
8.5 Discrete Axis 229
8.6 Legends 230
8.7 Insets 232
220 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“Never interrupt your enemy

when he is making a mistake.”
~ Napoleon Bonaparte
 Chapter 8 Axis, Legends, and Insets 221

Chapter 8: Axes, Legends, and Insets

In this chapter, we examine the options for controlling the axes, legends, and insets. You have
seen many of these options used throughout the examples in this book. This chapter will focus on
the capabilities you have not yet seen.

8.1 Introduction
The SGPLOT procedure supports up to four axes (XAXIS, YAXIS, X2AXIS, and Y2AXIS), while
the SGPANEL procedure supports only two (ROWAXIS and COLAXIS). The axis statements are
contained within the procedure code and can be specified in any order. For the secondary axes
to appear in SGPLOT, a plot must be assigned to them using the X2AXIS or Y2AXIS plot options.

The SGPLOT and SGPANEL procedures support four types of axes: LINEAR, LOG, TIME, and
DISCRETE. The axis type is chosen automatically based on the chart type, the data, and the
formats used; however, the LOG type is never chosen automatically. If you override the default
axis type, be sure you do not specify an axis type that is incompatible with your chart type;
otherwise, your chart will not draw. For example, if you request a vertical bar chart, the x-axis
type is automatically set to DISCRETE. If you attempt to set the x-axis type to any other value,
you will get a note in the log and the chart will not draw.

The following axis options can be used regardless of axis type. Axis options that apply to a
specific type will be addressed in the following sections.

Common Axis Options:

DISPLAY =display-value The features of the axis to display

GRID Control the display of grid lines
LABEL =string The label of the axis
LABELATTRS =text-attrs Text attributes for the axis label
OFFSETMIN =number Tick offset preceding the lowest value on the
axis
OFFSETMAX =number Tick offset following the highest value on the
axis
REFTICKS Mirrors the ticks on the opposite axis
REVERSE Reverses the order of the tick marks
TYPE =keyword LINEAR | LOG | TIME | DISCRETE
VALUEATTRS =text-attrs Text attributes for the axis tick values

The KEYLEGEND statement is used to specify the content and the attributes of custom legends,
as well as control the attributes of automatic legends. This statement works the same way in both
the SGPLOT and SGPANEL procedures, with the exception of the LOCATION option. The
LOCATION option is not available in the SGPANEL procedure because internal legends are not
222 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

supported. The SGSCATTER procedure has a LEGEND option that supports a subset of the
options in the following legend table. See the documentation for more details.

If the legend is located inside, and the POSITION option is not specified, the legend will
automatically position itself in an area with the least amount of data collision. This feature is very
useful for batch programs. If the plot data varies significantly from one run to the next, the legend
will move to the best possible position without any program change.

Legend Options:

ACROSS =integer Specifies the number of columns in the

legend
BORDER / NOBORDER Controls the display of the legend border
DOWN =integer Specifies the number of rows in the legend
LOCATION =keyword INSIDE | OUTSIDE
POSITION =keyword Specifies the legend position within the
legend location (TOP, BOTTOM, etc.)
TITLE =string Specifies a title for the legend
TITLEATTRS =text-attrs Specifies text attributes for the legend title
VALUEATTRS =text-attrs Specifies text attributes for the legend values

The INSET statement is used to add additional text or small tables to an SGPLOT graph. As the
statement name suggests, the inset is always drawn within the plot area of the graph. The
information in the inset is specified directly on the statement. See the inset examples for more
details.

If the POSITION option is not specified, the inset will automatically position itself in an area with
the least amount of data collision. This feature is very useful for batch programs. If the plot data
varies significantly from one run to the next, the inset will move to the best possible position
without any program change.

Inset Options:

BORDER / NOBORDER Controls the display of the inset border

LABELALIGN =keyword LEFT | RIGHT | CENTER
POSITION =keyword Specifies the inset position within the
data area (TOP, BOTTOM, etc.)
TEXTATTRS =text-attrs Text attributes for the inset text
TITLE = string Specifies a title for the inset
TITLEATTRS =text-attrs Text attributes for the inset title
VALUEALIGN =keyword LEFT | RIGHT | CENTER
 Chapter 8 Axis, Legends, and Insets 223

8.2 Linear Axis

Figure 8.2.1: Specifying Axis Min/Max

In this example, the MIN and MAX options

are used to set the data range of the x-axis.
These options set the extent of the data
range but not necessarily the first and last
tick values.

Appropriate tick values are automatically

calculated to fit within the requested range.
These options are available on all axis types
except DISCRETE.

title "Class Statistics";

proc sgplot data=sashelp.class;
xaxis min=75 max=125;
loess x=weight y=height / clm;
run;

Figure 8.2.2: Specifying a Tick List

In this example, the VALUES option is used

to set both the extents of the axis and the
tick values to display. The tick values can be
specified as a list, a sequence, or both.

If the values do not have equal intervals, the

values will still be placed in the correct
numerical location along the axis. This
option is supported by LINEAR and TIME
axes.

title "Class Statistics";

proc sgplot data=sashelp.class;
xaxis values=(70 to 120 by 10);
loess x=weight y=height / clm;
run;
224 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 8.2.3: The VALUESHINT Option

By default, the low and high values on the

VALUES option set the extents of the axis.

The VALUESHINT option can be used to

show the full data extent is displayed, and
the tick values in the list are drawn on the
axis if they fall within the data extent. This
option is supported on LINEAR and TIME
axes.

title "Class Statistics";

proc sgplot data=sashelp.class;
xaxis values=(70 to 120 by 10) valueshint;
loess x=weight y=height / clm;
run;

Figure 8.2.4: The INTEGER Option

If an axis displays a small data range, it is

possible that tick values with decimal values
can appear.

To prevent this, you can use the INTEGER

option to force the axis to display only
integer values.

title "Low Response Values";

proc sgplot data=sgbook.random;
yaxis integer;
vbar x / response=y dataskin=pressed;
run;
 Chapter 8 Axis, Legends, and Insets 225

8.3 Log Axis

Figure 8.3.1: Log Expand Style

In this example, the log axis shows the

expanded form of the log exponent.

The MINOR option is specified to display

the minor tick marks on the axis.

title "Plot of Y by X";

proc sgplot data=sgbook.logdata;
xaxis integer;
yaxis type=log logstyle=LogExpand label="log(y)" minor;
series x=x y=y / markers;
run;

Figure 8.3.2: Log Exponent Style

In this example, the log axis displays the log

exponent instead of the expanded form.

title "Plot of Y by X";

proc sgplot data=sgbook.logdata;
xaxis integer;
yaxis type=log logstyle=LogExponent label="log(y)";
series x=x y=y / markers;
run;
226 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 8.3.3: Log Linear Style

With the LINEAR style of log axis, the major

tick marks are placed at non-uniform
locations covering the range of the axis.

title "Plot of Y by X";

proc sgplot data=sgbook.logdata;
xaxis integer;
yaxis type=log logstyle=Linear label="log(y)";
series x=x y=y / markers;
run;

Figure 8.3.4: The LOGBASE Option

This option sets the base for the log

calculations. The three valid values are 10,
2, and e.

title "Plot of Y by X";

proc sgplot data=sgbook.logdata;
xaxis integer;
yaxis type=log logstyle=LogExpand label="log(y)" logbase=e;
series x=x y=y / markers;
run;
 Chapter 8 Axis, Legends, and Insets 227

8.4 Time Axis

Figure 8.4.1: The INTERVAL Option

By default, the time axis calculates the best

time interval and format. You can use the
INTERVAL option to override the default
interval.

title "Air Travel";

proc sgplot data=sashelp.air;
where date >= '01jan1959'd and date <='01jan1961'd;
xaxis offsetmin=0 interval=semiyear;
yaxis offsetmin=0 label="Intl. Air Travel";
band x=date upper=air lower=300;
run;

Figure 8.4.2: The NOTIMESPLIT Option

By default, the horizontal time axis extracts

the highest time order and displays it on a
second line to save space.

The NOTIMESPLIT option prevents the

extraction. Compare this example with
Figure 8.4.1. In this example, the year is
kept with each tick value.

title "Air Travel";

proc sgplot data=sashelp.air;
where date >= '01jan1959'd and date <='01jan1961'd;
xaxis offsetmin=0 interval=semiyear notimesplit;
yaxis offsetmin=0 label="Intl. Air Travel";
band x=date upper=air lower=300;
run;
228 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 8.4.3: The TICKVALUEFORMAT

Option

By default, the time axis determines the

best format to use based on the interval.

You can override this behavior by specifying

a format using the TICKVALUEFORMAT
option. This option is also supported by the
LINEAR axis.

title "Air Travel";

proc sgplot data=sashelp.air;
where date >= '01jan1959'd and date <='01jan1961'd;
yaxis offsetmin=0 label="Intl. Air Travel";
xaxis offsetmin=0 interval=semiyear tickvalueformat=monyy5.;
band x=date upper=air lower=300;
run;

Figure 8.4.4: The MINOR Option

By default, minor tick marks are not drawn

on the axis.

The MINOR option turns them on for both

TIME axes and LOG axes.

title "Air Travel";

proc sgplot data=sashelp.air;
where date >= '01jan1959'd and date <='01jan1961'd;
yaxis offsetmin=0 label="Intl. Air Travel";
xaxis offsetmin=0 interval=semiyear tickvalueformat=monyy5. minor;
band x=date upper=air lower=300;
run;
 Chapter 8 Axis, Legends, and Insets 229

8.5 Discrete Axis

Figure 8.5.1: Fit Policy

When collisions occur on a horizontal

discrete axis, the FITPOLICY option can be
used to control which policy or sequence of
policies are used to resolve the collision.

In this example, the STAGGER policy was

used instead of the ROTATE default.

title "Class Statistics";

proc sgplot data=sashelp.class (obs=10); 9.3
xaxis fitpolicy=stagger;
vbar name / response=height dataskin=pressed;
run;

Figure 8.5.2: Discrete Order

By default, the tick values of categorical

charts and box plots are sorted by
ascending unformatted value. The
DISCRETEORDER option can be used to
change the sort order.

In this example, the bars are drawn in

DATA order. For bar charts, line charts, and
dot plots, the CATEGORYORDER option
can be used to sort by response values.

title "Product Sales";

proc sgplot data=sashelp.prdsale; 9.3
xaxis discreteorder=data;
vbar product / response=actual dataskin=pressed;
run;
230 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

8.6 Legends
Figure 8.6.1: Automatic Legends

When grouped plots or multiple plots are

specified, the SG procedures evaluate the
request to determine the best legend to
display.

Only one automatic legend is drawn. If more

legends are needed, you need to create
custom legends (see Figure 8.6.2).

Title "Class Statistics";

proc sgplot data=sashelp.class;
scatter x=weight y=height / group=sex;
run;

Figure 8.6.2: Custom Legends

To customize the default legend or to create

your own, use the KEYLEGEND statement.

Assign names to the plots to be included in

the legend(s) and list their names on the
KEYLEGEND statement(s) in the order in
which you want them displayed. The
LEGENDLABEL plot option can be used to
customize the legend label for a plot without
groups.
Title "Class Statistics";
proc sgplot data=sashelp.class;
scatter x=weight y=height / group=sex name="s";
ellipse x=weight y=height / name="e1";
ellipse x=weight y=height / alpha=0.2 name="e2";
keylegend "e1" "e2";
keylegend "s" / location=inside title="Sex";
run;
 Chapter 8 Axis, Legends, and Insets 231

Figure 8.6.3: The LOCATION and

POSITION Options

The LOCATION option specifies whether a

legend is inside or outside of the data area.
The POSITION option species one of eight
positions in the legend’s location.

If the legend is inside the data area, and

POSITION is not specified, the legend will
automatically position itself in an area with
the least amount of data collision.

Title "Class Statistics";

proc sgplot data=sashelp.class;
scatter x=weight y=height / group=sex name="s";
ellipse x=weight y=height / name="e1";
ellipse x=weight y=height / alpha=0.2 name="e2";
keylegend "e1" "e2";
keylegend "s" / location=inside position=bottomright title="Sex";
run;

Figure 8.6.4: The ACROSS and DOWN

Options

The ACROSS and DOWN options control

the layout of the legend entries. The
ACROSS option limits how many legend
entries are drawn across.

In this example, setting ACROSS=1 forces

the legend entries to be drawn in one
column.

Title "Class Statistics";

proc sgplot data=sashelp.class;
scatter x=weight y=height / group=sex name="s";
ellipse x=weight y=height / name="e1";
ellipse x=weight y=height / alpha=0.2 name="e2";
keylegend "e1" "e2";
keylegend "s" / location=inside across=1 title="Sex";
run;
232 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

8.7 Insets
Figure 8.7.1: Inset Types

This example shows two kinds of insets: a

text inset and a table inset. Each quoted
string in a text inset creates a new line. Each
label/value pair in a table inset creates a
new entry in the table.

Insets can be positioned in any of eight

positions. However, if no position is
specified, the inset automatically positions
itself in an area with the least amount of
data collision.

Title "Class Statistics";

9.3
proc sgplot data=sashelp.class;
reg x=weight y=height / clm;
inset ("Date"="02/15/11" "nobs"="19") / title="Survey #1" border
textattrs=(size=7pt) titleattrs=(size=8 weight=bold);
inset "The class size survey will be taken"
"once a month for the next 6 months" / textattrs=(size=7pt);
run;

Figure 8.7.2: Rich-Text Inset

Not only do insets support Unicode

characters like other strings in the SG
procedures, but they also support
superscripts and subscripts.

Use the ODS ESCAPECHAR option to

insert rich text into your inset strings.

ods escapechar='~';
proc sgplot data=sashelp.class; 9.3
reg x=weight y=height / clm;
inset ( "Y~{unicode bar}"="62.34" "R~{sup '2'}"="0.94"
"~{unicode alpha}"=".05" ) / border textattrs=(size=7pt);
run;
Chapter 9
Annotation and Attribute Maps (SAS 9.3)
9.1 Annotation 235
9.2 Attribute Maps (9.3) 251
234 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“Only two things are infinite, the universe and human

stupidity, and I'm not sure about the former.”
~ Albert Einstein
 Chapter 9 Annotation and Attribute Maps (SAS 9.3) 235

Chapter 9: Annotation and Attribute Maps (SAS 9.3)

In SAS 9.3, two new features were added to the SG procedures that are driven by auxiliary data
sets. The first feature is an annotation facility that enables you to draw text and other graphical
primitives on your graph output to supplement or enhance the plot information. The second
feature is called an attribute map, which enables you to define the visual attributes of particular
group values.

9.1 Annotation
The basic idea for annotation is that you can define a data set that contains the custom drawing
actions you want to perform on a graph (see Figure 9.1.1). The data set has reserved column
names, each providing the specific information needed to create the graphical primitives. The
name of this data set is provided on the procedure statement using the SGANNO option.

Each observation draws one graphical primitive on the graph as defined by the FUNCTION
column. If the function is a POLYGON or a POLYLINE, then this observation, together with
subsequent POLYCONT values, defines the vertices of the figure. The information needed for
each function is provided in the other named columns of the data set. Figure 9.1.2 shows the
program code needed to create the graph shown in Figure 9.1.3.

Function drawspace x1 x2 yc1 yc2 title "Top Global Automakers…”;

Polyline datavalue 2.3 . Chrysler proc sgplot data=autos sganno=anno;
Polycont datavalue 3.5 . Chrysler hbarparm category=Automaker
Polycont datavalue 3.5 . Fiat response=million_units / <options>;
Polycont datavalue 2.8 . Fiat yaxis display=(nolabel) reverse;
Polyline datavalue 3.5 . Suzuki run;
Polycont datavalue 4.5 . Suzuki
Arrow datavalue 4.5 4.5 Suzuki Hyundai
Figure 9.1.1 Figure 9.1.2

Figure 9.1.3
236 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

When you create the annotation data set using DATA step code, be sure to check the length of
your columns. This annotation definition uses a number of keywords of varying lengths. You
might start with an attribute that is only four characters wide and change it to something eight
characters wide later in the DATA step. The eight-character value will be truncated to four
characters unless you use the LENGTH statement or set a column range in the INPUT statement.
If you see odd results in your graph, be sure to check the log to see if invalid attributes are
specified due to truncation.

Many SAS/GRAPH users will be familiar with the annotation concepts described in this chapter,
as they are based on the SAS/GRAPH annotation facility. However, this new facility has been
redesigned to take advantage of the capabilities of the ODS Graphics system, such as
transparency and rich text support. The data set column names and values have been defined in
a way to help make them more memorable and self-documenting. The facility uses various
drawing spaces in non-paneled plots to help simplify the placement of graphics primitives.

For the SAS 9.3 release, the following functions are supported in the FUNCTION column:
• TEXT – Used to draw a text string in the graph.
• TEXTCONT – Used to continue a text string from a previous TEXT or TEXTCONT
function. This function is typically to perform rich text operations, such as changing color
or font attributes.
• IMAGE – Used to place an image into the graph.
• LINE – Used to draw a line segment in the graph.
• ARROW – Used to draw an arrow in the graph.
• RECTANGLE – Used to draw rectangles or squares in the graph.
• OVAL – Used to draw ovals or circles in the graph.
• POLYGON – Used to draw a closed line figure in the graph. Because this is a closed
figure, the figure may be filled. This operation specifies the starting point of the polygon.
Each subsequent POLYCONT operation is used to draw one line segment from the
previous ending point.
• POLYLINE – Used to draw an open line figure in the graph. This operation specifies the
starting point of the polygon. Each subsequent POLYCONT operation is used to draw
one line segment from the previous ending point.
• POLYCONT – Used to specify the next drawing point for a POLYGON or POLYLINE
figure. All visual attributes for the figure must be specified on the initial POLYGON or
POLYLINE statement.
 Chapter 9 Annotation and Attribute Maps (SAS 9.3) 237

An important concept to understand when working with annotations is the concept of drawspace.
The drawspace specification is a combination of both an area and a unit. Not all areas are
supported by all SG procedures. The SGPLOT procedure supports all areas, while the SGPANEL
and SGSCATTER procedures support only the graph and layout areas. The graph below shows
where these areas are located in a typical SGPLOT output (Figure 9.1.4).

Figure 9.1.4

The units are either PERCENT or PIXEL, except the DATA area which also supports a VALUE
unit. As an example, if you wanted to draw some annotation in the wall area using percentages,
you could use the WALLPERCENT drawspace. Currently, annotations must be drawn using
absolute coordinates; there is not a relative drawing mode.

Annotations do not reserve space in the graph. If you need to reserve space in the graph for an
annotation, there are two techniques you can use. If you need space outside of the plot area, the
SG procedures support a new option called PAD that gives you the ability to padding to any of the
four edges of the graph. To reserve space within the plot area, use the OFFSETMIN and
OFFSETMAX options on the axis statements.
238 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 9.1.5: Creating Axis Tick Values with Unicode

This example uses a user-defined format to create age ranges for each box. To show these
ranges in the most compact way, you can use a Unicode “≤” in the tick value. To add these
characters, you can turn off the tick values and use annotation to create the tick values. The ODS
ESCAPECHAR functionality can be used to add Unicode, superscripts, or subscript to your
annotation strings. The data set contains a value for each range so that the values can be
positioned in data space. The label uses the Unicode function to generate the “≤” (‘2264’x). The
new PAD option is used to reserve space for the annotated axis values.

function y1space x1space y1 width x1 label

text graphpercent datavalue 7 15 21 20 ≤ 30

text graphpercent datavalue 7 15 31 30 ≤ 40

data anno;
retain function 'text' y1space 'graphpercent' x1space 'datavalue' y1 7 width 15;
input x1 label $ 4-33;
cards;
21 20 (*ESC*){unicode '2264'x} 30
31 30 (*ESC*){unicode '2264'x} 40
41 40 (*ESC*){unicode '2264'x} 50
51 50 (*ESC*){unicode '2264'x} 60
61 60 (*ESC*){unicode '2264'x} 70
;
run;

ods graphics / reset=index width=6in height=3in;

Title1 "Cholesterol Level by Age Range";
proc sgplot data=sashelp.heart sganno=anno pad=(bottom=8%);
format AgeAtStart agefmt.;
xaxis display=(nolabel novalues);
vbox cholesterol / category=AgeAtStart;
run;
 Chapter 9 Annotation and Attribute Maps (SAS 9.3) 239

Figure 9.1.6: Creating Multi-line Axis Tick Values

When tick values contain multiple words, it is sometimes more space-efficient to split the value
into multiple lines. There is a technique using annotation that can help you create these multi-line
values. The key is using the WIDTH column to specify a text width. When a text string reaches
the specified width, the system will automatically try to wrap the string. By combining the correct
width with the “center” value on the JUSTIFY column, you can create center-justified, multi-line
tick values.

xc1 x1space y1space anchor function width justify y1 label

North Carolina datavalue wallpercent top text 15 center -0.5 North Carolina

South Carolina datavalue wallpercent top text 15 center -0.5 South Carolina

data anno;
set sgbook.rainfall (keep=state) end=_last_;
length x1space $ 11 y1space $ 13 anchor $ 6;
rename state=xc1;
retain function "text" x1space "datavalue" y1space "wallpercent" width 15
justify "center" y1 -0.5 anchor "top";
label=state;
output;
if (_last_) then do;
x1space = "wallpercent"; title "Maximum Rainfall Caused by a Tropical Cyclone";
y1space = "graphpercent"; proc sgplot data=rainfall sganno=anno
anchor="bottom"; pad=(bottom=15%) noautolegend;
x1=50; xaxis display=(nolabel novalues);
y1=1; vbar state / response=rainfall dataskin=pressed
label="State"; datalabel;
output; run;
end;
240 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 9.1.7: Creating an Axis-aligned Statistics Table

This graph contains an axis-aligned statistics table of class weight and height. The Y-coordinates
and the table values are extracted from the bar chart’s input data set. After the last value is
processed (_last_ is set), we add the column headers for the table. The PROC SGPLOT call to
make the bar chart uses the new PAD option to provide space for the table.

Name Height Weight x1space y1space anchor label function yc1 x1

Alfred 69.0 112.5 wallpercent datavalue Right 113 text Alfred 111.0

Alfred 69.0 112.5 wallpercent datavalue Right 69 text Alfred 122.0

data anno;
set sashelp.class (obs=5 keep=name weight height) end=_last_;
length x1space $ 11 y1space $ 11 anchor $ 8 label $ 6;
retain function 'text' x1space 'wallpercent' y1space 'datavalue' anchor 'right';
yc1=name; /* Y-coordinate is the name */
x1=111; /* percent beyond the edge of the wall */
label=put(weight, F3.0);
output;
x1=122; /* percent beyond the edge of the wall */
label=put(height, F3.0);
output;
if (_last_) then do; /* Add table headers to the end of the data */
y1space = 'wallpercent';
width = 20;
anchor = 'top';
textweight='bold';
y1 = 103; title "Class Statistics";
x1=107.5; proc sgplot data=sashelp.class (obs=5) sganno=anno
pad=(right=40%);
label="Weight";
yaxis display=(nolabel);
output;
hbar name / response=weight categoryorder=respdesc
x1=119.5; dataskin=pressed nostatlabel;
label="Height"; run;
output;
 Chapter 9 Annotation and Attribute Maps (SAS 9.3) 241

Figure 9.1.8: Creating a Simple Forest Plot with Annotation

This example combines the techniques of data alignment and text width in the previous examples
to create a simple forest plot. Notice how the effect labels along the bottom axis are positioned
based on wall position, and the width keeps long labels from colliding. If you know you are going
to have long effect labels, you may want to specify more bottom padding in the SGPLOT
procedure. Also see Figure 12.2 for an alternate way to create a forest plot without annotation.

