Working With JMP

Release 8
Introductory Guide
The real voyage of discovery consists not in seeking new
landscapes, but in having new eyes.
Marcel Proust
JMP, A Business Unit of SAS

SAS Campus Drive
Cary, NC 27513
The correct bibliographic citation for this manual is as follows: SAS Institute Inc. 2008. JMP 8 Introductory
Guide. Cary, NC: SAS Institute Inc.
JMP 8 Introductory Guide
Copyright 2008, SAS Institute Inc., Cary, NC, USA
ISBN 978-1-59994-920-8
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.
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.
1st printing, October 2008
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 hardcopy 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
JMP Introductory Guide
1
Introducing JMP
Your First Look . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
What You Need to Know . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Learning About JMP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Using Tutorials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Searching in the Help . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Learning About Statistical and JSL Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Using the Context-Sensitive Help . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
Learning JMP Tips & Tricks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Using This Book in Combination with Other Included Books . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Conventions Used in this Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Step 1: Start JMP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Step 2: Open a JMP Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Step 3: Learn About the Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Specifying the Values Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Data Table Cursor Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Selecting Rows and Columns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Step 4: Select an Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Casting Columns Into Roles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Step 5: View the Output Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Graphs and Charts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Statistical Tables and Text . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Step 6: Save the JMP Output Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
A Practice Tutorial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Open a Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
Select an Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Spin the Cowboy Hat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Creating a JMP Data Table

Entering and Plotting Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Starting a JMP Session . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Creating Rows and Columns in a JMP Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Add Columns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
21
23
23
24
ii
Set Column Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Add Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Entering Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Plotting Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Document the Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Summarizing Data
Look Closely at the Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Look Before You Leap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Grouping Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Creating Statistics for Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Charting Statistics from Grouped Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Charting Statistics for Two Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Finding a Subgroup with Multiple Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Comparative Scatterplots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
What Has Been Discovered? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Finding the Best Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
31
33
34
35
36
38
39
39
41
41
42
Looking at Distributions
Histograms, moments, quantiles, and proportions . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Look Before You Leap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Displaying Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Understanding Histograms of Nominal and Ordinal Variables . . . . . . . . . . . . . . . . . . . . . . . .
Understanding Histograms of Continuous Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Learning About Report Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Reports for Continuous Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Frequency Table for Ordinal or Nominal Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Adding a Computed Column . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Creating Subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
25
25
25
26
29
30
45
47
47
48
49
51
51
52
52
55
56
Comparing Group Means

Testing Differences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Graphical Display of Grouped Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Choose Variable Roles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Show Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fit Means Option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fit Quantiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Comparison Circles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
57
59
59
60
60
61
63
64
iii
Analyzing Categorical Data

Comparing Proportions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Open a Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Address the Research Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Modify the Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Contingency Table Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Cast Variables Into Roles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Contingency Table Mosaic Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
69
71
71
71
71
75
75
75
78
Regression and Curve Fitting

Visualizing Relationships . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Open a JMP File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Select an Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Choose Variable Roles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fitting Models to Continuous Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fitting the Mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fitting a Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Understanding the Summary of Fit Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Understanding the Analysis of Variance Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Understanding the Parameter Estimates Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Excluding Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Journaling JMP Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Examining a Polynomial Fit (Linear Regression) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fitting a Spline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fitting By Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
65
65
66
67
81
83
83
83
83
84
85
85
86
87
87
88
88
90
91
93
94
A Factorial Analysis
Designed Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Look Before You Leap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Open a Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
What Questions Can Be Answered? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
The Fit Model Window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Graphical Display: Leverage Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Contents
Quantify Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Analysis of Variance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Mean Estimates and Statistical Comparisons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
iv
Quantify Results: Statistical Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Analysis of Variance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Summary Reports For The Whole Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Summary Reports for Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
102
103
105
105
107
Exploring Data
Finding Exceptions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Solubility Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
One-Dimensional Views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Two-Dimensional Views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Three-Dimensional Views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Principal Components and Biplots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Multivariate Distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
109
111
111
112
113
114
116
117
10 Multiple Regression
Examining Multiple Explanations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Aerobic Fitness Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fitting Plane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Fit Planes to Test Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Whole Model Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
More and More Regressors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Interpreting Leverage Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Collinearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
119
121
123
124
126
126
127
129
130
Index
JMP Introductory Guide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
Credits and Acknowledgments