Study Summary label anchor y1space x1space function width textsize yc1 x1

Modano 0.590 Modano left datavalue graphpercent text 50 8 Modano 1

(1967) (0.096, (1967) (1967)
3.634)

Modano 0.590 0.590 left datavalue wallpercent text 50 8 Modano 101

(1967) (0.096, (0.096, (1967)
3.634) 3.634)
242 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

/* Annotation for the study column, the stat column, */

/* and the effect labels. */
data anno;
set sgbook.labels end=_last_;
length label $ 50 anchor $ 6 y1space $ 13 x1space $ 12;
retain function "text" y1space "datavalue" width 50 anchor "left" ;
yc1=study;
label=study;
x1space="graphpercent";
x1=1;
output;
label=summary;
x1space="wallpercent";
x1=101;
output;
if (_last_) then do;
/* Add the effect labels */
yc1=" ";
x1space="wallpercent";
y1space="graphpercent";
anchor="bottom";
y1=2;
x1=25;
label="Favors Treatment";
output;
x1=75;
label="Favors Placebo";
output;
/* Add the column headers */
y1space="wallpercent";
x1space="graphpercent";
y1=100;
x1=9;
label="Study";
output;
x1=89;
label="OR (LCL, UCL)";
output;
end; run;
proc sgplot data=sgbook.meta nocycleattrs noautolegend sganno=anno
dattrmap=attrmap pad=(left=22% right=28% bottom=7%);
yaxis reverse display=none;
xaxis type=log logbase=10 min=.01 max=100 minor display=(nolabel)
offsetmin=0 offsetmax=0;
refline 1 / axis=x;
refline .01 .1 10 100 / axis=x lineattrs=GraphGridLines;
highlow y=study high=UpperCL low=LowerCL;
scatter y=study x=oddsratio / attrid=a group=study_type; run;
 Chapter 9 Annotation and Attribute Maps (SAS 9.3) 243

Figure 9.1.9: Adding a Company Logo

You can create a small annotate data set to add a company logo to all of the graphs in your
report, assuming you want the logo in the same location. A typical place to put this logo is in one
of the extreme corners of the graph. The key to using the data set is to specify the proper padding
in the SG procedure to prevent collisions with the plot axes or other graph features.

function anchor x1 y1 width drawspace image

Image bottomright 100 0.5 20 graphpercent logo.png

data anno;
retain function "Image" anchor "bottomright" x1 100 y1 0.5
width 20 drawspace "graphpercent" image "logo.png";
run;

title "Milk Production for 2007";

proc sgplot data=sgbook.milk_production pad=(bottom=5%) sganno=anno;
yaxis discreteorder=data;
hbar month / response=production dataskin=pressed;
run;
 244 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 9.1.10: Background Images in the Graph

In this example, the IMAGESCALE column is used to tile the texture image. The WIDTH and
HEIGHT columns define the size of the area in which to tile the image. The ANCHOR, X1, and Y1
columns determine where the tiling area is positioned. The LAYER column is used to push the
image to the back of the graph, even behind the wall of the plot area.

function anchor x1 y1 width height widthunit heightunit imagescale drawspace transparency layer Image

Image bottomright 100 0.5 640 480 pixel pixel tile graphpercent 0.5 back graywood.jpg

data anno;
retain function "Image" anchor "bottomright" x1 100 y1 0.5 width 640
height 480 widthunit 'pixel' heightunit 'pixel' imagescale 'tile'
drawspace "graphpercent" layer 'back' transparency 0.5
Image "graywood.png";
run;
 Chapter 9 Annotation and Attribute Maps (SAS 9.3) 245

Figure 9.1.11: Images as Curve Labels

In this example, an image is used to label each curve instead of a text label. All images are
positioned just outside of the x-axis data area. The Y position of the images comes from the Y
value of the plot points at the year 2000. The ability to position images in data space also opens
up the possibility of using images as plot points. Note that annotations do not reserve space;
therefore, you must provide an axis offset large enough to contain the images within the wall
area.

In addition to the image labels, the company logo and the footnote text are also annotated. The
text is annotated because the text from a FOOTNOTE statement would have been raised up with
plot when adding padding for the company logo.
246 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

x1space y1space anchor width widthunit function x1 y1 image textsize label

datapercent datavalue Left 40 pixel image 102 77 chicken.jpg .

datapercent datavalue left 40 pixel image 102 67 cow.jpg .

datapercent datavalue left 40 pixel image 102 51 pig.jpg .

graphpercent graphpercent bottomright 90 pixel image 99 1 Logo.png .

graphpercent graphpercent bottomleft 150 pixel text 1 1 Logo.png 6 Source:

USDA

data anno;
length x1space $ 13 y1space $ 13 anchor $ 11;
/* Query for the observation at year 2000 */
set meat_consumption (where=(year='01jan2000'd)) end=_last_;
retain anchor 'left' y1space 'datavalue' x1space 'datapercent'
width 40 widthunit 'pixel' function 'image' x1 102;
y1 = chicken;
image = chicken.jpg";
output; /* Chicken image */
y1 = beef;
image = cow.jpg";
output; /* Cow image */
y1 = pork;
image = pig.jpg";
output; /* Pig image */
if (_last_) then do;
/* The company logo */
x1space = "graphpercent";
y1space = "graphpercent";
anchor = "bottomright";
x1 = 99;
y1 = 1;
width=90;
image = Logo.png";
output;
/* The footnote text */
function = "text";
anchor = "bottomleft";
x1 = 1;
width=150;
textsize = 6;
label = "Source: USDA";
output;
end; run;
 Chapter 9 Annotation and Attribute Maps (SAS 9.3) 247

Figure 9.1.12: Using Arrows to Emphasize Data

In this example, arrows are used to emphasize outlying values in the plot. Notice how a common
X2 and Y2 value are use to draw the arrows to the same point. The single label is added after all
of the data is processed.

y1 x1 x1space y1space function x2space y2space x2 y2 direction scale

$128,420 14 datavalue datavalue Arrow wallpercent wallpercent 50 75 in 0.5

$121,770 15 datavalue datavalue Arrow wallpercent wallpercent 50 75 in 0.5

data anno;
set sgbook.expensive (keep=msrp mpg_city) end=_last_;
length x1space $ 11 y1space $ 11;
retain function "Arrow" x1space "datavalue" y1space "datavalue" x2space "wallpercent"
y2space "wallpercent" x2 50 y2 75 direction "in" scale 0.5;
rename msrp=y1 mpg_city=x1;
mpg_city = mpg_city + 1;
output;
if (_last_) then do;
function = "text";
x1space = "wallpercent";
y1space = "wallpercent";
mpg_city=50; Title "Is Good Gas Mileage Expensive?";
proc sgplot data=sashelp.cars sganno=anno;
Figure 9.1.13: Making a Bubble Legend
msrp=75;
scatter x=mpg_city y=msrp / group=type;
anchor="left";
loess x=mpg_city y=msrp / nomarkers;
width=50;
run;
label="These cars are expensive!";
output; end;
248 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 9.1.13: Making a Bubble Legend

This legend is constructed using four annotation functions: RECTANGLE, OVAL, LINE, and
TEXT. The minimum and maximum size of the annotated bubbles is synchronized with the
bubble plot by using the BRADIUSMIN and BRADIUSMAX options.

drawspace widthunit heightunit linethickness textsize function x1 y1 width height anchor

wallpercent pixel pixel 1 12 Rectangle 86 76.5 140 87 bottom

wallpercent pixel pixel 1 12 Oval 80 76.5 44 44 bottom

data anno;
retain drawspace "wallpercent" widthunit "pixel" heightunit "pixel"
linethickness 1 textsize 8;
length function $ 9;
input function $ x1 y1 width height x2 y2 textsize anchor $ label $ 48-66;
cards;
Rectangle 86 76.5 140 87 . . 12 bottom
Oval 80 76.5 44 44 . . 12 bottom
Oval 80 76.5 16 16 . . 12 bottom
Line 80 87.9 . . 87 87.9 12 bottom
Line 80 80.5 . . 87 80.5 12 bottom
Text 86 98 140 . . . 12 top Salary (in dollars)
Text 87 88.1 140 . . . 8 left $32,816
Text 87 80.4 140 . . . 8 left $18,444; run;
 Chapter 9 Annotation and Attribute Maps (SAS 9.3) 249

Figure 9.1.14: Making a Polyline Figure

In this example, a POLYLINE function is used with other functions to create an annotation
showing the result of two companies combining into one. A group variable is used to color the
bars differently from the rest of the companies.

yc1 drawspace function x1 x2 yc2 anchor width label

Chrysler datavalue polyline 2.5 . .

Chrysler datavalue polycont 3.5 . .

250 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

data anno;
length yc1 $ 15;
retain drawspace "datavalue";
function="polyline";
yc1="Chrysler";
x1=2.5;
output;
function="polycont";
x1=3.5;
output;
yc1="Fiat";
output;
x1=3.0;
output;
function="polyline";
yc1="Suzuki";
x1=3.5;
output;
function="polycont";
x1=6.0;
output;
yc1="Fiat + Chrysler";
output;
function="arrow";
x2=5.0;
yc2="Fiat + Chrysler";
output;
function="text";
yc1="Honda";
x1=6.1;
anchor="Left";
width=30;
label="Alliance creates the #6 global automaker by volume";
output;
run;

proc sgplot data=sgbook.autos noautolegend sganno=anno;

hbarparm category=automaker response=million_units / datalabel
group=colorvar dataskin=pressed;
run;
 Chapter 9 Annotation and Attribute Maps (SAS 9.3) 251

9.2 Attribute Maps (9.3)

Attribute maps are used to assign visual attributes to specific group data values. Like annotations,
this feature uses a data set to define the content of the feature; however, the definition is much
smaller.

Required Column(s):
ID The ID of the attrmap. This value is referenced from the
Char
ATTRID option on the plot statements.
VALUE The data value to be assigned to the attributes. The
keyword _OTHER_ in the column can be used to define
Char the attributes of any values that are not explicitly defined in
the map. Numeric values should represent their final
formatted form.

Optional Column(s):
A style element reference for marker attributes (e.g.
MARKERSTYLE Char
GraphData1, GraphData2, etc.).
Either a literal marker symbol (circle, plus, etc.) or an
MARKERSYMBOL Char
attribute style reference (GraphData1:markersymbol).
Either a literal marker color (red, cxff0000, etc.) or an
MARKERCOLOR Char
attribute style reference (GraphData1:contrastcolor).
A style element reference for line attributes (e.g.
LINESTYLE Char
GraphData1, GraphData2, etc.).
Either a literal line pattern (solid, dash, etc.) or an attribute
LINEPATTERN Char
style reference (GraphData1:linepattern).
Either a literal line color (red, cxff0000, etc.) or an attribute
LINECOLOR Char
style reference (GraphData1:contrastcolor).
A style element reference for fill attributes (e.g.
FILLSTYLE Char
GraphData1, GraphData2, etc.).
Either a literal fill color (red, cxff0000, etc.) or an attribute
FILLCOLOR Char
style reference (GraphData1:color).

You can define multiple attribute maps within the same data set; however, make sure that the
data set stays sorted by the ID column. The ID is referenced from the procedure’s PLOT
statement and the attributes applied to the GROUP variable. If there is not a GROUP variable on
the PLOT statement, the attribute map is ignored. Also note that values in the attribute map are
case-sensitive.

The optional columns are applied to the plot only if the attributes can be applied to the plot.
Otherwise, the extra columns are ignored. If a group value is not found in the map, the attributes
for that group will come from the ODS style.
252 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 9.2.1: Controlling the Appearance of Plot Lines

In this example, the attribute map is defined to control the appearance of both the plot lines and
the markers for each group value.

data attrmap;
length value $ 9 markersymbol $ 14;
retain id 'my_id' linepattern 'solid';
input value $ markersymbol;
cards;
Microsoft circlefille
IBM squarefilled
Intel trianglefilled
;
run;
 Chapter 9 Annotation and Attribute Maps (SAS 9.3) 253

The DATA step produces an attribute map with the following information:

value markersymbol id linepattern

Microsoft circlefille my_id solid

IBM squarefilled my_id solid

Intel trianglefilled my_id solid

The value “my_id” in the ID column is referenced by the ATTRID option in the following code to
identify which map in the data set to use. The values from the GROUP variable are compared to
the map values in the VALUE column to determine if the defined attributes in the attribute
columns should be applied.

ods listing style=journal;

Title1 "Stock prices from 2001-2003";
proc sgplot data=sashelp.stocks dattrmap=attrmap;
where Date >= '01jan2001'd and Date <= '01jan2003'd;
series x=Date y=close / markers group=stock attrid=my_id;
run;
254
Chapter 10
Classification Panels
10.1 Introduction 257
10.2 SGPANEL Procedure 258
10.3 PANELBY Statement 259
10.4 Classification Panels 261
10.5 Paging of Large Panels 266
256 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“I have not failed. I've just found

10,000 ways that won't work.”
~ Thomas Alva Edison
 Chapter 10 Classification Panels 257

Chapter 10: Classification Panels

In this chapter, we step into the world of multi-cell classification panels using the SGPANEL
procedure. These graphs are valuable tools for understanding the data by displaying the data
classified by one or more class variables.

Classification panel graphs allow us to examine the relationship between variables for all
crossings of the class variables. Therefore, we can examine effects of a drug over time classified
by the gender of the person in the study.

10.1 Introduction
In addition to the plots supported by the SGPLOT procedure, the SGPANEL procedure supports
a PANELBY statement. This is the key difference between the two procedures. As mentioned in
Section 2.3, the PANELBY statement allows us to specify the LAYOUT type, and the
classification variables. The graph space is subdivided into smaller “cells” based on the number
of levels for each class variable(s).

Each cell in the graph is populated with a subset of the data that corresponds to the class
variable value(s). The plot statements in the PROC step after the PANELBY statement define the
“prototype” or “rubber stamp” that is used to populate each cell, as shown in Figures 10.1.1 and
10.1.2 below.

title ‘Distribution of Cholesterol’;

proc sgpanel data=sashelp.heart;

panelby sex;

histogram cholesterol;
density cholesterol;

colaxis display=(nolabel);

run;

Figure 10.1.1 Figure 10.1.2

The key features of the example shown in Figures 10.1.1 and 10.1.2 are:
• The PANELBY variable is “Sex”. Since the data has two values for sex, this results in
two cells in the graph.
• Each cell is populated with a HISTOGRAM and a DENSITY statement for the associated
data.
• Options on other statements such as COLAXIS apply to the whole graph.
258 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Note: Each cell essentially contains a graph of the type created by the SGPLOT procedure. So,
all the techniques that we have learned in the previous chapters apply here, with just a few
exceptions.

10.2 SGPANEL Procedure

The SGPANEL procedure supports the following syntax:

Proc SGPANEL data=datasetname <options>;

Options:
CYCLEATTRS boolean Cycle through style elements for each plot
DATA =sas-data-set Optional data set
DESCRIPTION =string Description string
DATTRMAP =sas-data-set Data set defining an attribute map
NOAUTOLEGEND boolean Do not cycle style elements for each plot
NOCYCLEATTRS boolean Do not cycle style elements for each plot
PAD =value Padding around the outside of the graph
SGANNO =sas-data-set Data set containing the annotations
TMPLOUT =string File name for generated graph template code

The procedure supports the following statements:

• PANELBY
• Basic Plots: BAND, NEEDLE, SCATTER, STEP, SERIES, VECTOR,
BUBBLE, HIGHLOW, HBARPARM, VBARPARM
• Fit Plots: LOESS, PBSPLINE, REG, LINEPARM
• Distribution Plots: DENSITY, HISTOGRAM, HBOX, VBOX
• Categorical Plots: DOT, HBAR, HLINE, VBAR, VLINE
• Other: KEYLEGEND, REFLINE, COLAXIS, ROWAXIS

The procedure has the following syntax:

PROC SGPANEL < DATA= data-set > < options >;

PANELBY classvar1 < classvar2 … <classvarN > < / options >;
< plot-statement(s) >;
< refline-statement(s) >;
< axis-statement(s) >;
< keylegend-statement(s) >;
RUN;
 Chapter 10 Classification Panels 259

As seen in the examples in Chapter 3, one or more plot statements can be provided that will work
together to create the graph. See the table of permissible combinations shown in section 2.4.

The syntax and features of some of these plots are very similar to the same statements in the
SGPLOT procedure, with a few deviations related to positioning of data and curve labels. The
syntax for these plot statements has been discussed in the preceding chapters as listed below, so
we need not discuss them again here.
• Basic Plots : Chapter 4
• Fit and Confidence Plots : Chapter 5
• Distribution Plots : Chapter 6
• Categorization Plots : Chapter 7

The following statements are new to the SGPANEL procedure, or they behave differently than in
the SGPLOT procedure. We will discuss these statements in detail in this chapter:
• PANELBY : Defines the paneling structure of the graph
• KEYLEGEND : Customizes the external legend
• REFLINE : Adds common reference lines
• COLAXIS : Customizes the common external column axis
• ROWAXIS : Customizes the common external row axis

Uniform Data Ranges: For a panel graph, the axis ranges are always uniform as follows:
• All vertical axes in any one row always have a uniform data range.
• All horizontal axes in any one column always have a uniform data range.
• Vertical (or horizontal) data ranges may or may not be uniform between rows (or
columns).
• A Y reference line is drawn at the same Y value in all cells of a row.
• An X reference line is drawn at the same X value in all cells of a column.

10.3 PANELBY Statement

The PANELBY statement is required and must be provided before any

other plot, legend, axis, or refline statements. The syntax is:

panelby classvar1 <classvar2...classvarN> </ options>;

The space provided for the graph is subdivided into an even grid of
cells based on the number of crossings for the class variables.
Figure 10.3.1
Four different types of layouts are available. These are panel (default),
lattice, columnlattice and rowlattice as described below.

An example of the lattice layout is shown in figure 10.3.1, where the first class variable is used as
the column variable and the second as the row variable.
260 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Required Data Roles:

column-list classification variable(s)

Options:
COLHEADERPOS =keyword TOP | BOTTOM | BOTH
COLUMNS =value Number of columns
LAYOUT =keyword PANEL, LATTICE, ROWLATTICE, etc.
ROWHEADERPOS =keyword LEFT | RIGHT | BOTH
ROWS =value Number of rows
SPACING =value Spacing between cells
START =keyword TOPLEFT | BOTTOMLEFT
UNISCALE =keyword ALL | COLUMN | ROW

Boolean Options:
BORDER | NOBORDER Displays borders
MISSING Includes missing as a separate level
NOVARNAME Displays only the class value in the header
ONEPANEL Prevents paging of the panel

Layout Types – PROC SGPANEL supports four different layouts. Each layout subdivides the
available graph region into a regular grid of cells. The layout of the cells and the location of the
headers are different for each layout type as shown in Figure 10.3.2 below.

Each layout has its strengths as follows:

• Panel (Default) :Supports multiple class variables and is optimal for sparse data.
• Lattice : Supports column and row variables and similar layout.
• ColumnLattice : Supports one class variable to create a Lattice of columns (one row).
• RowLattice : Supports one class variable to create a aLattice of rows (one column).

Panel
(Multiple class
Variables)

RowLattice
Lattice
(One class var)
(Two Class
Variables)

ColumnLattice
(One class var)
Figure 10.3.2
 Chapter 10 Classification Panels 261

10.4 Classification Panels

Figure 10.4.1: Mileage by Horsepower
and Type

This graph shows mileage by horsepower

for vehicles classified by type. Having
separate cells for each class allows us to
plot the regression and confidence bands.

title 'Vehicle Mileage by Type';

proc sgpanel data=sashelp.cars;
where (type='Sedan' or type='Sports' or type='SUV' or type='Wagon');
panelby type;
reg x=horsepower y=mpg_city / nomarkers cli degree=2;
scatter x=horsepower y=mpg_city / transparency=0.7;
run;

Figure 10.4.2: Mileage by Type

This graph shows the city and highway

mileage for vehicles by type and origin.

The number of columns is set to 3, or else

the procedure will attempt to create a
balanced 2x2 grid.

The reference line at Y=20 is shown in all

the cells. NOVARNAME suppresses the
display of column labels in the cell headers.

title 'Vehicle Mileage';

proc sgpanel data=sashelp.cars;
where type='Sedan' or type='Sports';
panelby origin / novarname columns=3;
vbar type / response=mpg_city stat=mean fillattrs=graphdata7;
vbar type / response=mpg_highway barwidth=0.5 stat=mean
fillattrs=graphdata11;
refline 20;
colaxis grid display=(nolabel); rowaxis grid label='Mileage';
run;
262 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 10.4.3: Panel by Origin and Type

This graph shows a panel with two class

variables, Origin and Type. Each cell has a
cell header for each of the class variables.

The space available for display of the data

is quickly consumed by the cell headers.

This layout (panel) allows multiple class

variables; however, the space is quickly
consumed by the cell headers.

title 'Vehicle Mileage';

proc sgpanel data=sgbook.cars;
where origin='USA' or origin='Asia';
panelby origin type / novarname;
histogram mpgc;
colaxis grid; rowaxis grid;
run;

Figure 10.4.4: Lattice by Origin and Type

This graph also displays a panel with two

class variables, Origin and Type. Here we
have used LAYOUT=Lattice, so Origin is
used as the column variable and Type is
used as the row variable.

Each row and column has a header, with

the value of the class variable. This layout
uses less space for headers.

title 'Vehicle Mileage';

proc sgpanel data=sgbook.cars;
where origin='USA' or origin='Asia';
panelby origin type / layout=lattice novarname;
histogram mpgc;
colaxis grid; rowaxis grid;
run;
 Chapter 10 Classification Panels 263

Figure 10.4.5: Panel by Type

To create a one-column panel of horsepower by type, we

can use the default layout (panel). To ensure one column,
set COLUMN=1. We used option ONEPANEL to prevent
paging of the graph into multiple images.

Each cell gets a cell header at the top, taking up a

significant amount of the available space. See Figure
10.4.6 for an alternative layout.

title 'Vehicle Horsepower';

proc sgpanel data=sashelp.cars;
panelby type / onepanel novarname
columns=1;
histogram horsepower;
colaxis grid; rowaxis grid;
run;

Figure 10.4.6: Row Lattice by Type

In this graph, we used LAYOUT= RowLattice. This layout

creates a lattice of rows (with one column). Each row gets
a row header on one side (similar to the lattice layout).

Sometimes this layout is preferred when there are many

short cells. The cell header goes on the side and results in
an efficient usage of the graph space.

Note: All y-axis data ranges are uniform.

title 'Vehicle Horsepower';

proc sgpanel data=sashelp.cars;
panelby type / layout=rowlattice onepanel
novarname;
histogram horsepower;
colaxis grid; rowaxis grid;
run;
264 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 10.4.7: Panel by Type

This is a graph of the vital signs for a patient in a clinical

study. The vital signs are plotted by day of visit.