Origin
JMP was developed by SAS Institute Inc., Cary, NC. JMP is not a part of the SAS System, though portions of JMP were adapted from routines in the SAS System, particularly for linear algebra and probability calculations. Version 1 of JMP went into production in October, 1989.
Credits
JMP was conceived and started by John Sall. Design and development were done by John Sall,
Chung-Wei Ng, Michael Hecht, Richard Potter, Brian Corcoran, Annie Dudley Zangi, Bradley Jones,
Craige Hales, Chris Gotwalt, Paul Nelson, Xan Gregg, Jianfeng Ding, Eric Hill, John Schroedl, Laura
Lancaster, Scott McQuiggan, and Peng Liu.
In the SAS Institute Technical Support division, Wendy Murphrey and Toby Trott provide technical
support and conduct test site administration. Statistical technical support is provided by Craig
DeVault, Duane Hayes, Elizabeth Edwards, Kathleen Kiernan, and Tonya Mauldin.
Nicole Jones, Jim Borek, Kyoko Keener, Hui Di, Joseph Morgan, Wenjun Bao, Fang Chen, Susan
Shao, Hugh Crews, Yusuke Ono and Kelci Miclaus provide ongoing quality assurance. Additional testing and technical support is done by Noriki Inoue, Kyoko Takenaka, and Masakazu Okada from SAS
Japan.
Bob Hickey is the release engineer.
The JMP manuals were written by Ann Lehman, Lee Creighton, John Sall, Bradley Jones, Erin Vang,
Melanie Drake, and Meredith Blackwelder, with contributions from Annie Dudley Zangi and Brian
Corcoran. Creative services and production was done by SAS Publications. Melanie Drake implemented the help system.
Jon Weisz and Jeff Perkinson provided project management. Also thanks to Lou Valente, Ian Cox,
Mark Bailey, and Malcolm Moore for technical advice.
Thanks also to Georges Guirguis, Warren Sarle, Gordon Johnston, Duane Hayes, Russell Wolfinger,
Randall Tobias, Robert N. Rodriguez, Ying So, Warren Kuhfeld, George MacKensie, Bob Lucas, Warren Kuhfeld, Mike Leonard, and Padraic Neville for statistical R&D support. Thanks are also due to
Doug Melzer, Bryan Wolfe, Vincent DelGobbo, Biff Beers, Russell Gonsalves, Mitchel Soltys, Dave
Mackie, and Stephanie Smith, who helped us get started with SAS Foundation Services from JMP.
Acknowledgments
We owe special gratitude to the people that encouraged us to start JMP, to the alpha and beta testers of
JMP, and to the reviewers of the documentation. In particular we thank Michael Benson, Howard Yetter (d), Andy Mauromoustakos, Al Best, Stan Young, Robert Muenchen, Lenore Herzenberg, Ramon
Leon, Tom Lange, Homer Hegedus, Skip Weed, Michael Emptage, Pat Spagan, Paul Wenz, Mike
Bowen, Lori Gates, Georgia Morgan, David Tanaka, Zoe Jewell, Sky Alibhai, David Coleman, Linda
vi
Blazek, Michael Friendly, Joe Hockman, Frank Shen, J.H. Goodman, David Ikl, Barry Hembree, Dan
Obermiller, Jeff Sweeney, Lynn Vanatta, and Kris Ghosh.
Also, we thank Dick DeVeaux, Gray McQuarrie, Robert Stine, George Fraction, Avigdor Cahaner, Jos
Ramirez, Gudmunder Axelsson, Al Fulmer, Cary Tuckfield, Ron Thisted, Nancy McDermott, Veronica
Czitrom, Tom Johnson, Cy Wegman, Paul Dwyer, DaRon Huffaker, Kevin Norwood, Mike Thompson, Jack Reese, Francois Mainville, and John Wass.