The range of values for each vital sign is different, from

blood pressure, to weight (kg), to temperature (C). In such
a use case, it is useful for each row to have its own data
range. To get this, we set UNISCALE=Column.

title 'Vital Signs';

proc sgpanel data=sgbook.vitalsigns;
where vstestcd <> 'HEIGHT';
panelby vstestcd / layout=rowlattice onepanel
novarname uniscale=column;
series x=visitdy y=vsstresn;
colaxis grid; rowaxis grid display=(nolabel);
run;

Figure 10.4.8: Row Lattice by Type

Here we have added the display of the normal ranges for

some of the vital signs. These bands are different for each
cell. This is a common feature needed to indicate if the
measurements are within normal range.

To do this, we have added two new columns in the data

with values for systolic and diastolic pressure. A band plot
is added to display the normal range. If needed, data
driven REFLINEs can also be added.
title 'Vital Signs';
proc sgpanel data=sgbook.vitalsigns;
where vstestcd <> 'HEIGHT';
panelby vstestcd / layout=rowlattice onepanel
novarname uniscale=column;
band x=visitdy lower=low upper=high / fill
nooutline transparency=0.6;
series x=visitdy y=vsstresn;
colaxis grid; rowaxis grid display=(nolabel);
run;
 Chapter 10 Classification Panels 265

Figure 10.4.9: Column Lattice title 'Vehicle Mileage by Origin ...';

proc sgpanel data=sashelp.cars
noautolegend;
This graph shows a column lattice, which where type <> 'Hybrid';
is a lattice of multiple columns in one row. panelby origin / layout=columnlattice
onepanel novarname;
vbar type / response=mpg_city
stat=mean dataskin=gloss;
colaxis grid;
rowaxis grid display=(nolabel);
run; 9.3

Figure 10.4.10: Grouped Bar Chart title 'Vehicle Mileage by Origin ...';
proc sgpanel data=sashelp.cars;
where type <> 'Hybrid';
SG procedures with SAS 9.2 do not panelby origin / onepanel novarname
support adjacent “cluster” grouped bar layout=columnlattice noborder
charts. This graph type can be simulated colheaderpos=bottom;
using the Layout=ColumnLattice option vbar type / response=mpg_city
stat=mean dataskin=gloss group=type;
and other appropriate options as shown colaxis grid display=none
here. offsetmin=0.2 offsetmax=0.2;
rowaxis grid;
SAS 9.3 does support “cluster” grouped run;
9.3
bar charts as discussed earlier.
266 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

10.5 Paging of Large Panels

title 'Vehicle Mileage by Origin ...';
Figure 10.5.1: Layout Panel proc sgpanel data=sashelp.cars;
where (origin='USA' or
This graph is a panel classified by origin origin='Europe') and
and type. There are eight combinations, (type='Sedan' or type='Sports' or
and we have specified a 2x2 grid. type='SUV' or type='Truck');
panelby origin type / columns=2
rows=2;
The procedure creates two graph histogram mpg_city;
outputs to cover all the combinations. run;
Note the Europe-Truck combination is
completely dropped (including headers).

Figure 10.5.2: Layout Lattice title 'Vehicle Mileage by Origin ...’;

proc sgpanel data=sashelp.cars;
where (origin='USA' or
This graph is the same as Figure 10.5.1, origin='Europe') and
except LAYOUT=Lattice. (type='Sedan' or type='Sports' or
type='SUV' or type='Truck');
In this case, all combinations are shown panelby origin type / layout=lattice
(including headers). However, cells with columns=2 rows=2;
missing data are empty. histogram mpg_city; run;
Chapter 11
Comparative and Matrix Plots
11.1 Introduction 269
11.2 SGSCATTER Procedure 269
11.3 PLOT Statement 270
11.4 COMPARE Statement 276
11.5 MATRIX Statement 281
268 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“If you don't know where you are going,

you will wind up somewhere else.”
~ Yogi Berra
 Chapter 11 Comparative and Matrix Plots 269

Chapter 11: Comparative and Matrix Plots

In this chapter, we continue our journey through the world of multi-cell graphs. However, instead
of creating graphs classified by one or more class variables, we will look into grids of plots that
provide quick comparative views of the data.

11.1 Introduction
Comparative and matrix graphs primarily consist of scatter plots of multiple response variables or
measures in the data set. These graphs are especially valuable for understanding the raw data
received from the field. Often the analyst would like to get a feel for the data and find
relationships between variables of a data set using “visual” means. This can be a valuable tool in
the pre-analysis phase of the project.

The SGSCATTER procedure is specifically designed for the creation of grids of scatter plots in
the form of comparative or matrix graphs. Now, you certainly can use the SGSCATTER
procedure to create a single-cell scatter plot with fit curves and confidence bands. However, you
will be better off using the SGPLOT procedure to create the single-cell graph as seen in the many
examples shown in previous chapters. The SGSCATTER procedure is better suited for creation
of grids of scatter plots with a few fit and confidence plots.

11.2 SGSCATTER Procedure

This procedure is very different from the SGPLOT and SGPANEL procedures. The syntax for
this procedure is much more like the syntax for the GPLOT procedure and is as follows:

proc sgscatter <options>;

plot or compare or matrix statement;
run;

Options:
DATA =sas-data-set Optional data set
DESCRIPTION =string Description string
DATTRMAP =sas-data-set Data set defining an attribute map
PAD =value Padding around the outside of the graph
SGANNO =sas-data-set Data set containing the annotations
TMPLOUT =string File name for generated graph template code

The SGSCATTER procedure supports the following three independent statements:

• PLOT – Creates multi-plot panels

• COMPARE – Creates comparative panels
• MATRIX – Creates square scatter plot matrices
270 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

11.3 PLOT Statement

The PLOT statement creates regular grids of scatter plots based on the “plot request”. The
syntax for this statement is as follows:

plot plot-request </ options>;

Plot request is in one of the forms listed below. A regular grid of scatter plots is created based on
the plot request.
• y * x : creates a single scatter plot.
• y1 * x1 y2 * x2 : creates a graph with two scatter plots as defined.
• (y1 y2) * x : creates a graph with two separate scatter plots y1 * x and y2 * x.
• y * (x1 x2) : creates a graph with two separate scatter plots y * x1 and y * x2.
• (y1 y2) * (x1 x2): creates a graph with four separate scatter plots.

Required Plot Request:

Plot Request y * x or (y1 y2) * x or y * (x1 x2) or (y1 y2) * (x1 x2)

Options:
ATTRID =variable Associated Attr Map for visual attributes
COLUMNS =value Number of columns in panel
DATALABEL =variable Variable for labeling of scatter points
ELLIPSE <=(options) Fit a confidence ellipse to the data
GROUP = variable Grouping variable
JOIN <=options> Connect the data points
LEGEND = (options) Controls legend characteristics
LOESS = (options) Fit a Loess curve to the data
MARKERATTRS =marker-attrs Specify marker attributes
PBSPLINE =(options) Fit a penalized B-Spline to the data
REG =(options) Fit a regression line to the data
ROWS =value Number of rows in panel
SPACING =value Spacing between rows and columns
TRANSPARENCY =value Transparency value
UNISCALE =keyword Uniform scaling for X | Y | ALL. Default is NONE

Boolean Options:
GRID Display borders
NOLEGEND Suppress display of legend
REFTICK Display duplicate reference tick marks
 Chapter 11 Comparative and Matrix Plots 271

Figure 11.3.1: Basic Scatter Plots

This is a basic two-cell scatter plot of height

by weight and cholesterol by age at start.
Each plot is specified as a separate request.

Each cell has its own set of axes with

independent data ranges.

title "Blood Pressure by Age";

proc sgscatter data=sashelp.heart;
plot height*weight cholesterol*ageatstart;
run;

Figure 11.3.2: Scatter Plots with

Attributes

This is the same graph as shown in Figure

11.3.1 with the addition of marker attributes
and transparency.

The high transparency allows us to see

where the observations are clustered.

title "Blood Pressure by Age";

proc sgscatter data=sashelp.heart;
plot height*weight cholesterol*ageatstart/
markerattrs=(symbol=circlefilled) transparency=0.96;
run;
272 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 11.3.3: Plot Grid - 1 Row x 2

Columns

This graph displays diastolic and systolic

blood pressure by age. Gridlines have been
enabled.

Notice the data range for the two y-axes is

not uniform.

title " Patient Blood Pressure Profile";

proc sgscatter data=sashelp.heart;
where sex='Male';
plot (diastolic systolic) * ageatstart / grid
markerattrs=(symbol=circlefilled) transparency=0.96;
run;

Figure 11.3.4: Plot Grid - 1 Row x 2

Columns

This graph is the same as in Figure 11.3.3.

We have UNISCALE=Y to make all the y-
axes have the same scale visually showing
the relative values of the measures.

Each cell still has its own x and y-axes.

title " Patient Blood Pressure Profile";

proc sgscatter data=sashelp.heart;
where sex='Male';
plot (diastolic systolic) * ageatstart / grid uniscale=y
markerattrs=(symbol=circlefilled) transparency=0.96;
run;
 Chapter 11 Comparative and Matrix Plots 273

Figure 11.3.5: Plot with Fit

This is the same graph as in Figure 11.3.3,

with an added regression fit and confidence
band.

The transparency of the markers is

increased to make the fit line and bands
easier to see.

title "Male Patients Blood Pressure Profile";

proc sgscatter data=sashelp.heart;
where sex='Male';
plot (diastolic systolic) * ageatstart / grid transparency=0.98
markerattrs=(symbol=circlefilled) uniscale=Y reg=(degree=2 cli);
run;

Figure 11.3.6: Plot with Fit and Ellipse

This graph uses the penalized B-spline fit

and a confidence ellipse.

The transparency of the markers is

increased to make the fit line and ellipse
easier to see.

title "Male Patients Vital Signs Profile";

proc sgscatter data=sashelp.heart;
where sex='Male';
plot (diastolic systolic) * ageatstart / grid
pbspline=(clm degree=1) ellipse uniscale=Y
markerattrs=(symbol=circlefilled) transparency=0.98;
run;
274 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 11.3.7: Multiple Plot Request title "Male Patients Vital Signs
Profile";
proc sgscatter data=sashelp.heart;
This graph shows diastolic and systolic where sex='Male';
blood pressure by age and cholesterol by plot (diastolic systolic) *
weight in one row per the plot request. ageatstart cholesterol*weight
Regression fit is shown. / grid Reg=(cli degree=2)
markerattrs=(symbol=circlefilled)
The transparency of the markers is transparency=0.98 rows=1;
increased to make the fit line and bands run;
easier to see.

Figure 11.3.8: Multiple Plot Request title "Male Patients Vital Signs
Profile";
proc sgscatter data=sashelp.heart;
This graph shows diastolic and systolic where sex='Male';
blood pressure and “Cholesterol by plot (diastolic systolic cholesterol)
Age” and “Age at Death” by “Age at * ageatstart ageatdeath * ageatstart
Start”. A penalized B-spline fit is shown. / grid pbspline=(clm degree=1)
markerattrs=(symbol=circlefilled)
Marker transparency is set to make the transparency=0.98 rows=1 ellipse;
fit line and ellipses easier to see. run;
 Chapter 11 Comparative and Matrix Plots 275

Figure 11.3.9: Uniform Side-by-Side Plots

This is a graph of the response over time for

“Drug A” for female and male patients.

We have used the JOIN, UNISCALE and

GRID options.

title " Response over Time for Drug A";

proc sgscatter data=sgbook.seriesGroup2; 9.3
where drug='A';
plot (female male)*date / grid uniscale=Y join;
run;

Figure 11.3.10: Uniform Side-by-Side

Plots

This is a graph of the response over time by

treatment for female and male patients.

We have used the JOIN, UNISCALE and

GRID options. Markers are suppressed by
setting size=0.

title "Response by Treatment over Time";

9.3
proc sgscatter data=sgbook.seriesGroup2;
plot (female male)*date / group=drug grid uniscale=Y join
markerattrs=(size=0);
run;
276 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

11.4 COMPARE Statement

The COMPARE statement creates regular grids of scatter plots. Fit lines and ellipses can be
included. The graphs always have common row and column axis. By default, individual columns
have uniform data ranges for the x-axis. Individual rows have uniform data range for the y-axis.

compare x=(var <var> ...) y=(var <var> ...) </ options>;

Required Plot Request:

X =var-list List of variables for columns
Y =var-list List of variables for rows

Options:
ATTRID =variable Associated Attr Map for visual attributes
DATALABEL =variable Variable for labeling of scatter points
ELLIPSE <=(options) Fit a confidence ellipse to the data
GROUP = variable Grouping variable
JOIN <=options> Connect the data points
LEGEND = (options) Controls legend characteristics
LOESS = (options) Fit a Loess curve to the data
MARKERATTRS =attrs Specify marker attributes
PBSPLINE =(options) Fit a penalized B-spline to the data
REG =(options) Fit a regression line to the data
SPACING =value Spacing between rows and columns
TRANSPARENCY =value Transparency value

Boolean Options:
GRID Display grid lines
NOLEGEND Suppress display of legend
REFTICK Display duplicate reference tick marks
 Chapter 11 Comparative and Matrix Plots 277

Figure 11.4.1: Comparative Graph

This is a comparative graph of vehicle by

city and highway mileage by horsepower.
The x-axis is automatically made common,
so it is easier to compare the data in the
cells with uniform scale.

Gridlines are enabled for ease of

comparison. The JournalSmallFont2 style
has smaller axis label fonts that are needed
for the y-axis labels to fit.

title "Mileage by Horsepower";

proc sgscatter data=sashelp.cars;
compare y=(mpg_city mpg_highway) x= horsepower / grid;
run;

Figure 11.4.2: Comparative Graph

This is a comparative graph showing city

mileage by horsepower and weight.
Common y-axis reduces clutter and
provides more space to display the data.

The x-axis is not uniform. Options are used

for filled markers and transparency.

title "Mileage by Horsepower and Weight";

proc sgscatter data=sashelp.cars;
compare y=(mpg_city) x= (horsepower weight) / grid
markerattrs=(symbol=circlefilled) transparency=0.9;
run;
278 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 11.4.3: Grouped Comparative Plot

This is a comparative graph showing city

mileage by horsepower and weight and
grouped by country of origin.

Common y-axis reduces clutter and provides

more space to display the data. The x-axis
is not uniform.

Options are used for transparency and

marker size.

title "Mileage for Sedans by Origin";

proc sgscatter data=sashelp.cars;
where type='Sedan';
compare y=(mpg_city) x= (horsepower weight) / group=origin grid
transparency=0.9 markerattrs=(size=9);
run;

Figure 11.4.4: 2x2 Grouped ComparativeGraph

This is a comparative plot of city and highway

mileage by horsepower and weight. Common row
and column axes are used by default, so individual
row and column ranges are uniform.

The quadratic regression fit is not by group

(NOGROUP).

title "Profile for Sedans by Origin";

proc sgscatter data=sashelp.cars;
where type='Sedan';
compare y=(mpg_city mpg_highway)
x= (horsepower weight) /
group=origin grid
transparency=0.9
markerattrs=(size=9)
reg=(nogroup degree=2);
run;
 Chapter 11 Comparative and Matrix Plots 279

Figure 11.4.5: Comparative Graph

This is a comparative graph of male and

female response over time for “Drug A”.
Since two Y variables and one X variable
are provided, we get a single column with
two rows.

Common x-axis is displayed with grids.

title " Response over Time for Drug A";

proc sgscatter data=sgbook.seriesGroup2; 9.3
where drug='A';
compare y=(female male) x=date / grid join;
run;

Figure 11.4.6: Comparative Graph

This is a comparative graph of systolic

blood pressure by weight and age grouped
by sex for patients who died of coronary
heart failure.

Linear regression fit lines by sex are

displayed. Transparency is used to
unclutter the display.

Common y-axis is displayed with grids.

title "Blood Pressure by Weight and Age for CHD";

proc sgscatter data=sashelp.heart;
where deathcause = 'Coronary Heart Disease';
compare y=systolic x=(weight ageatstart) / group=sex
grid reg transparency=0.9;
run;
280 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 11.4.7: Heat Map

This is heat map showing the frequency of

occurrences for patients who died of
coronary heart disease (CHD) by
cholesterol status, blood pressure status,
sex, and weight status.

Darker squares represent higher

occurrence. This graph shows highest
levels of CHD deaths for overweight males
with high blood pressure and cholesterol.

title "Heat Map for CHD Deaths";

proc sgscatter data=sgbook.HeatMap;
label chol_status='Cholesterol';
label bp_status='BP';
label weight_status='Weight';
compare y=(chol_Status bp_Status) x=(sex Weight_Status) / grid
transparency=0.995 markerattrs=(symbol=squarefilled size=30);
run;

Figure 11.4.8: Rectangular Matrix

This is a rectangular matrix of scatter plots

with confidence ellipse showing the
relationship of age and weight by systolic
and diastolic blood pressure and cholesterol
for CHD patients.

Some correlation between weight and blood

pressure can be visually discerned.

title "Rectangular Scatter Matrix for CHD";

proc sgscatter data=sgbook.HeatMap;
compare x=(systolic diastolic cholesterol) y=(weight ageatdeath) /
markerattrs=(symbol=circlefilled size=10) ellipse
grid transparency=0.95;
run;
 Chapter 11 Comparative and Matrix Plots 281

11.5 MATRIX Statement

The MATRIX statement creates regular grids of scatter plots of pair-wise combinations of all the
variables in the list. Diagonals are used to label the variable name. Diagonals can alternatively
be used to display the distribution of each individual variable itself:

matrix var var <var> <var> ... </ options>;

Required Plot Request:

Plot Request Var list

Options:
ATTRID =variable Associated Attr Map for visual attributes
DATALABEL =variable Variable for labeling of scatter points
ELLIPSE <=(options) Fit a confidence ellipse to the data
GROUP =variable Grouping variable
JOIN <=options> Connect the data points
LEGEND =(options) Controls legend characteristics
MARKERATTRS =attrs Specify marker attributes
START =keyword Specify the starting position
– TOPLEFT | BOTTOMLEFT
TRANSPARENCY =value Specify the transparency for the markers

Boolean Options:
NOLEGEND Suppress display of legend
282 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 11.5.1: Basic Scatter Plot Matrix

This is a scatter plot matrix of “Weight” x

“Cholesterol” x “Systolic” for female patients
with coronary heart disease.

A scatter plot is drawn for each pair-wise

combination of the variables specified. The
diagonals display the variable name.

title "Scatter Plot Matrix for CHD Patients";

proc sgscatter data=sashelp.heart;
where sex='Female';
matrix weight cholesterol systolic;
run;

Figure 11.5.2: Basic Scatter Plot Matrix

This is a scatter plot matrix of “Weight” x

“Cholesterol” x “Systolic” for female patients
with coronary heart disease.

A filled marker is used with a transparency

of 80%. This provides us a better view of
where the data is clustered.

title "Scatter Plot Matrix for Female CHD Patients";

proc sgscatter data=sgbook.chd;
where sex='Female';
matrix weight cholesterol systolic / transparency=0.8
markerattrs=(symbol=circlefilled);
run;
 Chapter 11 Comparative and Matrix Plots 283

Figure 11.5.3: Scatter Plot Matrix with

Histograms

This is a scatter plot matrix of “Weight” x

“Cholesterol” x “Systolic” with a histogram of
each individual variable shown along the
diagonal.

Variable labels are moved out to the top and

side.

title "Scatter Plot Matrix for Female CHD Patients";

proc sgscatter data=sgbook.chd;
where sex='Female';
matrix weight cholesterol systolic / transparency=0.6
markerattrs=(symbol=circlefilled) diagonal=(histogram);
run;

Figure 11.5.4: Scatter Plot Matrix with

Diagonals and Ellipse

This is a scatter plot matrix of “Weight” x

“Cholesterol” x “Systolic” with histogram and
density plots of each individual variable
shown along the diagonal.

A confidence ellipse is shown for each

scatter plot.

title "Scatter Plot Matrix for Female CHD Patients";

proc sgscatter data=sgbook.chd;
where sex='Female';
matrix weight cholesterol systolic / transparency=0.6 ellipse
markerattrs=(symbol=circlefilled)
diagonal=(histogram normal kernel);
run;
284 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Figure 11.5.5: Grouped Scatter Plot

Matrix

This is a scatter plot matrix of “MPG” x

“Horsepower” x “Weight”. Country of origin
is set as the group variable.

Ellipses are clipped to fit the data, and a

legend is displayed.

title "Scatter Plot Matrix Trucks";

proc sgscatter data=sashelp.cars;
where type='Truck';
matrix mpg_city horsepower weight / group=origin ellipse
transparency=0.5;
run;

Figure 11.5.6: 4x4 Scatter Plot Matrix

This is a 4x4 scatter plot matrix of “MPG” x

“Horsepower” x “Weight” x “MSRP”.

The START option is set to BOTTOMLEFT,

so the matrix is laid out bottom to top. The
previous examples all had the matrix laid
out from top down.

title "Scatter Plot Matrix Trucks";

proc sgscatter data=sashelp.cars;
where type='Truck';
matrix mpg_city horsepower weight msrp/ start=bottomleft
transparency=0.5;
run;
Chapter 12
Health and Life Sciences Graphs
12.1 Introduction 287
12.2 Forest Plot 288
12.3 Forest Plot (9.3) 289
12.4 Survival Plot 290
12.5 Adverse Event Timeline 291
12.6 Adverse Event Timeline (9.3) 292
12.7 Maximum LFT Values by Treatment (9.3) 293
12.8 Median of Lipid Profile over Time by Treatment (9.3) 294
12.9 QTc Change from Baseline over Time by Treatment (9.3) 295
12.10 QTc Change Graph with Annotated “At Risk” Values (9.3) 296
12.11 QTc Change from Baseline over Time by Treatment 297
12.12 LFT Safety Panel, Baseline vs. Study 298
12.13 Immunology Profile by Treatment 299
12.14 Most Frequent On-Therapy Adverse Events by Frequency 300
12.15 LFT Patient Profile 301
12.16 Panel of LFT Values 302
12.17 Distribution of Eye Irritation Using PROC SGPANEL (9.3) 303
12.18 Distribution of Eye Irritation Using PROC SGPLOT (9.3) 304
12.19 Vital Signs by Time Point Name 305
12.20 Concomitant Medications 306
12.21 Creating a 2 x 2 Cell Graph Using PROC SGPLOT 307
286 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“We didn't lose the game; we just ran out of time.”

~ Vince Lombardi
 Chapter 12 Health and Life Sciences Graphs 287

Chapter 12: Health and Life Sciences Graphs

12.1 Introduction
The health and life sciences industry is an ideal candidate for use of graphs. Data is collected for
many different applications such as clinical trials, demographics, vital signs, laboratory test results
for chemistry and hematology, patient profiles, and adverse events. Such data is initially
collected in tabular form, and often consumed in the same form.

Evaluation of such data can be significantly enhanced by the use of graphical displays of the data
in the form of dot plots, box plots, and lattice and matrix displays. Graphical display of data
provides insights into trends and correlations that are just not possible through tabular data.
Often, combining the data and summary statistics on a single display allows for improved analysis
and interpretation of the results.

The examples include displays of lab results over time, distribution of tests by treatment, lattice
and matrix displays of liver function tests (LFT), patient profiles, adverse event plots sorted by
relative risk, hazard function plots, and displays for evaluation of blood chemistry and
hematology.

SG procedures are very well suited for the creation of such displays using the techniques
described in the previous chapters of this book. In this chapter, we will examine how you can
create some common graphs used in this industry. Some examples are shown in Figure 12.1.
Detailed code for creating such graphs is shown in the following pages.

Figure 12.1
288 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

12.2 Forest Plot

This graph shows a common version of the forest plot, showing the results from different studies
and an overall observation. The data is as shown in the table below. The key features are:
• Plot the graph on the X-Y axis (numeric), restricted to the lower 65% of the x-axis.
• Plot the data on the X2-Y axis (character), restricted to the upper 30% of the X2 axis.
• Use multiple scatter plots with MARKERCHAR option to display the statistics.
Study OddsRatio LowerCL UpperCL Weight Q1 Q3 ObsId study2 lcl2 ucl2 OR LCL UCL WT
Modano (1967) 0.590 0.096 3.634 5% 0.56 0.62 1 0.096 3.634 OR LCL UCL Weight
Borodan (1981) 0.464 0.201 1.074 18% 0.38 0.55 2 0.201 1.074 OR LCL UCL Weight
0.328 0.233 0.462 . . . 16 Overall . . OR LCL UCL Weight

title "Impact of Treatment on Mortality by Study";

title2 h=8pt 'Odds Ratio and 95% CL';
proc sgplot data=sgbook.forest noautolegend;
scatter y=study2 x=oddsratio / markerattrs=(symbol=diamondfilled size=10);
scatter y=study x=oddsratio / xerrorupper=ucl2 xerrorlower=lcl2
markerattrs=(symbol=squarefilled);
scatter y=study x=or / markerchar=oddsratio x2axis;
scatter y=study x=lcl / markerchar=lowercl x2axis;
scatter y=study x=ucl / markerchar=uppercl x2axis;
scatter y=study x=wt / markerchar=weight x2axis;
refline 1 100 / axis=x;
refline 0.01 0.1 10 / axis=x lineattrs=(pattern=shortdash) transparency=0.5;
inset ' Favors Treatment' / position=bottomleft;
inset 'Favors Placebo' / position=bottom;
xaxis type=log offsetmin=0 offsetmax=0.35 min=0.01 max=100 minor display=(nolabel) ;
x2axis offsetmin=0.7 display=(noticks nolabel);
yaxis display=(noticks nolabel) offsetmin=0.1 offsetmax=0.1;
run;
 Chapter 12 Health and Life Sciences Graphs 289

12.3 Forest Plot (9.3)

This graph shows the same forest plot from Figure 12.2 using the HighLow plot available with
SAS 9.3. The data is as shown below. The key features of this plot are:
• A HighLow plot with TYPE=Bar is used to display of the weights of the study.
• A HighLow plot with TYPE=Line (Default) is used to display the confidence limits.

Study OddsRatio LowerCL UpperCL Weight Q1 Q3 ObsId study2 lcl2 ucl2 OR LCL UCL WT
Modano (1967) 0.590 0.096 3.634 5% 0.56 0.62 1 0.096 3.634 OR LCL UCL Weight
Borodan (1981) 0.464 0.201 1.074 18% 0.38 0.55 2 0.201 1.074 OR LCL UCL Weight
0.328 0.233 0.462 . . . 16 Overall . . OR LCL UCL Weight

title "Impact of Treatment on Mortality by Study";

title2 h=8pt 'Odds Ratio and 95% CL';
proc sgplot data=sgbook.forest2 noautolegend nocycleattrs;
scatter y=study2 x=oddsratio / markerattrs=(symbol=diamondfilled);
highlow y=study low=lcl2 high=ucl2 / type=line;
highlow y=study low=q1 high=q3 / type=bar;
scatter y=study x=or / markerchar=oddsratio x2axis;
scatter y=study x=lcl / markerchar=lowercl x2axis;
scatter y=study x=ucl / markerchar=uppercl x2axis;
scatter y=study x=wt / markerchar=weight x2axis;
refline 1 100 / axis=x;
refline 0.01 0.1 10 / axis=x lineattrs=(pattern=shortdash) transparency=0.5;
inset ' Favors Treatment' / position=bottomleft;
inset 'Favors Placebo' / position=bottom;
xaxis type=log offsetmin=0 offsetmax=0.35 min=0.01 max=100 minor display=(nolabel) ;
x2axis offsetmin=0.7 display=(noticks nolabel);
yaxis display=(noticks nolabel) offsetmin=0.1 offsetmax=0.2 reverse;
run;
290 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

12.4 Survival Plot

This graph shows the survival estimates for multiple groups of patients by severity of disease.
The data is obtained using the “ods output” from the “SurvivalPlot” object of Sample 49.2.1 of the
LIFETEST procedure as shown below. The key features of the graph are:
• Plot survival curves on the X-Y axis (numeric), restricted to the upper 75% of the Y axis.
• Plot at risk data on the X-Y2 axis (character), restricted to the lower 17% of the Y2 axis.
• Survival curves are labeled using curve labels for clarity. Legend works for color style.
Obs Time Survival AtRisk Event Censored tAtRisk Stratum StratumNum
1 0 1.00000 38 0 . . 1: ALL 1
2 0 . 38 . . 0 1: ALL 1

ods output Survivalplot=sgbook.SurvivalPlot49_2_1;

proc lifetest data=BMT plots=survival(atrisk=0 to 2500 by 500);
time T * Status(0);
strata Group / test=logrank adjust=sidak; run;
proc format; value aml 3 = 'Low' 2 = 'High' 1 = 'All'; run;

title 'Product-Limit Survival Estimates';

title2 h=7pt 'With Number of AML Subjects at Risk';
proc sgplot data=sgbook.SurvivalPlot49_2_1;
format stratumNum aml.;
step x=time y=survival / group=stratumNum curvelabel lineattrs=(pattern=solid);
scatter x=time y=censored / markerattrs=(symbol=plus) name='censored';
scatter x=tatrisk y=stratumnum / markerchar=atrisk y2axis group=stratumnum;
keylegend 'censored' / location=inside position=topright;
refline 0;
yaxis offsetmin=0.2 min=0;
y2axis offsetmax=0.83 display=(nolabel noticks) valueattrs=(size=8);
run;
 Chapter 12 Health and Life Sciences Graphs 291

12.5 Adverse Event Timeline

This graph shows adverse events by time and severity. The data is extracted from Clinical Data
Interchange Standards Consortium (CDISC) data and is as shown in the table below. The key
features of the graph are:
• Vector plots are used to display the events.
• Scatter plots are used to show the start and end events.
• Scatter plot with MARKERCHAR option is used to display the event name.
• Note: X position (xc) for the label needs to be computed. See detailed program.
Obs aestdate aeendate aeseq aedecod aesev y xc aestdy aeendy startday endday
1 06MAR13 06MAR13 1 DUMMY MILD -100 -4.4 0 0 0 0
2 06MAR13 06MAR13 1 DUMMY MODERATE -100 -4.4 0 0 0 0
3 06MAR13 06MAR13 1 DUMMY SEVERE -100 -4.4 0 0 0 0
4 06MAR13 06MAR13 1 DIZZINESS MODERATE 1 -6.6 0 0 0 0
5 20MAR13 . 2 COUGH MILD 2 9.57 14 . 14 104

title "Adverse Events for Patient Id = xx-xxx-xxxx";

proc sgplot data=sgbook.ae1 noautolegend nocycleattrs;
refline 0 / axis=x lineattrs=(thickness=1 color=black);
vector x=endday y=y / xorigin=startday yorigin=y noarrowheads
lineattrs=(thickness=9px);
vector x=endday y=y / xorigin=startday yorigin=y group=aesev noarrowheads
lineattrs=(thickness=7px pattern=solid) transparency=0 name='sev';
scatter x=aestdy y=y / markerattrs=(size=13px symbol=trianglefilled);
scatter x=aestdy y=y / markerattrs=(size=9px symbol=trianglefilled) group=aesev;
scatter x=aeendy y=y / markerattrs=(size=13px symbol=trianglefilled);
scatter x=aeendy y=y / markerattrs=(size=9px symbol=trianglefilled) group=aesev;
scatter x=xc y=y / markerchar=aedecod;
scatter x=aestdate y=y / markerattrs=(size=0) x2axis;
yaxis display=(nolabel noticks novalues) min=0;
xaxis grid label='Study Days' offsetmin=0.02 offsetmax=0.02
values=(&minday10 to &maxday by 2);
x2axis notimesplit display=(nolabel) offsetmin=0.02 offsetmax=0.02
values=(&mindate10 to &maxdate);
keylegend 'sev'/ title='Severity :';
run;
292 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

12.6 Adverse Event Timeline (9.3)

This graph shows adverse events by time and severity. The data is extracted from CDISC data
and is as shown in the table below. The key features for the graph are:
• The HighLow option is used to display the events.
• The LOWLABEL option is used to display the event name.
• Low and high caps are used to display events that are continuing.

aestdate aeendate aeseq aedecod aesev y aestdy aeendy stday enday lcap hcap xs
06MAR13 06MAR13 . MILD -9 0 0 0 0 0
06MAR13 06MAR13 . MODERATE -9 0 0 0 0 0
06MAR13 06MAR13 . SEVERE -9 0 0 0 0 0
06MAR13 06MAR13 1 DIZZINESS MODERATE 1 0 0 0 0 0
20MAR13 . 2 COUGH MILD 2 14 . 14 104 ARROW 0

title "Adverse Events for Patient Id = xx-xxx-xxxx";

proc sgplot data=sgbook.ae2 noautolegend nocycleattrs;
highlow y=y low=xs high=xs / group=aesev type=bar barwidth=0.8 name='sev' Y2axis;
highlow y=aeseq low=stday high=enday / group=aesev lowlabel=aedecod type=bar
barwidth=0.8 lowcap=lcap highcap=hcap lineattrs=(pattern=solid color=black);
scatter x=aestdate y=aeseq / markerattrs=(size=0) x2axis;
refline 0 / axis=x lineattrs=(thickness=1 color=black);
yaxis display=(nolabel noticks novalues) type=discrete;
y2axis display=none min=0;
xaxis grid label='Study Days' offsetmin=0.02 offsetmax=0.02
values=(&minday10 to &maxday by 2);
x2axis notimesplit display=(nolabel) offsetmin=0.02 offsetmax=0.02
values=(&mindate10 to &maxdate);
keylegend 'sev'/ title='Severity :'; run;
 Chapter 12 Health and Life Sciences Graphs 293

12.7 Maximum LFT Values by Treatment (9.3)

This graph shows the “Distribution of Maximum Liver Function Obs Test Drug Value
Test Values by Treatment” using a grouped box plot on a discrete 1 ALAT A 1.05198
axis. The data is as shown on the right. 2 ALAT B 0.97755
3 ASAT A 0.78177
• Use a box plot grouped by “Drug”.
4 ASAT B 0.59554
• Use a reference line to display the clinical concern levels. 5 ALKPH A 0.20475

title h=10pt 'Distribution of Maximum Liver Function Test Values ...';

footnote1 h=8pt j=left "For ALAT, ASAT and ALKPH, the CCL is 2 ULN;";
footnote2 h=8pt j=left "For BILTOT, the CCL is 1.5 ULN: where ULN is the Upper Level of
Normal Range";
proc sgplot data=sgbook.LFT_Group;
format drug $drug.;
vbox value / category=test group=drug nofill lineattrs=(pattern=solid)
medianattrs=(pattern=solid);
keylegend / location=inside position=topright across=1;
refline 1 1.5 2 / lineattrs=(pattern=dot);
xaxis discreteorder=data display=(nolabel);
yaxis label='Maximum (/ULN)';
run;
294 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

12.8 Median of Lipid Profile over Time by Treatment (9.3)

In this graph we are displaying the lipid profile over time by

X Drug Median Lcl Ucl
treatment group. Here again, we have used the “Y-Y2 Split”
technique to split the graph region into two parts: Baseline Drug A 5.21 5.04 5.52
Baseline Drug B 5.17 4.94 5.47
• The upper region is used to display the graph. Baseline Drug C 5.24 4.97 5.33
Baseline Placebo 5.08 4.81 5.35
• The lower region is used to display the statistics.
Day 14 Drug A 4.90 4.60 5.79
• A cluster grouped scatter plot is used for the values.
Day 14 Drug B 6.65 4.81 7.51
• A cluster grouped series plot is used for the lines.
• The cluster width is reduced to tighten the clusters.
• A scatter plot with y=Drug and MARKERCHAR option is used to display the statistics.
• A reference line on the Y2 axis, with DISCRETEOFFSET, is used to separate the two
regions.

title 'Median of Lipid Profile over Time by Treatment';

proc sgplot data=sgbook.lipid_grp;
where by='Test 1';
scatter x=x y=median / group=drug yerrorupper=ucl yerrorlower=lcl
groupdisplay=cluster clusterwidth=0.5 markerattrs=(size=9);
series x=x y=median / group=drug groupdisplay=cluster clusterwidth=0.5;
scatter x=x y=drug / markerchar=median y2axis;
refline 'Placebo' / discreteoffset=0.5 axis=y2;
xaxis display=(nolabel);
yaxis grid offsetmin=0.30;
y2axis offsetmax=0.8 display=(noticks nolabel) valueattrs=(size=7);
run;
 Chapter 12 Health and Life Sciences Graphs 295

12.9 QTc Change from Baseline over Time by Treatment (9.3)

This graph displays the change in corrected QT (QTc) Week Week2 atRiskA atRiskB valueA valueB
interval over time for heartbeats by treatment. The 1 0.75 209 . 72.1929 .
1 0.75 209 . -3.1966 .
visits are irregularly spaced, with higher frequency in 1 1.25 . 405 . -29.5119
the beginning, so it is desirable to have a linear axis 1 1.25 . 405 . -10.0056
for time. This graph has the following features: 2 1.75 206 . 52.2286 .
2 1.75 206 . 57.0051 .
• An overlay of two VBOX statements is used for the data.
• The x-axis values are jittered by 0.25 for each treatment to get adjacent overlays.
• A user-defined format is used to label the last value as “Max”.
• A reference line is added to separate the last value.
• Ideally, we would like to display the “atRisk” values for each week of the study. To do so,
we will use the Annotation feature for SGPLOT as shown in Figure 12.10.

title 'QTc Change from Baseline by Week';

proc sgplot data=sgbook.QtcChange;
vbox valueA / category=Week2 lineattrs=(pattern=solid) meanattrs=graphdata1
whiskerattrs=(pattern=solid) medianattrs=(pattern=solid);
vbox valueB / category=Week2 lineattrs=(pattern=solid) meanattrs=graphdata2
whiskerattrs=(pattern=solid) medianattrs=(pattern=solid);
refline 26 / axis=x;
xaxis type=linear values=(1 2 4 8 12 16 20 24 28) display=(nolabel)
tickvalueformat=myWeek. offsetmin=0.03 offsetmax=0.03;
yaxis values=(-120 to 90 by 30) grid label=’QTc Change’;
run;
296 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

12.10 QTc Change Graph with Annotated “At Risk” Values (9.3)
Here we have added the “At Risk” table to the graph shown in Figure 12.9 using the Annotation.
• An annotation data set is created to draw the at-risk values for “Drug A” and “Drug B”.
• The X coordinate is in DataValue space so they line up with the X axis values.
• The Y coordinate is in WallPixel space, 62 and 74 pixels below the y-axis.
• Annotations are also created for the labels and the “At Risk” label.
• Space is created for the annotation by adding a bottom pad of 80 pixels.

X1Space Y1Space Label Function Anchor TextSize X1 Y1

DataValue WallPixel 209 Text Center 8 1 -62
DataValue WallPixel 405 Text Center 8 1 -74
WallPercent WallPixel At Risk : Text Right 8 -1 -50
WallPercent WallPixel Drug A : Text Right 8 -1 -62
WallPercent WallPixel Drug B : Text Right 8 -1 -74

title 'QTc Change from Baseline by Week';

proc sgplot data=sgbook.QtcChange sganno=sgbook.QTcAnno pad=(bottom=80);
vbox valueA / category=Week2 lineattrs=(pattern=solid) meanattrs=graphdata1
whiskerattrs=(pattern=solid) medianattrs=(pattern=solid);
vbox valueB / category=Week2 lineattrs=(pattern=solid) meanattrs=graphdata2
whiskerattrs=(pattern=solid) medianattrs=(pattern=solid);
refline 26 / axis=x;
xaxis type=linear values=(1 2 4 8 12 16 20 24 28) display=(nolabel)
tickvalueformat=myWeek. offsetmin=0.03 offsetmax=0.03;
yaxis values=(-120 to 90 by 30) grid label='QTc Change';
run;
 Chapter 12 Health and Life Sciences Graphs 297

12.11 QTc Change from Baseline over Time by Treatment

This graph displays the hazard function over time by treatment. The features of the graph are:
• The Y-Y2 Split technique is used to display the day value Drug Low High Risk DrugLabel
axis aligned at risk table at the bottom. 10 .0005 Drug A .0001 .0009 212
• We could have displayed the Y2 labels, but that 10 .0078 Drug B .0070 .0086 428
190 .0030 Drug B .0000 .0060 19 Drug A
takes up extra space. We want all labels on 190 .0030 Drug B .0000 .0060 19 Drug B
one side.
• So, we used the REFLINE statement, with “DrugLabel” as the variable. This has only two
values, and these are displayed. The lines are suppressed by setting thickness=0.
• The separator refline uses a value=-0.0005. This is data dependent. It is better to use
the refline from the Y2 axis value=’Drug B’ and DiscreteOffset=0.5, a SAS 9.3 feature.

title 'Hazard Function for Adverse Events of Special Interest';

proc sgplot data=sgbook.hazard;
step x=day y=value / yerrorupper=high yerrorlower=low group=drug
justify=center lineattrs=(thickness=2) name='step';
scatter x=day y=drug / markerchar=risk y2axis;
refline druglabel / label labelpos=min axis=y2 lineattrs=(thickness=0);
refline -0.0005 / noclip;
keylegend 'step' / location=inside position=topright across=1;
xaxis values=(0 to 200 by 10);
yaxis offsetmin=0.15;
y2axis offsetmax=0.9 display=none;
run;
298 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

12.12 LFT Safety Panel, Baseline vs. Study

This graph shows the distribution of LFT visitnum labtest drug pre result
values by baseline classified by test and 1 Week ALAT Drug A (n=240) 0.27216 0.64893
time. The data is as shown in the table. 1 Week Bilirubin Total Drug A (n=240) 1.28536 0.54600
1 Week Alk Phosphatase Drug A (n=240) 0.21586 0.01842
• Here we use the SGPANEL 1 Week ASAT Drug A (n=240) 0.46319 0.64387
procedure with Type=Lattice. 3 Months Alk Phosphatase Drug B (n=195) 1.67485 0.18387
3 Months ASAT Drug B (n=195) 0.07650 0.21848
• Column class variable is Labtest
and Row class variable is Visit.
• Each cell has a scatter plot of Result x Baseline using filled markers with transparency.
• Reflines for multiple clinical concern levels are displayed.

title 'LFT Safety Panel, Baseline vs. Study';

footnote j=l italic height=8pt "* For ALAT, ASAT and Alkaline Phosphatase, the...";
footnote2 j=l italic height=8pt " For Bilirubin Total, the CCL is 1.5 ULN: where ...”;
proc sgpanel data=sgbook.labs;
panelby labtest visitnum / layout=lattice onepanel novarname;
scatter x=pre y=result/ group=drug markerattrs=(symbol=circlefilled size=9)
transparency=0.7;
refline 1 1.5 2 / axis=Y lineattrs=(pattern=dash);
refline 1 1.5 2 / axis=X lineattrs=(pattern=dash);
rowaxis integer min=0 max=4 label='Study (/ULN)';
colaxis integer min=0 max=4 label='Baseline (/ULN) *';
keylegend/title=" " noborder;
run;
 Chapter 12 Health and Life Sciences Graphs 299

12.13 Immunology Profile by Treatment

This graph shows the immunology profile for sival trt cyc pt lbparm xval labelX labelY
patients in a study over time by treatment and lab 1.120 Drug A Cycle 1 1 C3 0 2 1.60
test. The study period is in four cycles. 0.147 Drug A Cycle 1 1 C4 0 2 0.45
1.440 Drug A Cycle 2 1 C3 0 5 1.60
0.278 Drug A Cycle 2 1 C4 0 5 0.45
• This is created using SGPANEL procedure 1.190 Drug A Cycle 3 1 C3 0 8 1.60
with TYPE=Lattice. 0.282 Drug A Cycle 3 1 C4 0 8 0.45
0.917 Drug A Cycle 4 1 C3 0 11 1.60
• The column var is treatment and row var is 0.163 Drug A Cycle 4 1 C4 0 11 0.45
lbparm. 1.130 Drug B Cycle 1 5 C3 0 2 1.60
• The “Cycle” labels at the top are displayed 0.220 Drug B Cycle 1 5 C4 0 2 0.45
using a scatter plot with MarkerChar. The positions of the labels are set in the data as
LabelX and LabelY.
• The 12 observations on the x-axis across four cycles use a format to display the
repeating 0, 15, 30 axis tick values.

title 'Immunology Profile by Treatment';

proc sgpanel data=sgbook.immune;
panelby trt lbparm / layout=lattice uniscale=column novarname;
series x=xval y=sival / group=pt lineattrs=(pattern=solid thickness=2);
scatter x=xval y=sival / group=pt markerattrs=(size=9 color=cx4f4f4f) name='a';
scatter x=labelx y=labely / markerchar=cyc markercharattrs=(size=9);
refline 3.5 6.5 9.5 / axis=x;
keylegend 'a' / title='Patient: ';
colaxis values=(1 to 12 by 1) offsetmin=0.02 offsetmax=0.02 label='Cycle Day' grid;
rowaxis grid;
run;
300 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

12.14 Most Frequent On-Therapy Adverse Events by Frequency

This graph shows the most frequent adverse events for the data shown below:
• The graph is split into two regions using the X-X2 split technique discussed earlier.
• The left side has a plot of percent by adverse event (AE) and treatment on x-axis with
OffsetMax=0.5.
• The right side has a plot of the relative risk and CL on X2 axis with OffsetMin=0.5.
• A wide, soft refline is used for the category bands.