We also thank the following individuals for expert advice in their statistical specialties: R. Hocking and
P. Spector for advice on effective hypotheses; Robert Mee for screening design generators; Greg Piepel,
Peter Goos, J. Stuart Hunter, Dennis Lin, Doug Montgomery, and Chris Nachtsheim for advice on
design of experiments; Jason Hsu for advice on multiple comparisons methods (not all of which we
were able to incorporate in JMP); Ralph OBrien for advice on homogeneity of variance tests; Ralph
OBrien and S. Paul Wright for advice on statistical power; Keith Muller for advice in multivariate
methods, Harry Martz, Wayne Nelson, Ramon Leon, Dave Trindade, Paul Tobias, and William Q.
Meeker for advice on reliability plots; Lijian Yang and J.S. Marron for bivariate smoothing design;
George Milliken and Yurii Bulavski for development of mixed models; Will Potts and Cathy
Maahs-Fladung for data mining; Clay Thompson for advice on contour plotting algorithms; and Tom
Little, Damon Stoddard, Blanton Godfrey, Tim Clapp, and Joe Ficalora for advice in the area of Six
Sigma; and Josef Schmee and Alan Bowman for advice on simulation and tolerance design.
For sample data, thanks to Patrice Strahle for Pareto examples, the Texas air control board for the pollution data, and David Coleman for the pollen (eureka) data.
Translations
Erin Vang coordinated localization. Noriki Inoue, Kyoko Takenaka, and Masakazu Okada of SAS
Japan were indispensable throughout the project. Special thanks to Professor Toshiro Haga (retired, Science University of Tokyo) and Professor Hirohiko Asano (Tokyo Metropolitan University for reviewing
our Japanese translation. Special thanks to Dr. Fengshan Bai, Dr. Xuan Lu, and Dr. Jianguo Li, professors at Tsinghua University in Beijing, and their assistants Rui Guo, Shan Jiang, Zhicheng Wan, and
Qiang Zhao, for reviewing the Simplified Chinese translation. Finally, thanks to all the members of our
outstanding translation teams.
Past Support
Many people were important in the evolution of JMP. Special thanks to David DeLong, Mary Cole,
Kristin Nauta, Aaron Walker, Ike Walker, Eric Gjertsen, Dave Tilley, Ruth Lee, Annette Sanders, Tim
Christensen, Jeff Polzin, Eric Wasserman, Charles Soper, Wenjie Bao, and Junji Kishimoto. Thanks to
SAS Institute quality assurance by Jeanne Martin, Fouad Younan, and Frank Lassiter. Additional testing
for Versions 3 and 4 was done by Li Yang, Brenda Sun, Katrina Hauser, and Andrea Ritter.
Also thanks to Jenny Kendall, John Hansen, Eddie Routten, David Schlotzhauer, and James Mulherin.
Thanks to Steve Shack, Greg Weier, and Maura Stokes for testing JMP Version 1.
Thanks for support from Charles Shipp, Harold Gugel (d), Jim Winters, Matthew Lay, Tim Rey,
Rubin Gabriel, Brian Ruff, William Lisowski, David Morganstein, Tom Esposito, Susan West, Chris
Fehily, Dan Chilko, Jim Shook, Ken Bodner, Rick Blahunka, Dana C. Aultman, and William Fehlner.
Technology License Notices
The ImageMan DLL is used with permission of Data Techniques, Inc.
vii
RECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF

USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS
ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
XRender is Copyright 2002 Keith Packard. KEITH PACKARD DISCLAIMS ALL WARRANTIES WITH
REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL KEITH PACKARD BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS,
WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT
OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
Credits
Scintilla is Copyright 1998-2003 by Neil Hodgson <neilh@scintilla.org>. NEIL HODGSON DISCLAIMS

ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL NEIL HODGSON BE LIABLE FOR ANY SPECIAL, INDI-
Chapter 1
Introducing JMP
Your First Look
JMP uses an extraordinary graphical interface to display and analyze data. JMP is software for interactive statistical graphics and includes:
a data table window for editing, entering, and manipulating data
a broad range of graphical and statistical methods for data analysis
an extensive design of experiments module
options to highlight and display subsets of data
a formula editor for each table column to compute values as needed
a facility for grouping data and computing summary statistics
special plots, charts, and communication capability for quality improvement techniques
tools for printing and for moving analyses results between applications
a scripting language for saving and creating frequently used routines
This introductory chapter gives basic information about using JMP.
Contents
What You Need to Know . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Learning About JMP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Using Tutorials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Searching in the Help . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Learning About Statistical and JSL Terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
Using the Context-Sensitive Help . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .5
Learning JMP Tips & Tricks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Using This Book in Combination with Other Included Books . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Conventions Used in this Book . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Step 1: Start JMP . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Step 2: Open a JMP Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Step 3: Learn About the Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
Specifying the Values Type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10
Data Table Cursor Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .10
Selecting Rows and Columns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
Step 4: Select an Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Casting Columns Into Roles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Step 5: View the Output Report. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
Graphs and Charts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .14
Statistical Tables and Text . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
Step 6: Save the JMP Output Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .16
A Practice Tutorial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .16
Open a Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .16
Select an Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .17
Spin the Cowboy Hat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
Chapter 1
Introducing JMP
What You Need to Know
Before you begin using JMP, you should be familiar with:

Standard operations and terminology such as click, double-click, Ctrl-click and Alt-click (Command-click and Option-click on the Macintosh), Shift-click, drag, select, copy, and paste.
How to use menu bars and scroll bars, how to move and resize windows, and how to manipulate
files in the desktop. If you are using your computer for the first time, consult the reference guides
that came with it for more information.
Minimal statistics. Even though JMP has many advanced features, you only need a minimal background of formal statistical training. All analyses include graphical displays with options that help
you review and interpret the results. Each analysis also includes access to help windows that offer
general help and some statistical details.
Learning About JMP

If you are familiar with JMP, you might want to know only whats new. The JMP New Features document gives a summary of general changes and additions. To learn more about JMP, use the recommendations in the following sections.
Using Tutorials
JMP provides three types of tutorials:
Beginners Tutorial The beginners tutorial steps you through the JMP interface and explains the
basics of how to use JMP. It is accessible through JMPs Tip of the Day window, which appears when
you start JMP. To start the tutorial from the Tip of the Day window, click Enter Beginners
Tutorial. Or, start the tutorial by selecting Help (View on the Macintosh) > Tutorials > Beginners
Tutorial.
Specific Analysis Tutorials Tutorials that step you through creating an analysis in JMP are found
under Help (View on the Macintosh) > Tutorials. Tutorials describe how to create a chart, compare
means, how to design an experiment, and more.
JMP Introductory Guide The JMP Introductory Guide is a collection of tutorials designed to help
you learn JMP strategies. If you did not receive a printed copy of this book, view the .pdf file by
selecting Help > Books > JMP Introductory Guide. By following along with these step-by-step
examples, you can quickly become familiar with JMP menus, options, and report windows.
Searching in the Help

You might want help on a specific topic, and you want to search JMPs online help for that topic. The
main menu bar contains a Help menu, which provides the appropriate searching capabilities:
1 Introducing JMP
What You Need to Know
Introducing JMP
Learning About JMP
Chapter 1
Windows and Linux