AE A B Low Mean High RefAE

DYSPNEA 7.0 2.5 0.13 0.30 0.7 DYSPNEA
CHRONIC OBSTRUCTIVE AIRWAY 22.0 38.0 0.60 0.70 0.8
BACK PAIN 5.0 6.0 0.80 1.04 2.0 BACK PAIN
GASTROESOPHAGEAL REFLUX 3.0 4.0 0.50 1.05 3.8
HEADACHE 7.0 10.0 0.80 1.10 2.0 HEADACHE

footnote j=l 'Most Frequent On Therapy Adverse Events Sorted By Relative Risk';
proc sgplot data=sgbook.MostFrequentAE;
refline refae / lineattrs=(thickness=12) transparency=0.8;
scatter y=ae x=a / name='a' legendlabel='Drug A (N=216)';
scatter y=ae x=b / name='b' legendlabel='Drug B (N=431)';
scatter y=ae x=mean / xerrorlower=low xerrorupper=high x2axis;
refline 40 / axis=x;
keylegend 'a' 'b';
xaxis offsetmax=0.5 grid labelattrs=(size=8) valueattrs=(size=7)
label='Percent ';
x2axis offsetmin=0.5 type=log logbase=2 logstyle=logexpand grid max=64
labelattrs=(size=8) valueattrs=(size=7)
label=' Relative Risk with 95% CL';
yaxis display=(nolabel noticks);
run;
 Chapter 12 Health and Life Sciences Graphs 301

12.15 LFT Patient Profile

This graph shows LFT patient profiles for the data shown below using the SGPANEL procedure.
• The default panel layout is used with patient as the class variable.
• A band plot is used to display the range of the study days.

patient alat biltot alkph asat days dval sdays

Patient:5152 White Male Age: 48; Drug: A 0.50000 0.40000 0.40000 0.50000 -25 . .
Patient:5152 White Male Age: 48; Drug: A 1.36533 0.64286 0.68972 1.08498 0 -0.5 0
Patient:6416 White Male Age: 64; Drug: A 1.50000 1.50000 0.50000 1.00000 -25 . .
Patient:6416 White Male Age: 64; Drug: A 1.67422 1.63046 0.76092 1.08711 -10 . .
Patient:6969 White Female Age: 48; Drug: B 0.50000 0.30000 0.70000 0.60000 -25 . .
Patient:6969 White Female Age: 48; Drug: B 0.71729 0.45636 0.83046 0.79538 0 -0.5 0

title "LFT Patient Profile";

footnote1 j=l "For ALAT, ASAT and ALKPH, the Clinical Concern Level is 2 ULN;";
footnote2 j=l "For BILTOT, the CCL is 1.5 ULN: where ULN is the Upper Level of ..”;
proc sgpanel data=sgbook.lft cycleattrs;
panelby patient / novarname;
series x=days y=alat / markers lineattrs=(thickness=1px pattern=solid) name='a';
series x=days y=asat / markers lineattrs=(thickness=1px pattern=solid) name='b';
series x=days y=alkph / markers lineattrs=(thickness=1px pattern=solid) name='c';
series x=days y=biltot / markers lineattrs=(thickness=1px pattern=solid) name='d';
series x=days y=dval / lineattrs=(thickness=2px pattern=solid) name='e';
band x=sdays lower=dval upper=4.75 / transparency=0.9;
refline 1 1.5 2 / axis=Y lineattrs=(pattern=dash);
colaxis min=-50 max= 200;
rowaxis label="LFT (/ULN)";
keylegend 'a' 'b' 'c' 'd' 'e';
run;
302 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

12.16 Panel of LFT Values

This graph shows a classification panel of LFT values by lab Drug Group Pl Sl
test. Drug A (N=209) ALAT 0.21429 0.66250
• The default panel layout is used with Group as the Drug B (N=405) ALAT -0.02798 0.67263
Drug A (N=209) ASAT 0.65714 0.81250
class variable. Drug B (N=405) ASAT 0.44884 1.17656
• Each cell contains a scatter plot of sl * ps by drug. Drug A (N=209) BILTOT 0.78857 0.97500
• Three reference lines are added on each axis to Drug B (N=405) BILTOT 1.32225 1.08905
Drug A (N=209) ASAT 0.64286 0.77500
display the clinical concern levels. Drug B (N=405) ASAT 0.58379 0.67525

title "Panel of LFT Shift from Baseline to Maximum by Treatment";

footnote1 j=l "For ALAT, ASAT and ALKPH, the Clinical Concern Level ...";
footnote2 j=l "For BILTOT, the CCL is 1.5 ULN: where ULN is the Upper...";
proc sgpanel data=sgbook.Panel_LFT pad=(top=40 bottom=40);
panelby group / layout=panel columns=4 spacing=10 novarname;
scatter x=pl y=sl / group=drug;
refline 1 1.5 2 / axis=Y lineattrs=(pattern=dash);
refline 1 1.5 2 / axis=X lineattrs=(pattern=dash);
rowaxis integer min=0 max=4;
colaxis integer min=0 max=4;
keylegend / title="" noborder;
run;
 Chapter 12 Health and Life Sciences Graphs 303

12.17 Distribution of Eye Irritation Using PROC SGPANEL (9.3)

This graph shows the distribution of eye irritation by time and treatment. We have used PROC
SGPANEL to simulate the clustered grouped bar chart:

• We use a OnePanel ColumnLattice with class=time.

• Uniscale=Row is set to allow each cell to display only the categories it has.
• The bar labels are displayed using a scatter plot with MarkerChar option.

PARAM PERCENT time chartvar lcl ucl n

EYES ITCHY/GRITTY 40.0000 Week 1 Placebo 25.6864 54.3136 45
EYES ITCHY/GRITTY 60.8696 Week 1 Drug A 49.3541 72.3850 69
EYES ITCHY/GRITTY 52.0548 Week 1 Drug B 40.5947 63.5149 73
EYES ITCHY/GRITTY 62.2222 End Point Placebo 48.0567 76.3878 45
EYES ITCHY/GRITTY 69.5652 End Point Drug A 58.7083 80.4221 69
EYES ITCHY/GRITTY 57.5342 End Point Drug B 46.1954 68.8731 73

proc sgpanel data=sgbook.EyeIrritation;

format percent 4.1;
panelby time / layout=columnlattice onepanel novarname
noborder colheaderpos=bottom uniscale=row;
vbarparm category=n response=percent / group=chartvar datalabel
dataskin=pressed;
scatter x=n y=percent / group=chartvar yerrorupper=ucl markerattrs=(size=0)
yerrorlower=lcl groupdisplay=cluster;
colaxis display=(nolabel);
rowaxis grid max=100;
run;
304 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

12.18 Distribution of Eye Irritation Using PROC SGPLOT (9.3)

This graph shows the distribution of eye irritation by time and treatment using the SGPLOT
procedure. We have used the Y-Y2 split technique to display the statistics at the top:
• We used the VBarParm statement with cluster groups and error bars on the y-axis.
• For stats, use multiple scatter plots of lbl x time with MarkerChar option on the Y2 axis.
• We use OffsetMax=0.25 for Y axis and OffsetMin-0.8 on Y2 axis.

PARAM PERCENT time chartvar lcl ucl n Nlbl Lcllbl UclLbl PctLbl
EYES ITCHY/GRITTY 40.0 Week 1 Placebo 25.7 54.3 45 N LCL UCL %
EYES ITCHY/GRITTY 60.9 Week 1 Drug A 49.4 72.4 69 N LCL UCL %
EYES ITCHY/GRITTY 52.1 Week 1 Drug B 40.6 63.5 73 N LCL UCL %
EYES ITCHY/GRITTY 62.2 End Point Placebo 48.1 76.4 45 N LCL UCL %
EYES ITCHY/GRITTY 69.6 End Point Drug A 58.7 80.4 69 N LCL UCL %
EYES ITCHY/GRITTY 57.5 End Point Drug B 46.2 68.9 73 N LCL UCL %

title 'Distribution of Eye Irritation';

title2 h=9pt 'Parameter = Eye Redness (No / Yes)';
proc sgplot data=sgbook.EyeIrritationstats;
vbarparm category=time response=percent / group=chartvar limitupper=ucl
limitlower=lcl dataskin=pressed;
scatter x=time y=nlbl / group=chartvar groupdisplay=cluster markerchar=n Y2axis;
scatter x=time y=lcllbl / group=chartvar groupdisplay=cluster markerchar=lcl Y2axis;
scatter x=time y=pctlbl / group=chartvar groupdisplay=cluster markerchar=percent
Y2axis;
scatter x=time y=ucllbl / group=chartvar groupdisplay=cluster markerchar=ucl Y2axis;
xaxis display=(nolabel);
yaxis grid offsetmax=0.25;
y2axis offsetmin=0.8 offsetmax=0.03; run;
 Chapter 12 Health and Life Sciences Graphs 305

12.19 Vital Signs by Time Point Name

This is a graph of some of the vital signs of a patient in a study. The data is extracted from
CDISC form and plotted over time by test. In this case we have included only the blood pressure
and pulse readings as shown in the table below.
• We use the SGPANEL procedure with Layout=RowLattice with class var=vstestcd.
• We set Uniscale=Column to ensure each row gets its own data range.
• We used a band plot to add the normal ranges for easy evaluation.

vstestcd VISITDY VSSTRESN vstpt low high

DIABP -7 68 AFTER LYING DOWN FOR 5 MINUTES 80 100
DIABP -7 68 AFTER STANDING FOR 1 MINUTE 80 100
DIABP -7 62 AFTER STANDING FOR 3 MINUTES 80 100
PULSE -7 54 AFTER LYING DOWN FOR 5 MINUTES . .
PULSE -7 60 AFTER STANDING FOR 1 MINUTE . .
SYSBP -7 122 AFTER STANDING FOR 1 MINUTE 120 140
SYSBP -7 134 AFTER STANDING FOR 3 MINUTES 120 140

title 'Vital Signs by Point Name';

proc sgpanel data=sgbook.vitalsigns;
where vstestcd <> 'HEIGHT' and vstestcd <> 'TEMP' and vstestcd <> 'WEIGHT';
panelby vstestcd / layout=rowlattice onepanel novarname uniscale=column;
band x=visitdy lower=low upper=high / fill nooutline transparency=0.6;
series x=visitdy y=vsstresn / group=vstpt;
colaxis grid values=(0 to 120 by 10) valueshint;
rowaxis grid display=(nolabel);
run;
306 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

12.20 Concomitant Medications

This graph displays the medications administered to a patient in the study over time and by
medication name. The data is shown in the table below.
• We use the SGPLOT procedure and turn off automatic attribute cycling.
• Use a vector plot to draw the duration for each medication.
• We use two scatter plots to label the start and end events for the medications.
• We use a scatter plot with data label to display the medication name.

STARTDATE STARTDAY ENDDAY cmtrt CMMED Y lblday

30MAR13 27 112 BECONASE BECONASE 1 SPRAY 0.0 27
16FEB13 -48 -1 GINKGO BILOBA GINKGO BILOBA 1 Tab 0.1 -1
30MAR13 24 112 HYDROCORTISONE HYDROCORTISONE 1 Unit 0.2 24
16FEB13 -7 112 HYTRIN HYTRIN 5 Mg 0.3 -7

title "Concomitant Medications for Patient Id = XX-XXX-XXXX";

proc sgplot data=meds noautolegend nocycleattrs;
refline 0 / axis=x lineattrs=(thickness=1 color=black);
vector x=STARTDAY y=y / xorigin=ENDDAY yorigin=y group=cmtrt noarrowheads
lineattrs=(thickness=5px pattern=solid);
scatter x=STARTDAY y=y / group=cmtrt markerattrs=(size=13px symbol=trianglefilled);
scatter x=ENDDAY y=y / group=cmtrt markerattrs=(size=13px symbol=trianglefilled);
scatter x=STARTDATE y=y / markerattrs=(size=0) x2axis;
scatter x=lblday y=y / markerattrs=(size=0) datalabel=cmmed;
xaxis grid label='Study Days' offsetmin=0.02 offsetmax=0.02
values=(&minday2 to &maxday by 2);
x2axis notimesplit display=(nolabel) offsetmin=0.02 offsetmax=0.02
values=(&mindate2 to &maxdate);
yaxis display=(nolabel noticks novalues) min=0;
run;
 Chapter 12 Health and Life Sciences Graphs 307

12.21 Creating a 2 x 2 Cell Graph Using PROC SGPLOT

Previously, we have used either X-X2 or Y-Y2 split to create two-cell graphs. Here we take
the next step and use the X-X2 and Y-Y2 split technique together to create a four-cell (2 x 2)
graph using the SGPLOT procedure. Important features of the graph are as follows:

• X and Y axes are restricted to the lower 75% of the space (OffsetMax=0.25).
• X2 and Y2 axes are restricted to the upper 15% of the axis space (OffsetMin=0.85).
• Each graph is drawn using the appropriate combinations of X, Y, X2, and Y2 axes.
• Reference lines are used to demarcate the cells.

title 'Patient Profile';

proc sgplot data=sgbook.heart3 nocycleattrs;
scatter x=systolic y=weight / group=sex transparency=0.8 name='scat';
reg x=systolic y=weight / cli degree=2 nomarkers;
highlow y=deathcauseBP low=minbp high=maxbp / y2axis type=bar transparency=0.4;
scatter y=deathcauseBP x=meanbp / y2axis markerattrs=(symbol=diamondfilled)
name='bp' legendlabel="Mean";
highlow x=sexWt low=minWt high=maxWt / x2axis group=sexwt type=bar
transparency=0.4;
scatter y=meanWt x=sexWt / x2axis group=sexWt name='wt';
scatter x=sex y=deathcause / y2axis x2axis group=sex;
refline 'Female' / discreteoffset=-0.5 axis=x2 lineattrs=(thickness=2);
refline 'Cancer' / discreteoffset=-0.5 axis=y2 lineattrs=(thickness=2);
xaxis offsetmin=0.05 offsetmax=0.25 grid label='Systolic BP';
x2axis offsetmin=0.85 offsetmax=0.05 display=(nolabel) grid valueattrs=(size=7);
yaxis offsetmin=0.05 offsetmax=0.25 grid label='Weight';
y2axis offsetmin=0.85 offsetmax=0.03 display=(nolabel) grid valueattrs=(size=7);
keylegend 'scat' 'bp';
run;
308
Chapter 13
Business Graphs
13.1 Introduction 311
13.2 Stock Price and Volume Chart 312
13.3 Financial Trend and Bond Maturity Graph (9.3) 313
13.4 Danger of High P/E Ratios 314
13.5 Oil Consumption Trend by Country 315
13.6 Product Sales and Target Graph (9.3) 316
13.7 Social Network (9.3) 317
310 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“You can observe a lot just by watching.”

~ Yogi Berra
 Chapter 13 Business Graphs 311

Chapter 13: Business Graphs

13.1 Introduction
The graphs commonly used in the banking, financial, and insurance industries have different
requirements than graphs used for the statistical analysis of data. These graphs are more
“creative” in nature. While the principles of effective graphics continue to be applicable for these
graphs, they also generally need more visual aesthetics.

The graphs shown in this book have been customized for gray-scale rendering. However, color
graphs are particularly useful for business graphics, and the visual appearance of these graphs
can be further enhanced by the use of color styles.

The techniques described in the earlier chapters of this book can be leveraged in creative ways to
create such graphs. Here we have used the axis-splitting technique to display multiple graphs in
one cell, like the graph for bond yields. We have used combinations of HighLow and VBar to do
the “Product Sales and targets” graph, and we used a bubble plot to do the “Social Graph”.

In this chapter, we will examine how you can create some common graphs used in this industry.
Some examples are shown in Figure 13.1. Detailed code for creating such graphs is shown in
the following pages.

Stock Plot Bond Yields

Product Sales and targets Social Graph

Figure 13.1
312 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

13.2 Stock Price and Volume Chart

This is a common stock price chart from the data shown below with the following features:

• Y-Y2 split technique is used to draw price data above the volume data.
• Price data is an overlay of the Bollinger band, 25 and 50 event moving average, and
price.
• Volume data includes volume and moving average.
• The key legend is placed inside in the empty upper right corner with all the details.

Stock Date Open High Low Close Volume avg25 avg50 upper lower vavg
IBM 01AUG88 $126.00 $126.87 $110.37 $112 5,256,886 $23 . $29 $16 7560307.56
IBM 01SEP88 $111.12 $116.50 $109.50 $115 5,433,352 $22 . $29 $16 7552677.08
IBM 03OCT88 $115.12 $124.87 $112.75 $123 5,790,742 $22 . $29 $16 7462149.84

title 'Stock Plot of Share Price and Volume';

proc sgplot data=sgbook.Stock_IBM_Stats;
where date > '01Jan2000'd;
band x=date upper=upper lower=lower / y2axis transparency=0.5 legendlabel='Boll…';
highlow x=date high=high low=low / y2axis open=open close=close
lineattrs=(color=black thickness=1) legendlabel='Value';
series x=date y=avg25 / y2axis lineattrs=graphdata1 legendlabel='25 Moving Avg.';
series x=date y=avg50 / y2axis lineattrs=graphdata2 legendlabel='50 Moving Avg.';
needle x=date y=volume / lineattrs=graphdata3(pattern=solid);
series x=date y=vavg / lineattrs=graphdata4(opts) legendlabel='Avg Volume';
keylegend / location=inside position=topright across=1 valueattrs=(size=8);
yaxis max=100000000 display=none;
y2axis display=(nolabel) grid min=20 values=(40 to 160 by 20) valueshint;
xaxis display=(nolabel) grid;
run;
 Chapter 13 Business Graphs 313

13.3 Financial Trend and Bond Maturity Graph (9.3)

This graph shows the financial trend over time and bond yields by maturity. The features of the
graphs are: Maturity Rate Year Value High Low Zero
6 Months .60% . . . . 0
1 Year 1.0% . . . . 0
• X-X2 and Y-Y2 split technique is used to display 2 Year 1.5% . . . . 0
the financial trend and yield in the same graph. 5 Year 1.7% . . . . 0
• The trend is drawn in the X-Y2 space using two 10 Year 2.0% . . . . 0
series, one band and one HighLow plot. 15 Year 2.5% . . . . 0
30 Year 3.0% . . . . 0
• The bond yields are drawn in the X2-Y space . 2004 1.4% 2.0% .94% 0
using a HighLow plot. . 2005 1.6% 2.2% 1.5% 0
. 2006 2.0% 2.7% 1.6% 0
. 2007 2.5% 2.8% 2.3% 0
. 2008 3.0% 3.6% 2.6% 0
. 2009 3.5% 4.0% 3.2% 0

title 'Financial Trend over Time and Current Bond Yield';

proc sgplot data=sgbook.Trend noautolegend;
band x=year lower=0 upper=low / transparency=0.4 y2axis;
series x=year y=high / lineattrs=(thickness=5) y2axis curvelabel='High';
series x=year y=low / lineattrs=(thickness=5 color=cxc0c0c0)
y2axis curvelabel='Low';
highlow x=year high=value low=zero / highcap=filledarrow y2axis
type=bar barwidth=0.3 transparency=0.0 highlabel=value;
highlow y=maturity low=zero high=rate / type=bar barwidth=0.8 x2axis
highlabel=rate transparency=0.3 ;
xaxis type=discrete display=(nolabel);
x2axis offsetmin=0 offsetmax=0.3 display=(nolabel noticks novalues);
y2axis offsetmin=0 offsetmax=0.4 display=(nolabel);
yaxis offsetmin=0.45 offsetmax=0.05 type=discrete display=(nolabel) grid;
run;
314 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

13.4 Danger of High P/E Ratios

This graph displays the price-earnings (PE) ratios for pe10 Year re pe re10 label
stocks over the last 125 years. The features of the graph 18.2148 1881 8.7483 12.6327 7.5378 .
are as follows: 15.6153 1882 8.7123 13.4545 7.7364 .
15.2162 1883 8.7730 13.5116 7.8609 .
14.4526 1884 7.5805 12.9500 7.8101 .
• A band plot is used to draw a filled graph of the 13.2753 1885 6.8382 13.6774 7.8221 .
PE over time. 26.7007 1929 19.0579 18.0145 11.9516 1929
• A second band plot is drawn to show the PE=20 zone.
• A series plot is drawn to display the actual PE values over time.
• The PE values are displayed using the “Label” variable. This column contains non-
missing values only for the ones we want to include in the graph.

title "The Danger of High Price-Earnings Ratios";

footnote j=l "Source: Robert J. Shiller";
proc sgplot data=sgbook.PE_DATA noautolegend;
band x=year upper=pe10 lower=0 / fillattrs=graphdata2
name='high' legendlabel='PE > 20';
band x=year upper=20 lower=0;
refline 20 / label="PE=20";
refline 10 20 30 40 / lineattrs=GraphGridLines;
keylegend 'high' / location=inside position=topleft;
series x=year y=pe10 / lineattrs=GraphData1(thickness=3) datalabel=label;
yaxis offsetmin=0;
run;
 Chapter 13 Business Graphs 315

13.5 Oil Consumption Trend by Country

This graph displays the oil Year Country value total subtotal subtotalpct valuepct valuelabel
consumption trend over 30 years by 1965 USA 11522 29524 11992 40% 38% .
1965 China 217 29524 11992 40% 1% .
country as a percentage of 2005 1965 India 253 29524 11992 40% 1% .
consumption.The data is shown in 2005 USA 20655 77099 30128 100% 69% 69%
the table. 2005 China 6988 77099 30128 100% 23% 23%
2005 India 2485 77099 30128 100% 8% 8%
• The SGPANEL procedure is used with PANELBY = country.
• The empty bar chart plots the combined trend for all three countries.
• The filled bar chart plots the consumption of the individual country as a percentage of the
total.
• A bar chart is used to display the bar label for just the most recent year.

title 'Oil Consumption Trend Comparison by Country';

proc sgpanel data=sgbook.oil nocycleattrs;
where country NE 'Japan' and year > 1971;
panelby country / columns=1 onepanel layout=rowlattice novarname spacing=6;
vbar year / response= subtotalpct transparency=0.8
legendlabel='Total Consumption' name="a" fillattrs=graphdata5 ;
vbar year / response= valuepct transparency=0.2 fillattrs=graphdata2
legendlabel='Country Consumption' name="b" barwidth=0.5 ;
vbar year / response= valuelabel datalabel fillattrs=graphdata2;
rowaxis grid label='Percent of Total for 2005' values=(0 to 1 by 0.25);
colaxis fitpolicy=thin;
keylegend "a" "b";
run;
316 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

13.6 Product Sales and Target Graph (9.3)

This graph displays sales by Status Sales Target High Low Product High2 Low2
product, with target and the Exceeded $20,000 $19,000 $21,768 $15,256 Bulbs 23945.27 13730.65
Exceeded $25,000 $22,000 $30,991 $18,882 CFL 34089.85 16993.41
historical low and high range. The ShortFall $12,000 $14,000 $13,624 $9,174 LED 14986.82 8256.79
data is shown on the right. Exceeded $16,000 $14,000 $19,844 $12,535 Spot 21828.51 11281.64
ShortFall $18,000 $21,000 $19,882 $15,484 Tube 21870.44 13935.41
• A grouped VBARPARM statement is used to display the sales data.
• Two HIGHLOW plots are used to draw the historical range for each product.
• A SCATTER plot is used to draw the target arrow.
• Another HIGHLOW bar is used to draw the target line.
• All items are displayed in an inside legend.
• Note: Some hard-coded colors that work for this style may need adjusting.

title 'Product Sales and Targets';

proc sgplot data=sgbook.target;
highlow x=product high=high low=low / type=bar barwidth=0.8 transparency=0.5
name='range' legendlabel='Historical Range' nofill;
highlow x=product high=high2 low=low2 / type=bar barwidth=0.7
lineattrs=(color=white thickness=2) nofill;
vbarparm category=product response=sales / group=status datalabel
dataskin=gloss barwidth=0.6 name='bar';
highlow x=product high=target low=target / type=bar barwidth=0.6
nofill lineattrs=(color=black);
scatter x=product y=target / markerattrs=(symbol=trianglerightfilled size=12)
discreteoffset=-0.3 name='target' legendlabel='Target';
keylegend 'range' 'target' 'bar' / location=inside position=topright across=1;
yaxis display=(nolabel) offsetmin=0 grid;
xaxis display=(nolabel);
run;
 Chapter 13 Business Graphs 317

13.7 Social Network (9.3)

This graph displays social relationships between
Node x y size Group Link x1 y1 x2 y2
persons in a social network.
A -1.5 2 1.0 Direct A -2.5 0 0.0 0
B 1.0 3 1.0 Indirect B -2.5 0 -1.5 2
• A BUBBLE statement is used to draw the C 2.0 2 1.0 Indirect C -1.5 2 1.0 3
nodes, with a size variable. D 2.0 -1 1.0 Indirect D 0.0 0 1.0 3
E 1.0 -2 1.0 Indirect F 0.0 0 2.0 2
• A VECTOR statement is used to draw the links. F -1.0 -2 1.0 Direct G 0.0 0 2.0 -1
The arrowheads are turned off. G -2.5 0 1.1 Person H -2.5 0 -1.0 -2
• A legend is used to show the type of H 0.0 0 1.3 Direct I -1.0 -2 1.0 -2
relationship.

title h=14pt 'Social Graph';

title2 h=11pt ' The pattern of social relationships between people';
proc sgplot data=sgbook.social;
vector x=x2 y=y2 / xorigin=x1 yorigin=y1 lineattrs=(thickness=1)
noarrowheads;
bubble x=x y=y size=size / group=group bradiusmin=30 bradiusmax=40
fill outline lineattrs=(thickness=3) name='b';
keylegend 'b' / title='Relationship:' location=inside position=topleft
across=1;
yaxis display=none;
xaxis display=none offsetmin=0.1 offsetmax=0.1;
run;
318
Chapter 14
Styles
14.1 Introduction 321
14.2 Using Styles 321
14.3 Style Elements 321
14.4 Using Style Elements 324
14.5 Style Element Usage Precedence 324
320 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“The theory of probabilities is at bottom nothing

but common sense reduced to calculus.”
~Laplace
 Chapter 14 Styles 321

Chapter 14: Styles

A brief introduction to usage of styles with ODS Graphics was included in Section 2.5. Let us
review that information and expand on the topic of styles and their usage.

14.1 Introduction
Graphs created by the SG procedures derive their visual look and feel from the active style for the
output destination. Each ODS destination has a default associated style. Styles supplied by SAS
are designed to ensure effectiveness and an aesthetically pleasing appearance by default.
Default styles for common destinations are shown in Figure 14.1.

You can create your own style to suit the look you ODS Destination Style Name
prefer or to present a consistent corporate look by LISTING Listing
using the TEMPLATE procedure. The description of
HTML HtmlBlue
PROC TEMPLATE is beyond the scope of this book.
RTF RTF
However, since styles are intimately related to the
PDF PRINTER
topic of graphs, we will cover the related relevant
topics here. Figure 14.1

For a discussion of styles and the TEMPLATE procedure, see the SAS documentation.

14.2 Using Styles

To change the overall visual appearance of the graphs, you can assign a different style by using
the STYLE option on the ODS destination statement.

ods destination style=style-name;

When you do this, all subsequent output sent to the destination will use this assigned style until
the destination is closed or the style is changed.

14.3 Style Elements

A style is a collection of named elements. Each element is a bundle of named visual attributes
such as color, font, marker symbols, and so on. Output tables derive visual attributes such as
fonts, background colors for the headers, size, color of titles, etc. from the style.

The default visual attributes of various graph elements are also derived from the style. Plot
colors, marker symbols, line thickness, axis label fonts, etc. are derived from specific named
elements of the style. The association between the element of the graph and the style element
is well defined and described in detail in the product documentation. Some common style
element names and their associated graph elements are shown in Figures 14.2 and 14.3.
322 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

GraphLabelText GraphDataText GraphTitleText

GraphWalls

GrapValueText GraphFootnoteText
Graph
Figure 14.2: Graph and Text Elements

Figure 14.2 above lists some style elements that are commonly used for various elements in
the graph. These elements are used to determine the visual attributes for the graph
background, walls, and the various text statements such as titles, footnotes, etc. The
elements are listed below. This graph was created with SAS 9.3 PROC SGPLOT.

Element Name Usage

Graph Graph background fill and border
GraphWalls Wall background fill and border
GraphTitleText Title font, size, color, etc.
GraphFootnoteText Footnote font, size, color, etc.
GraphLabelText Axis Label font, size, color, etc. Also Legend Title
GraphValueText Axis tick value and Legend entry font, size, color, etc.
GraphDataText Data labels font, size, color, etc.

If you want to create a new style that has smaller fonts for the tick values, you can customize
the GraphValueText element appropriately. Then, all the graphs using this style will display
the new font you have selected.

A new style can be created using the TEMPLATE procedure. It is generally good practice to
derive a new style from a parent style. In this way, you only need to customize the elements
that you need to change. For details on how to create a new style, see the SAS
documentation for the TEMPLATE procedure.
 Chapter 14 Styles 323

GraphGridLines GraphData1 - GraphDataN

GraphDataDefault GraphReference

Figure 14.3: Plot Elements

Figure 14.3 above shows some style elements that are commonly used for various plot
elements in the graph. These elements are used to determine the visual attributes for the
parts of the bar chart or series plot.

The GraphDataDefault element is used to display plots that do not have a GROUP option. In
Figure 14.3, the bar chart does not have a GROUP option set, so all the bars for this plot use
the GraphDataDefault style element.

Plot statements that use a GROUP option utilize the visual attributes from GraphData1 to
GraphData12 style elements. Most styles supplied by SAS have 12 data elements.
However, there is no limit, and you can define more or fewer elements as you please. For a
full list, see the SAS documentation.

Element Name Usage

GraphDataDefault Non-Grouped data elements
GraphData1 - N Grouped data elements – or when using CycleAttrs
GraphFit Fit lines
GraphFit2 A second element for Fit lines, not used by default
GraphReference Reference lines
GraphGrid Grid lines
GraphConfidence Confidence plots
324 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

14.4 Using Style Elements

If you customize and use your own style, this new style is used for all procedure output to the
ODS destination, until the destination is closed, or the style is changed again. Often you do not
want to change the style setting of a destination, but only want to make a specific change for one
graph.

For example, the default font size for the x-axis tick values may be just a little too big to fit the
space available. Shrinking the font just a bit will do just fine. One way to do this would be to
change the size of the currently used font:

xaxis valueattrs=(size=7);

You will likely have seen such syntax in many places in the examples. This works fine, but in
other cases you want to set the visual properties of one plot to the same as another item in the
graph. You can use the Data Label font for the Axis Tick Values as follows:

xaxis valueattrs=GraphDataText;

This technique is especially useful when using color output. If you have two separate measures
plotted on the same graph, one to the y-axis and one to the Y2 axis, it can be useful to set the
color of the axis tick values to the same color as the plot line. You can do that as follows:

yaxis valueattrs=GraphData1;
y2axis valueattrs=GraphData2;

14.5 Style Element Usage Precedence

When you use a color style, hard-coded color values are not recommended. This is because the
value specified on the plot option wins over the visual attributes derived from the ODS style. You
may set the color of a line to “Black” because the wall color of the current style is white. But if the
same graph is run with a style that has a dark wall color, this “Black” color will blend into the wall
color, resulting in a bad color choice for the graph.

The precedence for usage of visual attributes is as follows:

1. First the procedure derives the visual attributes from active Style.
2. This is replaced by attributes from Discrete Attribute Map (if applicable).
3. This is replaced by visual attributes specified on the statement option.

For the reason above, and to ensure your graph works well for all use cases, it is recommended
you always use “style relative” settings. If you do not like a color used in the graph, try one of the
other colors defined in the style. Hard-coded colors are best used for one-off graphs that are not
likely to be used with different style settings (see Figure 13.6).
Chapter 15
ODS Destination and ODS Graphics
Options
15.1 Introduction 327
15.2 ODS Destination Options 327
15.3 ODS Graphics Options 328
326 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“Not everything that can be counted counts,

and not everything that counts can be counted.”
~ Albert Einstein
 Chapter 15 ODS Destination and ODS Graphics Options 327

Chapter 15: ODS Destination and ODS Graphics

Options
15.1 Introduction
All graph output from SG procedures is sent to the open ODS destinations. If multiple
destinations are open, output is sent to each destination using the various options that are in
effect for each destination.

Global options that apply to the graphs are specified on the ODS GRAPHICS statement. In
Chapter 2, we discussed the use of the ODS GRAPHICS statement to switch on the creation of
automatic graphs from the SAS procedures. With SAS 9.3, ODS Graphic is ON by default for all
procedures including the SG Procedures. Global options that control certain aspects of the
graphs can be provided on this statement. These options are discussed below.

15.2 ODS Destination Options

The options relevant to graphs are discussed in Figure 15.1 below. For complete details on all
options for each ODS destination, see the SAS documentation for ODS Language Statements.

Option Values Description

DPI Value DPI for generated document
GPATH Location The location for all graphics output for this destination.
IMAGE_DPI Value DPI for generated image.
PATH Location The location for the output file for this destination.
SGE OFF | ON Enable / disable creation of editable graph output
STYLE Style name The ODS Style to be used for this destination.
Figure 15.1

SGE option: This option is used to enable or disable the creation of editable graphical output in
the format used by the ODS Graphics Editor. For SAS 9.2, only the LISTING destination
supports this option. For SAS 9.3 this option is available on all ODS destinations.

By default, the SGE option is OFF. When SGE=ON, an editable graph is created along with the
image file. This output is listed in the Results window. You can open this file to launch the ODS
Graphics Editor to make non-persistent customizations to the ODS Graphics output.

DPI and Image_DPI: Graphs are rendered at default DPI for the output destination. For some
destinations, the DPI applies only to the graph image. The size of the graph is based on the
active DPI. If the size is specified in pixels (the default is in pixels), then the baseline for scaling
is the active DPI for the destination.
328 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

15.3 ODS Graphics Options

Certain attributes of the ODS graphics output are controlled by options on the ODS GRAPHICS
statement. These options stay in effect until reset or the ODS Graphics system is turned off. The
syntax for the statement is as follows:

ods graphics <on | off> </ options;

The options are listed in Figure 15.2 below.

Option Values Description

ANTIALIAS ON | OFF Enable or disable anti-aliasing of lines and markers
ANTIALIASMAX Value Maximum number of observations for anti-aliasing
BORDER ON | OFF Display or hide the graph border
DISCRETEMAX Value Maximum number of discrete values
GROUPMAX Value Maximum number of group values
HEIGHT Value Height of the graph in px, in, cm, etc.
OUTPUTFMT PNG, GIF, etc. Any one of the supported formats for the destination
IMAGEMAP ON | OFF Enable or disable creation of image map for HTML
IMAGENAME string Name for the output file
LABELMAX Value Maximum number of labels displayed
MAXLEGENDAREA Value Maximum % of the height occupied by legends
PANELCELLMAX Value Maximum number of cells in a panel
RESET Reset all (or individual) options to default
SCALE ON | OFF Enable or disable scaling of graph elements
TIPMAX Value Maximum number of tips displayed
WIDTH Value Width of the graph in px, in, cm, etc.
Figure 15.2

For complete details on all options for each ODS destination see the SAS documentation for ODS
Language Statements.

Anti-aliasing: This is a technique used to improve the rendering of various elements in the
graph such as the markers, lines, text, and so on. Anti-aliasing improves the visual appearance
of the elements but consumes more computing resources. Text elements in the graph are always
anti-aliased for high quality. Anti-aliasing of the line and marker elements can be controlled by
these options. When there are too many observations in the data, anti-aliasing becomes less
effective and more costly. By default, anti-aliasing is turned off for markers and lines when the
number of observations exceeds 600. The ANTIALIAS option can be used to enable or disable
anti-aliasing. The ANTIALIASMAX option can be used to change the level when anti-aliasing is
turned off.

DISCRETEMAX, GROUPMAX: If the number of discrete values in a graph, such as number of

categories on the axis or the number of group values becomes large, the procedure can slow
 Chapter 15 ODS Destination and ODS Graphics Options 329

down or even run out of memory. This can happen if the user inadvertently selects a variable that
has a large number of discrete values for a discrete role, such as zip code, etc. In normal usage,
variables used for such roles have < 1000 unique values. In this case, the procedure will stop
processing and log a message. If you really need to run such a use case, you can set these
settings to a larger value to run your use case.

OUTPUTFMT: The default output format is determined by the ODS destination. For many
destinations like LISTING, HTML, etc., the graph is created in industry-standard image formats.
You can specify the format you want by using this option. The value you provide will only be
honored if the output destination supports the format.

IMAGEMAP: For the HTML destination, tool tips can be displayed for the data values in the
graph. By default, this setting is OFF since it consumes computing resources. You can enable
creation of the image map for display of tool tips for HTML destinations by using this option.

IMAGENAME: The default name for the graph output file is “sgplot.xxx” or “sgpanel.xxx” based
on the procedure used. File names are automatically appended with an incremented counter to
avoid overwriting of the files from multiple procedure executions. You can provide your own
preferred name for the filename prefix.

LABELMAX: Often you can request the display of data labels with a scatter plot. The
procedures will attempt to place the data labels close to the markers, with minimal collision
between labels. At some point, there are too many labels in the graph, and the graph becomes
ineffective. By default, the display of labels is disabled when the number of labels exceeds 200.
You can use this limit to control this feature.

MAXLEGENDAREA: By default, the legends for the graph are dropped if the area occupied by
them exceeds 20% of the total area of the graph. This is to prevent cases where the legend is so
big that there is not much space left for the display of the data itself. You can control this limit by
using this option.

PANELCELLMAX: The SGPANEL procedure subdivides the graph into cells based on the
number crossings of the classification variables. By default, if the number of cells exceeds
10,000, the graph is not created and a message is logged. You can control this limit by using this
option.

RESET: You can reset all the options to the default settings by using this option. You can also
use reset=option-name to reset one or more specific options. You can use reset=index to
reset just the counter that is used to postfix an index to the output file name.

SCALE: The output graph size is determined by the procedure. See the WIDTH or HEIGHT
option. If you request a different height of the graph, the internal details of the graph are scaled
using a nonlinear algorithm designed to render graphs with reasonable font sizes. In such a
case, the actual font size may be different from what was specified for the text in the graph. You
can enable or disable this scaling for height by using this option.
330 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

TIPMAX: Tooltips can be displayed for the data values for the HTML destination. By default, if
the number of observations exceeds 500, tooltips are disabled. You can control this limit by using
this option.

WIDTH and HEIGHT: The default width and height for the graph is determined by the procedure,
and is often 640px x 480px. In case of paneled graphs, the size may vary based on the number
of cells. You can control the size of the graph using these options.
Chapter 16
Tips for Graph Output

16.1 Introduction 333

16.2 Creating Small Graphs for Use in Documents 333
16.3 Creating Large Graphs for Use in Presentations 334
16.4 Combining Graph Size and DPI 336
16.5 Impact of Graph Size on System Resources 336
332 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

“Each problem that I solved became a rule which

served afterwards to solve other problems.”
~ Rene Descartes
 Chapter 16 Tips for Graph Output 333

Chapter 16: Tips for Graph Output

16.1 Introduction
After completing the data analysis, the analyst needs to share the results of the analysis, along
with graphs, with various consumers. This means the graphs will be inserted into Web pages,
publications, documents and overhead presentations. It is critical that these graphs look their
best and deliver the information effectively in the final form.

For the LISTING destination, the default graph size is 640 pixels x 480 pixels at 96 DPI. These
default settings work well for viewing the graphs on the computer screen. For inclusion of such
graphs in various documents, some customization is useful.

16.2 Creating Small Graphs for Use in Documents

Often you will need to insert a graph into a document where the graph must fit inside a small inset
or in one-column of a two-column layout. Figures 16.1 and 16.2 show two graphs, both with a
width of 2.875 inches. For Figure 16.1, we have inserted the default output from the SGPLOT
procedure into the available region. This default output is rendered as a 640px x 480px image.
The publication software linearly scales the image to fit the box, thus shrinking all the graph
elements, including the fonts, making the font size less than optimal.

Figure 16.1 Figure 16.2

For the graph for Figure 16.2, we need to fit the graph inside a 2.875-inch-wide box. So, we have
created the graph of the correct size by using the ODS Graphics WIDTH option. When this graph
is inserted into the document, no image scaling is required.

The graph in Figure 16.2 has fonts that are easy to read. The line thickness is also bolder. If the
graph had markers, they would be bigger. Here the graph is doing the scaling, and it uses a font-
friendly way to scale these items, resulting in a more readable graph.
334 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

Design and Render Size: Each graph has a default “design” size. For most cases, this is 640px
x 480px. This size cannot be changed directly in SG procedures, but it can be changed if you are
using GTL. A different “render” size can be specified by using the WIDTH and HEIGHT options
on the ODS GRAPHICS statement.

When the render size is different from the design size, the graph is scaled to fit the new size. All
individual elements inside the graph are scaled in a sub-linear fashion. The scale is:
0.25
Scale-Factor = (Render-Height / Design-Height)

This is done specifically to slow the shrinking or growth of the font sizes. Therefore, as the graph
height shrinks, the font size shrinks too, but at a slower rate. Similarly, as the graph height
grows, the font size grows too, but at a slower rate.

For our use case shown above, Design-Height = 480px. The Render-Width is 2.875. If we retain
the aspect ratio, the render width is 2.875 *.0.75 = 2.16in at 96 DPI, resulting in a graph width of
207 pixels. The Linear scale for the image size is 207 / 480 = 0.43. However, the Scale-Factor
shown above is 0.43 ** 0.25 = 0.81.

In Figure 16.1, the scale factor for the image and the fonts is 0.43. Therefore, a 10-point font
effectively shows as a 4-point font, which is quite small. For the graph in Figure 16.2, the scale
factor for image is 0.43, but the scale factor for the fonts is 0.81. So, a 10-point font is drawn at 8
point, which is still quite readable.

Recommendation for Creating Small Graphs: Always render your graph at a size close to its
final usage. Using the default size graph in a document will be fine if the graph fills a width of 6 to
7 inches. But if the graph is to fit a smaller size, do not use the default size graph.

Caveat: This does mean that the final rendered font sizes may differ from those specified in the
code or the style. If it is important that the font size be exactly as specified, then you can use the
NOSCALE option on the ODS GRAPHICS statement. Even if the graph render height is different
from the default design height, the fonts, lines, and markers will not scale.

16.3 Creating Large Graphs for Use in Presentations

The dimensions of the output image are also controlled by the DPI or IMAGE_DPI for the
destination. Each destination has its own default DPI
used by the document or just the image. Some default ODS Destination Default DPI
values are shown in Figure 16.3. LISTING 96
HTML 96
For LISTING and HTML destinations, the graph image RTF 150
is created as a separate image file. In this case, the PDF 150
default image format is PNG, and the DPI information Figure 16.3
is included in the file.
 Chapter 16 Tips for Graph Output 335

If you want to create graph output that can be embedded in another document or
presentation, it is convenient to create a separate file using the LISTING destination. With
SAS 9.2, all graph output is in image format. Even for PDF and RTF output, the graph is
essentially an image embedded into the document.

With SAS 9.3, support for vector graphics output has been added. You can now generate
EMF, PDF, PS, SVG, and PCL output, depending on the ODS destination you use. In some
special cases, the graphs use data skins or have gradient legends. In such cases, the output
is in image formats. This is also the case when using transparency for EMF and PS.

Graph size is controlled by setting the WIDTH and HEIGHT options on the ODS GRAPHICS
statement and DPI, as seen in section 16.2. When creating smaller graphs for embedding
into documents, it is useful to change the graph size. It is also useful to increase the graph
size when creating a graph that is more detailed in nature and needs more space for
rendering all the details.

Recommendations for Creating Large Graphs: When presenting a graph to a large

audience using a projection system, it is necessary that the details of the graphs are visible
from a distance. The same graph that looks nice on the screen from two feet away should be
viewable on the large screen from 50 to 100 feet away. To achieve this, we need to “scale
up” the graph so that all aspects of the graph are made bolder and bigger linearly.

Vector format output is inherently scalable and ideal for use in some cases. So you can use
the default output in EMF or PDF formats for such use cases. You can enlarge the graph on
the computer screen to fill the window, and these vector-based graphs will scale up nicely.

Image-based output created at a low (or default) resolution or DPI does not scale up very
well. When scaled, you will see “jaggies” along the edges of the marker, lines and fonts. To
include such graphs in a presentation using a projection system to a large audience, you
should increase the DPI of the graph. This will make the graph larger and also make all the
elements of the graph (for example, markers, line thickness, and fonts) linearly bigger.
Therefore, this graph, when viewed from afar, will look readable.

Creating a graph with a size of 6.4in x 4.8in at 200 DPI will generate an image that is 1280px
x 960px. The PNG file will include the DPI information in the file; therefore, when you include
this graph in a document, it will occupy a region 6.4in x 4.8in by default. However, if you view
this file in some image viewers or Internet Explorer, you may see the image in the full 1280 x
960 size because these viewers do not honor the DPI setting in the PNG file.

Recommendations for Graphs in Web Pages: Some Web browsers do not honor the DPI
setting in the PNG file. A graph intended to be 6in x 4in (200 DPI) may be shown as 12in x
8in in the browser. Other browsers do scale, but poorly. So, it is best to use the ODS HTML
destination, or create graphs at 96 DPI for usage in Web pages.
336 Statistical Graphics Procedures by Example: Effective Graphs Using SAS

16.4 Combining Graph Size and DPI

As discussed in sections 16.2 and 16.3, scaling of the graph for height and DPI are done
differently. Each is useful for different use cases. You can also mix the two judiciously to achieve
the custom look you need. High DPI is also useful for documents since they are usually printed
at a higher resolution (300 to 600).

In this book, we have used a size of 3in x 2in at 300 DPI for most of the graphs. Specifying the
size shrinks the fonts a little. Using high DPI gives us graphs that will scale well when printed on
a high-resolution printer. For graphs that occupy a larger area, like those in Chapters 12 and 13,
we have used a larger size of 6in x 3in at 200 DPI.

16.5 Impact of Graph Size on System Resources

When creating graphs as image output, higher system resources are consumed to create larger
graphs with high DPI.

• More system memory is needed to create the graph. The rendering system is Java based.
SAS system settings for Java specify the amount of memory that can be used by Java.
Using higher size and DPI requires a higher amount of memory. If the graph cannot be
rendered, a warning is logged. If you need the higher resolution graph, you can change the
memory settings using the JREOPTIONS option specified in the SAS system options. See
the SAS documentation for details.

-jreoptions '(-Xmx512m)'

• If sufficient memory resources are available, the graph will be rendered. However, it will
likely consume more system and CPU time to render the graph.

• If the graph is successfully rendered, the output file size will be bigger.
Index
A ARROWHEADSHAPE= option, NEEDLE
statement (SGPLOT) 99
accuracy of magnitude perception 6
arrows, emphasizing data with 247
ACROSS= option, KEYLEGEND statement
attribute maps 235, 251–253
(SGPLOT) 231
automatic graphs from procedures 10–11
adverse event timelines 291–292, 300
automatic paging of panels 28
ALPHA= option
axes
CLM option 140
See also discrete axis
DOT statement (SGPLOT) 215
See also linear axis
ELLIPSE statement (SGPLOT) 153,
See also log axis
155
See also time axis
HBAR statement (SGPLOT) 194
about 20, 221
HLINE statement (SGPLOT) 208
axis assignment examples 134, 141,
LOESS statement (SGPLOT) 140
148
PBSPLINE statement (SGPLOT)
creating axis-aligned statistics table
146–147
240
REG statement (SGPLOT) 132
creating multi-line tick values 239
VBAR statement (SGPLOT) 186
creating tick values with Unicode 238
VLINE statement (SGPLOT) 201
options supported 221
annotations
SGPLOT procedure 22
about 235–237
AXIS= option, REFLINE statement
adding company logos 243
(SGPLOT) 114
background images in graphs 244
bubble legends 248 B
creating axis-aligned statistics table
background images in graphs 244
240
band plots
creating axis tick values with Unicode
about 89
238
graph examples 61
creating forest plots with 241–242
grouped 91, 124
creating multi-axis tick values 239
grouped with transparency 91
images as curve labels 245–246
overlay 92, 124
making polyline figures 249–250
overlay with constant lower limit 92
QTc change graph with annotated at
overlay with curve labels 93
risk values 296
overlay with scatter and series plots
using arrows to emphasize data 247
122
ANTIALIAS option, ODS GRAPHICS
overlay with scatter plots 123
statement 328
overlay with series plots 123
ANTIALIASMAX= option, ODS GRAPHICS
overlay with step plots 93
statement 44, 79, 328
roles and options supported 89
ARROW function 236
338 Index

BAND statement, SGPLOT procedure BORDER option, ODS GRAPHICS

CURVELABELLOWER= option 93 statement 328
CURVELABELPOS= option 93 box plots
CURVELABELUPPER= option 93 grouped 42
LEGENDLABEL= option 92 horizontal 174–176
LOWER= option 92 interval 43
MODELNAME= option 90, 93 row lattice example 55
TRANSPARENCY= option 124 SGPLOT procedure 22
bar charts vertical 42, 161, 169–173
See also horizontal bar charts BOXWIDTH= option, VBOX statement
See also vertical bar charts (SGPLOT) 171
bar-bar overlay 49 BRADIUSMAX= option, BUBBLE statement
bar-line overlay 49, 218 (SGPLOT) 54, 248
classification panel example 55 BRADIUSMIN= option, BUBBLE statement
decorative skins and 15 (SGPLOT) 248
discrete order for 229 BREAK option
graph examples 47 HLINE statement (SGPLOT) 207
grouped 265 SERIES statement (SGPLOT) 82
grouped with skins 48, 51 STEP statement (SGPLOT) 88
HighLow 110–111 VLINE statement (SGPLOT) 200
multiple with patterns, fill colors, skins bubble legends 248
51 bubble plots
parametric 100–104 about 105
principles of effective graphics and 7–8 graph examples 54, 61, 106
SGPLOT procedure 22 grouped 106–107
stacked with skins 48 roles and options supported 105
with multiple responses 50 with negative response data 107
with patterns 50 BUBBLE statement, SGPLOT procedure
BARWIDTH= option BRADIUSMAX= option 54, 248
VBAR statement (SGPLOT) 50–51 BRADIUSMIN= option 248
VBARPARM statement (SGPLOT) 125 NOFILL option 54
BASELINE= option, NEEDLE statement bullet charts 190, 197
(SGPLOT) 96 business graphs
Berra, Yogi 310 about 311
BINSTART= option, HISTOGRAM danger of high P/E ratios 314
statement (SGPLOT) 164–165 financial trend and bond maturity graph
BINWIDTH= option, HISTOGRAM 311, 313
statement (SGPLOT) 164–165 oil consumption trend by country 315
Bonaparte, Napoleon 220 product sales and target graph 311,
bond maturity and financial trend graph 316
311, 313 social network 311, 317
stock price and volume chart 311–312
 Index 339

C CLM option
LOESS statement (SGPLOT)
Carr, Daniel 4
140–141, 157
categorization plots
PBSPLINE statement (SGPLOT)
See also bar charts
146–147
See also dot plots
REG statement (SGPLOT) 40, 132
See also line charts
CLMTRANSPARENCY= option, REG
about 181
statement (SGPLOT) 40
combining statements 29
CLOSE= option, HIGHLOW statement
roles and options supported 182
(SGPLOT) 53, 109
SGPANEL procedure 181, 258
CLUSTERWIDTH= option
SGPLOT procedure 22, 62, 181
HBOX statement (SGPLOT) 176
CATEGORYORDER= option
NEEDLE statement (SGPLOT) 96
discrete order and 229
SERIES statement (SGPLOT) 83
DOT statement (SGPLOT) 216
COLAXIS statement, SGPANEL procedure
HBAR statement (SGPLOT) 193
about 25–26, 33, 257, 259
VBAR statement (SGPLOT) 187
box plot row lattice example 55
VLINE statement (SGPLOT) 209
histogram lattice example 56
cell graphs 20, 307
color graphs 14–15
classification panels
COLUMN= option, PANELBY statement
See also SGPANEL procedure
(SGPANEL) 263
about 20–21, 37, 257–258
COLUMNLATTICE layout (SGPANEL
automatic paging 28
procedure) 28
bar chart panels 55
comparative and matrix graphs 37, 57
box plot row lattices 55
comparative and matrix plots
column lattice 265
See also SGSCATTER procedure
grouped bar chart 265
about 269
histogram lattice 56
comparative graph 277, 279
lattice by origin and type 262
grouped comparative graph 278
mileage by horsepower and type 261
grouped comparative plot 278
mileage by type 261
grouped scatter plots 284
paging of large panels 266
heat map 280
panel by origin and type 262
multiple plot requests 274
panel by type 263–264
plot grids 272
row lattice by type 263–264
plot with fit 273
scatter panels 56
plot with fit and ellipse 273
Cleveland, William S. 3
rectangular matrix 280
CLI option
scatter plot matrix 282, 284
PBSPLINE statement (SGPLOT)
scatter plot matrix with diagonals and
147–148
ellipse 283
REG statement (SGPLOT) 40,
scatter plot matrix with histograms 283
133–134
scatter plots 271
CLIP option, ELLIPSE statement (SGPLOT)
scatter plots with attributes 271
155
340 Index

comparative and matrix plots (continued) CURVELABELUPPER option, BAND

uniform side-by-side plots 275 statement (SGPLOT) 93
COMPARE statement, SGSCATTER
procedure
D
about 31, 269, 276 data labels
comparative graph 277, 279 dot plots with 214
grouped comparative graph 278 horizontal bar charts with 193
grouped comparative plot 278 horizontal line charts with 207
heat map 280 parametric bar charts with 101–102
options supported 276 scatter plots with 74–77, 135, 142, 149
rectangular matrix 280 series plots with 81
confidence plots vertical bar charts with 185
See fit and confidence plots vertical box plots with 173
CONNECT= option, VBOX statement vertical line charts with 200, 202
(SGPLOT) 173 waterfall charts with 119
Creating More Effective Graphs (Robbins) DATALABEL option
3 DOT statement (SGPLOT) 214
curve labels HBAR statement (SGPLOT) 193
horizontal line charts with 210 HLINE statement (SGPLOT) 207
images as 245–246 LOESS statement (SGPLOT) 142
Loess plots and 144 PBSPLINE statement (SGPLOT) 149
overlay band plots with 93 REG statement (SGPLOT) 135
overlay step plots with 88 SCATTER statement (SGPLOT)
penalized b-spline plots and 151 74–75, 156
regression plots and 137 SERIES statement (SGPLOT) 81
vertical line charts with 203 VBAR statement (SGPLOT) 47, 185
CURVELABEL= option VBARPARM statement (SGPLOT) 102
HLINE statement (SGPLOT) 210 VBOX statement (SGPLOT) 173
LINEPARM statement (SGPLOT) 116 VLINE statement (SGPLOT) 200
LOESS statement (SGPLOT) 144 DATALABELATTRS= option
PBSPLINE statement (SGPLOT) 151 DOT statement (SGPLOT) 214
REG statement (SGPLOT) 137 HBAR statement (SGPLOT) 193
SERIES statement (SGPLOT) 80, 123 HLINE statement (SGPLOT) 207
STEP statement (SGPLOT) 88 VBAR statement (SGPLOT) 185
VLINE statement (SGPLOT) 203 VLINE statement (SGPLOT) 200
CURVELABELLOC= option, LINEPARM DATALABELPOS= option
statement (SGPLOT) 116 VBAR statement (SGPLOT) 187
CURVELABELLOWER option, BAND VBARPARM statement (SGPLOT) 101
statement (SGPLOT) 93 VLINE statement (SGPLOT) 202
CURVELABELPOS= option DATASKIN= option
BAND statement (SGPLOT) 93 HBAR statement (SGPLOT) 192
LINEPARM statement (SGPLOT) 116 VBAR statement (SGPLOT) 50–51,
SERIES statement (SGPLOT) 123 184
 Index 341

VBARPARM statement (SGPLOT) 102 DISPLAY= option

decorative skins, effective graphics and 15 X2AXIS statement (SGPLOT) 77
DEGREE= option YAXIS statement (SGPLOT) 52
PBSPLINE statement (SGPLOT) 149 distribution plots
REG statement (SGPLOT) 40, 135 See also box plots
density plots See also density plots
about 166 See also histograms
density curve options 167 about 161
graphic example 161 combining 29, 177
histograms with 177 SGPANEL procedure 161, 258
normal density curve 167 SGPLOT procedure 22, 62, 161
roles and options supported 166 documents, small graphs in 333–334
DENSITY statement dot plots
SGPANEL procedure 25, 56, 257 about 212
SGPLOT procedure 23, 166–168 combining 218
Descartes, Rene 332 discrete order for 229
device-based graphics 33 graph examples 181, 213
discrete axis grouped 216
about 221 HLINE overlay 218
discrete order 229 overlaid 217
fit policy 229 overlaid with discrete offsets 217
reference line on 114 roles and options supported 212
scatter plots with clustered groups on SGPLOT procedure 22
72 with confidence limits 215
scatter plots with groups on 72 with data labels 214
DISCRETEMAX= option, ODS GRAPHICS with limits 46
statement 328–329 with marker attributes 213
DISCRETEOFFSET= option with reference line 214
DOT statement (SGPLOT) 217 with response sorting 216
HBAR statement (SGPLOT) 197 with upper limits 215
HLINE statement (SGPLOT) 211 DOT statement, SGPLOT procedure
REFLINE statement (SGPLOT) 77, about 212
114, 294 ALPHA= option 215
SCATTER statement (SGPLOT) 73 CATEGORYORDER= option 216
SERIES statement (SGPLOT) 82, 125 DATALABEL option 214
VBAR statement (SGPLOT) 50–51, DATALABELATTRS= option 214
190 DISCRETEOFFSET= option 217
VBARPARM statement (SGPLOT) 102 GROUP= option 216
VBOX statement (SGPLOT) 171 GROUPDISPLAY= option 212
VLINE statement (SGPLOT) 204 LEGENDLABEL= option 217
DISCRETEORDER= option, XAXIS LIMITS= option 215
statement (SGPLOT) 229 MARKERATTRS= option 213
NUMSTD= option 215
342 Index

DOWN= option, KEYLEGEND statement FILLATTRS= option, ELLIPSE statement

(SGPLOT) 231 (SGPLOT) 155
DPI financial trend and bond maturity graph
graph size and 336 311, 313
output destination and 327 fit and confidence plots
See also ellipse plots
E See also Loess plots
Edison, Thomas Alva 256 See also penalized b-spline plots
effective graphics See also regression plots
decorative skins and 15 about 129
principles of 3–9 combining 29, 157
Einstein, Albert 60, 160, 234, 326 comparing fits 157
ellipse plots comparing fits with CLM 157
about 152 fit plots 39
attribute control 155 fit plots with confidence limits 40
clipping 154 fit plots with transparent markers 39
filled 153 grouped fit plot with limits 40
graph examples 129, 153, 273 roles and options 130
overlaid 156 SGPANEL procedure 129, 258
overlaid filled 156 SGPLOT procedure 22, 62, 129
roles and options supported 152 FITPOLICY= option, XAXIS statement
scatter plot matrix with 283 (SGPLOT) 119, 229
setting type 155 floating bars and bands 125
with confidence level 154 FOOTNOTE statement 33, 245
ELLIPSE statement, SGPLOT procedure forest plots 241–242, 288–289
about 152 FREQ= option
ALPHA= option 153–154 HISTOGRAM statement (SGPLOT)
CLIP option 154 163
FILL option 153, 156 SCATTER statement (SGPLOT) 81
FILLATTRS= option 155
G
LINEATTRS= option 155
OUTLINE option 153 Gates, W.I.E. 2
TYPE= option 155 GCHART procedure 33
ERRORBARATTRS= option, STEP GOPTIONS global statement 33
statement (SGPLOT) 45 GPLOT procedure 33
eye irritation, distribution of graph output
using SGPANEL procedure 303 about 333
using SGPLOT procedure 304 combining graph size and DPI 336
creating large graphs for presentations
F 334–335
FILL option, ELLIPSE statement (SGPLOT) creating small graphs for documents
153, 156 333–334
 Index 343

impact of graph size on system SCATTER statement (SGPLOT) 72

resources 336 SERIES statement (SGPLOT) 83
Graph Template Language (GTL) 11 VBAR statement (SGPLOT) 48, 51,
graphs 188
See also business graphs VBOX statement (SGPLOT) 170
See also classification panels GROUPMAX= option, ODS GRAPHICS
See also health and life sciences graphs statement 328–329
See also single-cell graphs GSEG output format 33
about 20, 37 GTL (Graph Template Language) 11
automatically producing from
procedures 10–11
H
background images in 244 HBAR statement, SGPLOT procedure
color 14–15 about 191
comparative and matrix 37, 57 ALPHA= option 194
decorative skins and 15 CATEGORYORDER= option 193
Graph Template Language 11 DATALABEL option 193
gray-scale 14–15 DATALABELATTRS= option 193
principles of effective graphics 3–9 DATASKIN= option 192
statistical graphics procedures and DISCRETEOFFSET= option 197
11–12 GROUP= option 195
gray-scale graphs 14–15 GROUPDISPLAY= option 191, 195
GRID option LIMITS= option 52, 194
PLOT statement (SGPANEL) 275 NOFILL option 192
X2AXIS statement (SGPLOT) 134, NOOUTLINE option 192
141, 148 NUMSTED= option 194
Y2AXIS statement (SGPLOT) 134, STAT= option 52
141, 148 HBARPARM statement, SGPLOT procedure
GROUP= option about 100
DOT statement (SGPLOT) 216 LIMITATTRS= option 104
HBAR statement (SGPLOT) 195 with statistics 126
HBOX statement (SGPLOT) 176 HBOX statement, SGPLOT procedure
HLINE statement (SGPLOT) 210 about 174
LOESS statement (SGPLOT) 143 CLUSTERWIDTH= option 176
PBSPLINE statement (SGPLOT) 150 GROUP= option 176
REG statement (SGPLOT) 136 LINEATTRS= option 176
SCATTER statement (SGPLOT) 65, NOTCHES option 175
72, 76 health and life sciences graphs
style elements and 323 about 287
VBAR statement (SGPLOT) 51, 188 adverse event timelines 291–292
GROUPDISPLAY= option concomitant medications 306
DOT statement (SGPLOT) 212 creating 2 x 2 cell graph using SGPLOT
HBAR statement (SGPLOT) 191, 195 procedure 307
HLINE statement (SGPLOT) 205
344 Index

health and life sciences graphs (continued) HIGH= option 109

distribution of eye irritation using HIGHCAP option 53
SGPANEL procedure 303 HIGHLABEL= option 110
distribution of eye irritation using LOW= option 109
SGPLOT procedure 304 LOWCAP option 53
forest plots 288–289 LOWLABEL= option 110–111, 292
immunology profile by treatment 299 OPEN= option 53, 109
LFT patient profile 301 TYPE= option 53, 109
LFT safety panel, baseline vs. study HISTOGRAM statement
298 SGPANEL procedure 25, 56, 257
maximum LFT values by treatment 293 SGPLOT procedure 23, 41, 162–165
median of lipid profile over time by histograms
treatment 294 about 162
most frequent on-therapy adverse classification panel example 56
events by frequency 300 graph examples 41, 161, 163
panel of LFT values 302 multiple 165
QTc change from baseline over time by roles and options supported 162
treatment 295 scatter plot matrix with 283
QTc change graph with annotated at SGPLOT procedure 22
risk values 296 with BINSTART 164
survival plots 290 with BINWIDTH 164
vital signs by time point name 305 with density plots 41
heat maps 280 with frequency 163
HEIGHT= option, ODS GRAPHICS with NBINS 164
statement 15, 328 with normal density plots 177
HIGH= option, HIGHLOW statement HLINE statement, SGPLOT procedure
(SGPLOT) 109 about 205
HIGHCAP option, HIGHLOW statement ALPHA= option 208
(SGPLOT) 53 BREAK option 207
HIGHLABEL= option, HIGHLOW statement CURVELABEL= option 210
(SGPLOT) 110 DATALABEL option 207
HighLow bar charts 110–111 DATALABELATTRS= option 207
HighLow plots DISCRETEOFFSET= option 211
about 108 dot plot overlay 218
bar chart examples 110–111 GROUP= option 210
graph examples 53, 61, 109 LEGENDLABEL= option 211
roles and options supported 108 LIMITS= option 209
vertical vector plots and 99 LINEATTRS= option 206
with end caps 110 MARKERATTRS= option 206
with open and close values 109 MARKERS option 206
HIGHLOW statement, SGPLOT procedure NUMSTD= option 209
about 108 horizontal bar charts
CLOSE= option 53, 109 about 191
 Index 345

bullet charts 197 IMAGENAME= option, ODS GRAPHICS

graph examples 52, 181 statement 328–329
grouped using SGPANEL procedure images as curve labels 245–246
196 immunology profile by treatment 299
overlaid with discrete offset 197 Information Visualization (Carr) 4
roles and options supported 191 INSET statement, SGPLOT procedure 24,
with adjacent groups 195 85, 106, 222
with confidence limits 194 insets
with data labels 193 about 222
with fill attributes and data skins 192 options supported 222
with groups 195 rich-text 232
with groups and pattern fills 196 SGPLOT procedure 22
with limits 52 types of 232
with no fill 192 INTEGER option, YAXIS statement
with reference line 193 (SGPLOT) 224
with upper limit 194 INTERPOLATION= option, LOESS
horizontal box plots statement (SGPLOT) 142
about 174 interval box plots 43
graph examples 161, 175 INTERVAL= option, XAXIS statement
grouped 176 (SGPLOT) 227
grouped with CLUSTERWIDTH 176
roles and options supported 174
J
with notches 175 JOIN option, PLOT statement (SGPANEL)
horizontal line charts 275
about 205 JPG output format 33
graph examples 181, 206 JUSTIFY= option, STEP statement
grouped 210 (SGPLOT) 86
overlaid 211
overlaid with discrete offset 211 K
roles and options supported 205 Kay, Alan 18
with break 207 KEYLEGEND statement, SGPANEL
with confidence limits 208 procedure 33, 221, 259
with curve labels 210 KEYLEGEND statement, SGPLOT
with data labels 207 procedure
with markers 206 about 23–24, 33, 221
with reference line 208 ACROSS= option 231
with response sorting 209 bubble plot examples 106
with upper limits 209 custom legends 230
I DOWN= option 231
fit plot examples 40
IMAGE function 236 LOCATION= option 231
IMAGEMAP option, ODS GRAPHICS POSITION= option 231
statement 328–329 scatter plot examples 38, 65–66, 121
346 Index

KEYLEGEND statement, SGPLOT grouped series plots with 79

procedure (continued) LOCATION option 231
series plot examples 79–80 options supported 222
step plot examples 85 POSITION option 231
VALUEATTRS= option 122 SGPLOT procedure 22
LFT patient profile 301
L LFT safety panel, baseline vs. study 298
LABEL= option LFT values, panel of 302
XAXIS statement (SGPLOT) 206, 213 life sciences graphs
YAXIS statement (SGPLOT) 199 See health and life sciences graphs
LABELMAX= option, ODS GRAPHICS LIFETEST procedure 290
statement 328–329 LIMITATTRS= option, HBARPARM
LABELPOS= option, REFLINE statement statement (SGPLOT) 104
(SGPLOT) 113 LIMITS= option
LATTICE layout (SGPANEL procedure) 27 DOT statement (SGPLOT) 215
LAYOUT= option, PANELBY statement HBAR statement (SGPLOT) 52, 194
(SGPANEL) HLINE statement (SGPLOT) 209
about 257 VBAR statement (SGPLOT) 186
box plot row lattice 55 VLINE statement (SGPLOT) 202
grouped bar chart 189, 196 line charts
lattice by origin and type 262 bar-line overlay 49, 218
paging of large panels 266 discrete order for 229
row lattice by type 263 horizontal 181
LEGEND global statement 33 principles of effective graphics and 9
LEGEND option, SGSCATTER procedure vertical 181, 198–204
33, 222 LINE function 236, 248
LEGENDLABEL= option line plots, parametric 115–116
BAND statement (SGPLOT) 92 linear axis
custom legends and 230 about 221
DOT statement (SGPLOT) 217 specifying MIN/MAX 223
HLINE statement (SGPLOT) 211 specifying tick lists 223
PBSPLINE statement (SGPLOT) 151 TICKVALUEFORMAT option 228
REG statement (SGPLOT) 137 VALUESHINT option 224
SCATTER statement (SGPLOT) 71 LINEATTRS= option
VLINE statement (SGPLOT) 204 ELLIPSE statement (SGPLOT) 155
legends HBOX statement (SGPLOT) 176
about 20, 221–222 HLINE statement (SGPLOT) 206
ACROSS option 231 NEEDLE statement (SGPLOT) 95
automatic 230 REG statement (SGPLOT) 39
bubble 248 STEP statement (SGPLOT) 45, 86
custom 230 VECTOR statement (SGPLOT) 98
disabling 133, 140, 147 VLINE statement (SGPLOT) 49, 199
DOWN option 231 LineParm plots 61
 Index 347

LINEPARM statement, SGPLOT procedure LOGBASE= option, YAXIS statement

CURVELABEL= option 116 (SGPLOT) 226
CURVELABELLOC= option 116 LOGSTYLE= option, YAXIS statement
CURVELABELPOS= option 116 (SGPLOT) 226
lipid profile, median of, over time by LOW= option, HIGHLOW statement
treatment 294 (SGPLOT) 109
LOCATION= option, KEYLEGEND LOWCAP option, HIGHLOW statement
statement (SGPLOT) 231 (SGPLOT) 53
Loess plots LOWER= option, BAND statement
about 138 (SGPLOT) 92
attribute control 141 LOWLABEL= option, HIGHLOW statement
automatic legend 140 (SGPLOT) 110–111, 292
axis assignment 141
basic fit 139
M
CLM band 140 magnitude perception, accuracy of 6
cubic Loess fit 142 MARKERATTRS= option
curve labels 144 DOT statement (SGPLOT) 213
fit plot with smooth curves 139 HLINE statement (SGPLOT) 206
graph examples 129 SCATTER statement (SGPLOT) 66–67
grouped fits 143 VLINE statement (SGPLOT) 199
overlaid fits 144 MARKERCHAR= option, SCATTER
roles and options supported 138 statement (SGPLOT)
scatter data labels 142 about 74
SGPLOT procedure 22 bubble plot with negative response data
single fit with a group 143 107
LOESS statement, SGPLOT procedure forest plots 288
ALPHA= option 140 HBARPARM with statistics 126
CLM option 140–141, 157 median of lipid profile over time by
CURVELABEL= option 144 treatment 294
DATALABEL option 142 scatter with axis-aligned statistics table
GROUP= option 143 77
INTERPOLATION= option 142 scatter with marker characters 77
NOLEGFIT option 140 VBARPARM with statistics 126
NOMARKERS option 139, 143–144 MARKERS option, HLINE statement
SMOOTH= option 139 (SGPLOT) 206
log axis matrix graphs 37, 57
about 221 matrix plots
LINEAR style 226 See comparative and matrix plots
log expand style 225 MATRIX statement, SGSCATTER
log exponent style 225 procedure
MINOR option 228 about 31, 269, 281
specifying axis MIN/MAX 223 grouped scatter plots 284
options supported 281
348 Index

MATRIX statement, SGSCATTER BUBBLE statement (SGPLOT) 54

procedure (continued) HBAR statement (SGPLOT) 192
scatter plot matrix 282, 284 HISTOGRAM statement (SGPLOT)
scatter plot matrix with diagonals and 164–165
ellipse 283 VBAR statement (SGPLOT) 184
scatter plot matrix with histograms 283 VBOX statement (SGPLOT) 171
START= option 284 NOLEGFIT option
MAX= option, XAXIS statement (SGPLOT) LOESS statement (SGPLOT) 140
223 PBSPLINE statement (SGPLOT) 147
MAXLEGENDAREA= option, ODS REG statement (SGPLOT) 133
GRAPHICS statement 328–329 NOMARKERS option
medications, concomitant 306 LOESS statement (SGPLOT) 139,
MIN= option 143–144
XAXIS statement (SGPLOT) 223 PBSPLINE statement (SGPLOT) 146,
YAXIS statement (SGPLOT) 79, 85 150–151
MINOR option REG statement (SGPLOT) 132,
XAXIS statement (SGPLOT) 228 136–137
YAXIS statement (SGPLOT) 225 NOTCHES option
MODELNAME= option, BAND statement HBOX statement (SGPLOT) 175
(SGPLOT) 90, 93 VBOX statement (SGPLOT) 173
NOTIMESPLIT option, XAXIS statement
N (SGPLOT) 227
needle plots NOUTLINE option
about 94 HBAR statement (SGPLOT) 192
graph examples 61, 95 VBAR statement (SGPLOT) 184
grouped 95 NOVARNAME option, PANELBY statement
roles and options supported 94 (SGPANEL) 261
with baseline 96 NUMSTD= option
with discrete groups 96 DOT statement (SGPLOT) 215
NEEDLE statement, SGPLOT procedure HBAR statement (SGPLOT) 194
BASELINE= option 96 HLINE statement (SGPLOT) 209
CLUSTERWIDTH= option 96 VBAR statement (SGPLOT) 186
LINEATTRS= option 95 VLINE statement (SGPLOT) 202
Nietzsche, Friedrich 180
O
NOARROWHEADS option, NEEDLE
statement (SGPLOT) 99 ODS destination options 32, 321, 327
NOAUTOLEGEND option, SGPLOT ODS ESCAPECHAR option 232, 238
procedure 68, 107, 133, 140, 147 ODS GRAPHICS statement
NOBORDER option, PANELBY statement ANTIALIAS option 328
(SGPANEL) 189, 196 ANTIALIASMAX= option 44, 79, 328
NOCYCLEATTRS option, SGPLOT BORDER option 328
procedure 68 DISCRETEMAX= option 328–329
NOFILL option GROUPMAX= option 328–329
 Index 349

HEIGHT= option 15, 328 PANELBY statement, SGPANEL procedure

IMAGEMAP option 328–329 about 24–25, 257, 259
IMAGENAME= option 328–329 COLUMN= option 263
LABELMAX= option 328–329 LAYOUT= option 55, 189, 196, 257,
MAXLEGENDAREA= option 328–329 266
OUTPUTFMT= option 328–329 NOBORDER option 189, 196
PANELCELLMAX= option 328–329 NOVARNAME option 261
RESET option 328–329 ONEPANEL option 263
SCALE option 328–329 roles and options supported 260
TIPMAX= option 328, 330 UNISCALE= option 264
WIDTH= option 15, 328, 330 PANELCELLMAX= option, ODS GRAPHICS
ODS Graphics system statement 328–329
annotation and 236 parametric bar charts
automatically producing graphs 10–11 graph examples 101
options supported 328–330 grouped 102
SG procedures and 19 grouped with data labels 102
template-based graphs and 33 horizontal 104
ODS SELECT statement 10 horizontal with limits 104
OFFSETMAX= option overlay 103
annotations and 237 overlay with skins and offsets 103
YAXIS statement (SGPLOT) 46, 95, vertical 101–102
102, 173 with limits and data labels 101
OFFSETMIN= option parametric line plots 115–116
annotations and 237 Pareto chart 125
XAXIS statement (SGPLOT) 71, 114 PATTERN global statement 33
YAXIS statement (SGPLOT) 118 PBSPLINE statement, SGPLOT procedure
oil consumption trend by country 315 ALPHA= option 146–147
ONEPANEL option, PANELBY statement CLI option 147–148
(SGPANEL) 263 CLM option 146–147
OPEN= option, HIGHLOW statement CURVELABEL= option 151
(SGPLOT) 53, 109 DATALABEL option 149
OUTLINE option, ELLIPSE statement DEGREE= option 149
(SGPLOT) 153 GROUP= option 150
OUTPUTFMT= option, ODS GRAPHICS LEGENDLABEL= option 151
statement 328 NOLEGFIT option 147
OVAL function 236, 248 NOMARKERS option 146, 150–151
PDF output format 33
P penalized b-spline plots
P/E ratios 314 about 145
PAD= option, SGPLOT procedure 238, 240 attribute control 148
paging of large panels 266 automatic legend 147
PANEL layout (SGPANEL procedure) axis assignments 148
26–27 basic parametric 146
350 Index

penalized b-spline plots (continued) short-term memory and 9

changing degree of 149 simplifying and reducing clutter 8
CLI band 147 usage of 3-D graphs 7
CLM band 146 visual perception 4–5
curve labels 151 procedures, automatic graphs from 10–11
graph examples 129, 273 product sales and target graph 311, 316
grouped fits 150 PS output format 33
overlaid fits 151
roles and options supported 145
Q
scatter data labels 149 QTc change from baseline over time by
single fit within a group 150 treatment 295
pie charts 3–4, 6 QTc change graph with annotated at risk
PLOT statement, SGSCATTER procedure values 296
about 30, 269–270
GRID option 275 R
JOIN option 275 RECTANGLE function 236, 248
options supported 270 reference line
scatter plots 271 about 112
UNISCALE= option 272, 275 dot plots with 214
plot statements graph examples 61
See SGPLOT procedure horizontal bar charts with 193
See specific statements horizontal line charts with 208
plots multiple 114
See also reference line on discrete axis 114
See also SGPLOT procedure roles and options supported 112
See also specific types of plots scatter plots with 113
about 61 SGPANEL procedure 25
combining 29, 120–126 SGPLOT procedure 22
SGPANEL procedure 258 vertical bar charts with 185
waterfall charts 117–119 vertical line charts with 201
PNG output format 33 REFLINE statement, SGPANEL procedure
POLYCONT function 235–236 259
POLYGON function 235–236 REFLINE statement, SGPLOT procedure
POLYLINE function 235–236, 249–250 about 24, 71
POSITION= option, KEYLEGEND statement AXIS= option 114
(SGPLOT) 231 bar charts and 185, 193
presentations, large graphs in 334–335 DISCRETEOFFSET= option 77, 114,
principles of effective graphics 294
about 3–4 dot plots and 214
accuracy of magnitude perception 6 LABELPOS= option 113
additional information 3 line charts and 201, 208
proximity increases accuracy of step and scatter plot example 122
comparisons 8 REG procedure 10, 134
 Index 351

REG statement, SGPLOT procedure S

ALPHA= option 132–133
CLI option 40, 133–134 SAS/GRAPH 236
CLM option 40, 132 SCALE option
CLMTRANSPARENCY= option 40 ODS GRAPHICS statement 328–329
CURVELABEL= option 137 SCALE= option
DATALABEL option 135 DENSITY statement (SGPLOT) 167
DEGREE= option 40, 135 scatter plot matrix
GROUP= option 136 graph examples 57, 282, 284
LEGENDLABEL= option 137 with diagonals and ellipse 283
LINEATTRS= option 39 with histograms 283
NOLEGFIT option 133 scatter plots
NOMARKERS option 132, 136–137 about 64
regression plots classification panel example 56
about 131 comparative 57
attribute control 134 graph examples 38, 61, 65, 271
automatic legend 133 grouped 65, 284
axis assignment 134 grouped with grid lines 38
basic linear fit 132 grouped with X and Y limits 69
CLI band 133 overlaid fits 137, 144, 151
CLM band 132 overlay of 73, 121
curve labels 137 overlay with band plots 123
graph examples 129 overlay with discrete offset 73
grouped fits 136 overlay with series and band plots 122
overlaid fits 137 overlay with step plots 122
polynomial regression fit 135 roles and options supported 64
roles and options supported 131 SGPLOT procedure 22
scatter data labels 135 single fit with a group 136, 143, 150
SGPLOT procedure 22 two plot overlay 66
single fit with a group 136 with attributes 271
RESET option with clustered groups on discrete axis
GOPTIONS statement 33 72
ODS GRAPHICS statement 328–329 with custom attributes 66
REVERSE option with data labels 74–77, 135, 142, 149
XAXIS statement (SGPLOT) 197 with groups on discrete axis 72
YAXIS statement (SGPLOT) 190 with lower and upper limits for X and Y
Robbins, Naomi 3 68
ROWAXIS statement, SGPANEL procedure with reference line 113
about 25–26, 33, 259 with reflected ticks on Y2 71
histogram lattice example 56 with X, Y, and Y2 axis 71
ROWLATTICE layout (SGPANEL with X2 and Y2 axes 70
procedure) 28 with X lower and upper error limits 68
with Y2 axis 70
352 Index

scatter plots (continued) about 11

with Y lower and upper error limits 67 analysis of data 12
with Y upper error limit 67 annotation support 237
SCATTER statement, SGPLOT procedure automatic procedures 21
about 39 combining statements 29
DATALABEL option 74–75, 156 key concepts 19–21
DISCRETEOFFSET= option 73 legends and 230
FREQ= option 81 post-analysis data presentation 12
GROUP= option 65, 72, 76 pre-analysis data exploration 11
LEGENDLABEL= option 71 styles and their usage 32
MARKERATTRS= option 66–67 template-based vs. device-based
MARKERCHAR= option 74, 77, 107, graphics 33
126, 288, 294 SGANNO= option, SGPLOT procedure
overlaid fits 137, 144, 151 235
single fit with a group 136, 143, 150 SGPANEL procedure
series plots See also classification panels
graph examples 43, 61, 79 See also COLAXIS statement,
grouped 44 SGPANEL procedure
grouped with curved labels 80 See also PANELBY statement,
grouped with legends 79 SGPANEL procedure
grouped with markers 80–81 about 19, 24, 258–259
multi-response 44 annotation support 237
overlay with band plots 123 automatic paging 28
overlay with discrete offset 82 axis support 221
overlay with scatter and band plots 122 COLUMNLATTICE layout 28
overlay with scatter plots 121 DENSITY statement 25, 56, 257
roles and options supported 78 distribution of eye irritation using 303
SGPLOT procedure 22 fit and confidence plots 129
with breaks 82 grouped bar charts 189
with data labels 81 HISTOGRAM statement 25, 56, 257
with group cluster 83 KEYLEGEND statement 33, 221, 259
with group overlay 83 LATTICE layout 27
SERIES statement, SGPLOT procedure PANEL layout 26–27
BREAK option 82 plots supporting 61
CLUSTERWIDTH= option 83 REFLINE statement 259
CURVELABEL= option 80, 123 ROWAXIS statement 25–26, 33, 56,
CURVELABELPOS= option 123 259
DATALABEL option 81 ROWLATTICE layout 28
DISCRETEOFFSET= option 82, 125 SGPLOT procedure
GROUPDISPLAY= option 83 See also KEYLEGEND statement,
Pareto chart 125 SGPLOT procedure
SG (Statistical Graphics) procedures See also REFLINE statement, SGPLOT
See also specific procedures procedure
 Index 353

See also SCATTER statement, VBAR statement 47–48, 50–51,

SGPLOT procedure 183–190
See also XAXIS statement, SGPLOT VBARPARM statement 100–103,
procedure 125–126
See also YAXIS statement, SGPLOT VBOX statement 169–173, 295
procedure VECTOR statement 98–99
about 19, 22, 62 VLINE statement 49, 198–204
annotation support 237, 295 WHERE statement 43
axis support 221 X2AXIS statement 24, 33, 77
BAND statement 90, 92–93, 124 Y2AXIS statement 24, 33
BUBBLE statement 54, 248 SGRENDER procedure 11
creating cell graphs with 307 SGSCATTER procedure
DENSITY statement 23, 166–168 about 19, 30
distribution of eye irritation using 304 annotation support 237
DOT statement 212–217 comparative and matrix graphs and 21,
fit and confidence plots 129 37
HBAR statement 52, 191–197 COMPARE statement 31, 269,
HBARPARM statement 100, 104, 126 276–280
HBOX statement 174–176 controlling axes 33
HIGHLOW statement 53, 108–111, LEGEND option 33, 222
292 MATRIX statement 31, 269, 281–284
HISTOGRAM statement 23, 41 options supported 269
HLINE statement 205–211 PLOT statement 30, 269–275
INSET statement 24, 85, 106 SHOWBINS option, HISTOGRAM statement
LINEPARM statement 115–116 (SGPLOT) 41
LOESS statement 139–144, 157 SIGMA= option, DENSITY statement
NEEDLE statement 95–96 (SGPLOT) 168
NOAUTOLEGEND option 68, 107, single-cell graphs
133, 140, 147 about 37
NOCYCLEATTRS option 68 bar-bar overlay 49
options supported 62 bar charts 47
PAD= option 238, 240 bar charts with multiple responses 50
PBSPLINE statement 146–151 bar charts with patterns 50
plot roles and options 63 bar-line overlay 49
plot statements supported 62 basic scatter plot 38
plots supporting 61 bubble plots 54
REG statement 39–40, 132–137 dot plots with limits 46
SERIES statement 80–83, 123, 125 fit plots 39
SGANNO= option 235 fit plots with confidence limits 40
single-cell graphs and 19–20, 37 fit plots with transparent markers 39
STEP statement 45, 86, 88 grouped bar charts with skins 48, 51
typical use case 22–23 grouped box plots 42
grouped fit plot with limits 40
354 Index

single-cell graphs (continued) roles and options supported 84

grouped scatter plot with grid lines 38 with limits 45
grouped series plots 44 STEP statement, SGPLOT procedure
grouped step plots 45 BREAK option 88
HighLow plots 53 CURVELABEL= option 88
histograms 41 ERRORBARATTRS= option 45
histograms with density plots 41 JUSTIFY= option 86
horizontal bar charts 52 LINEATTRS= option 45, 86
horizontal bar charts with limits 52 Steven's power law 6
interval box plots 43 stock price and volume chart 311–312
multi-response series plots 44 style elements
multiple bar charts with patterns, fill about 321–323
colors, skins 51 usage considerations 324
series plots 43 usage precedence 324
SGPLOT procedure and 19–20, 37 STYLE= option (ODS destination) 32, 321
stacked bar charts with skins 48 styles
step plots with limits 45 about 321
vector plots 46 appearance options 32
vertical box plots 42 custom 32
SMOOTH= option, LOESS statement pre-defined 32
(SGPLOT) 139 SG procedures and 32
social network (business graphs) 311, 317 survival plots 290
START= option, MATRIX statement SYMBOL global statement 33
(SCSCATTER) 284 system resources, impact of graph size on
STAT= option, HBAR statement (SGPLOT) 336
52
Statistical Graphics procedures
T
See SG (Statistical Graphics) template-based graphics 33
procedures TEMPLATE procedure 322
step plots TEXT function 236, 248
about 84 TEXTCONT function 236
graph examples 61, 85 3-D graphs, usage considerations 7
grouped 45, 85 tick lists, specifying 223
grouped with center justified markers TICKVALUEFORMAT= option, XAXIS
86 statement (SGPLOT) 228
grouped with limits and error bar time axis
attributes 87 about 221
grouped with markers 86 INTERVAL option 227
grouped with upper limit 87 MINOR option 228
overlay with band plots 93 NOTIMESPLIT option 227
overlay with break 88 specifying axis MIN/MAX 223
overlay with curve labels 88 specifying tick lists 223
overlay with scatter plots 122 TICKVALUEFORMAT option 228
 Index 355

VALUESHINT option 224 DATALABELPOS= option 187

TIPMAX= option, ODS GRAPHICS DATASKIN= option 50–51, 184
statement 328, 330 DISCRETEOFFSET= option 50–51,
TITLE statement 33 190
TRANSPARENCY= option, BAND GROUP= option 51, 188
statement (SGPLOT) 124 GROUPDISPLAY= option 48, 51, 188
treatment LIMITS= option 186
immunology profile by 299 NOFILL option 184
maximum LFT values by 293 NOUTLINE option 184
median of lipid profile over time by 294 NUMSTD= option 186
QTc change from baseline over time by VBARPARM statement, SGPLOT procedure
295 about 100
Tufte, Edward 3, 8 BARWIDTH= option 125
TYPE= option DATALABEL option 102
DENSITY statement (SGPLOT) 168 DATALABELPOS= option 101
ELLIPSE statement (SGPLOT) 155 DATASKIN= option 102
HIGHLOW statement (SGPLOT) 53, DISCRETEOFFSET= option 103
109 Pareto chart 125
XAXIS statement (SGPLOT) 43, 172 with statistics 126
VBOX statement, SGPLOT procedure
U about 169
Unicode characters 232, 238 BOXWIDTH= option 171
UNISCALE= option CONNECT= option 173
PANELBY statement (SGPANEL) 264 DATALABEL option 173
PLOT statement (SGPANEL) 272, 275 DISCRETEOFFSET= option 171
GROUPDISPLAY= option 170
V NOFILL option 171
VALUEATTRS= option NOTCHES option 173
KEYLEGEND statement (SGPLOT) QTc change from baseline over time by
122 treatment 295
XAXIS statement (SGPLOT) 125 vector plots 46–47, 61, 97–99
YAXIS statement (SGPLOT) 125 VECTOR statement, SGPLOT procedure
VALUES= option, XAXIS statement ARROWHEADSHAPE= option 99
(SGPLOT) 223 LINEATTRS= option 98
VALUESHINT option, XAXIS statement NOARROWHEADS option 99
(SGPLOT) 224 XORIGIN= option 98
VBAR statement, SGPLOT procedure YORIGIN= option 98
about 183 vertical bar charts
ALPHA= option 186 bar-bar overlay 49
BARWIDTH= option 50–51 bar-line overlay 49
CATEGORYORDER= option 187 bullet charts 190
DATALABEL option 47, 185 graph examples 47, 181
DATALABELATTRS= option 185
356 Index

vertical bar charts (continued) with CLM and data label position 202
grouped using SGPANEL procedure with confidence limits 201
189 with curve labels 203
grouped with data labels 102 with data labels 200
grouped with skins 48, 51 with markers 199
multiple with patterns, fill colors, skins with reference line 201
51 with upper limits 202
overlaid with discrete offset 190 vertical vector plots 99
overlay 103 The Visual Display of Quantitative
overlay with skins and offsets 103 Information (Tufte) 3
roles and options supported 183 visual perception, effective graphics and
stacked with skins 48 4–5
with adjacent groups 188 Visualizing Data (Cleveland) 3
with an upper limit 186 vital signs by time point name 305
with confidence limits 186 VLINE statement, SGPLOT procedure
with data labels 185 about 198
with fill attributes and data skins 184 ALPHA= option 201
with groups and pattern fills 189 BREAK option 200
with limits and label positioning 187 CATEGORYORDER= option 209
with no fill 184 CURVELABEL= option 203
with patterns 50 DATALABEL option 200
with reference line 185 DATALABELATTRS= option 200
with response sorting 187 DATALABELPOS= option 202
with stacked groups 188 DISCRETEOFFSET= option 204
vertical box plots LEGENDLABEL= option 204
about 169 LIMITS= option 202
graph examples 42, 161, 170 LINEATTRS= option 49, 199
grouped 170 MARKERATTRS= option 199
grouped unfilled 171 NUMSTD= option 202
on linear axis 172
overlay 171
W
overlay for linear data 172 waterfall charts
roles and options supported 169 about 117
with labels 173 graph examples 118
with notches 173 grouped 119
vertical line charts roles and options supported 117
about 198 with data labels 119
graph examples 181, 199 with initial value 118
grouped 203 WATERFALL statement, SGPLOT
overlaid 204 procedure 117
overlaid with discrete offset 204 WHERE statement, SGPLOT procedure 43
roles and options supported 198 WIDTH= option, ODS GRAPHICS statement
with break 200 15, 328, 330
 Index 357

X Yoda 128
YORIGIN= option, NEEDLE statement
XAXIS statement, SGPLOT procedure (SGPLOT) 98
about 23–24, 33
DISCRETEORDER= option 229
FITPOLICY= option 119, 229
INTERVAL= option 227
LABEL= option 206, 213
MAX= option 223
MIN= option 223
MINOR option 228
NOTIMESPLIT option 227
OFFSETMIN= option 71, 114
REVERSE option 197
TICKVALUEFORMAT= option 228
TYPE= option 43, 172
VALUEATTRS= option 125
VALUES= option 223
VALUESHINT option 224
X2AXIS statement, SGPLOT procedure
about 24, 33
DISPLAY option 77
GRID option 134, 141, 148
XORIGIN= option, NEEDLE statement
(SGPLOT) 98
Y
YAXIS statement, SGPLOT procedure
about 24, 33
DISPLAY= option 52
INTEGER option 224
LABEL= option 199
LOGBASE= option 226
LOGSTYLE= option 226
MIN= option 79, 85
MINOR option 225
OFFSETMAX= option 46, 95, 102, 173
OFFSETMIN= option 118
REVERSE option 190
VALUEATTRS= option 125
Y2AXIS statement, SGPLOT procedure
about 24, 33
GRID option 134, 141, 148