Help Menu
Macintosh Help Menu
On Windows and Linux, the Help > Contents, Help > Search, and Help > Index commands access
the JMP help system. The help system provides navigable online JMP documentation.
On the Macintosh, the Help > JMP Help command displays a list of JMP help items with search capabilities and a table of contents.
Learning About Statistical and JSL Terms

The Help > Indexes command displays the following sources for your reference:
Statistics Index Accesses references that give definitions of statistical terms. Once you are in the
Statistics Index window, click the Topic Help button to go to the place in the online help that
describes the highlighted topic. Click the Script button to view or hide the script associated with
the highlighted topic. Click Launch to run the script that corresponds to the item you have highlighted in the list. Click Example to view an example of the highlighted topic.
Chapter 1
Introducing JMP
Learning About JMP
description
JSL value/script
JSL Operators Index Presents a list of JSL operators, such as Sin, Cos, Sqrt, and Abbrev Date that
you would use when writing JSL. Highlight an operator name to see a description of the operator
appear in the window on the right. Click the Topic Help button to locate the topic in the online
help.
Object Scripting Index Presents a list of JSL objects. These are scriptable JSL building blocks.
Highlight an object name and messages the object recognizes appear in the window on the right.
DisplayBox Scripting Index Presents a list of the elements that make up a JMP report. These elements are the JSL building blocks with which you build output. Highlight a Display Box and available messages for each object appear in the window on the right.
Using the Context-Sensitive Help

To use JMPs online help system, select one of the following methods:
Select Help from analysis construction windows (as shown in Figure 1.1) and report windows.
1 Introducing JMP
list of topics
Introducing JMP
Learning About JMP
Chapter 1
Figure 1.1 Help Is Available
Select the help tool ( ) from the Tools menu and click a place in a data table or report on which
you need assistance (Figure 1.2). Context-sensitive help tells about the items in the area you clicked.
Figure 1.2 Use the Help Tool for Context-Sensitive Help
In some reports, make a small circle with your cursor to reveal information about the item in the
area.
Figure 1.3 Making a Circle with the Cursor Displays Help
In some menus, hold the cursor on menu items to reveal information about the menu item.
Figure 1.4 Display a Description of Menu Items
Chapter 1
Introducing JMP
Conventions Used in this Book
When you first start JMP, you see the Tip of the Day window. This window provides tips about using
JMP that you might not know.
To turn off the Tip of the Day, un-check Show tips at startup. To view it again, select Help (View on
the Macintosh) > Tip of the Day.
Also use the JMP Quick Reference Guide to learn more advanced commands in JMP. View this document by selecting Help > Books > JMP Quick Reference Card.
Using This Book in Combination with Other Included Books

The book you are reading now is the JMP Introductory Guide. See the following manuals for further
documentation of JMP:
The JMP User Guide has complete documentation of all JMP menus, an explanation of data table
manipulation, and a description of the formula editor. There are chapters that show how to do common tasks such as manipulating files, transforming data table columns, and cutting and pasting
JMP data, statistical text reports, and graphical displays.
The JMP Statistics and Graphics Guide gives documentation of the Analyze and Graph menus. It
documents analyses, discusses statistical methods, and describes all report windows and options.
The JMP Design of Experiments covers the DOE menu, the experimental design facility in JMP.
The JMP Scripting Guide is a reference guide to the JMP scripting language (JSL) that lets you automate action sequences.
If you did not receive printed copies of these books, view the .pdf files by selecting Help > Books.
Conventions Used in this Book

Conventions used in this manual were devised to help relate written material to information that
appears on-screen:
The .jmp extension follows file names on the PC. When you installed JMP, a folder named Sample
Data was also installed. On the Macintosh, JMP sample data files have the same name, but show
without an extension. Reference to names of JMP files, data tables, variable names, and items in
reports appear in Helvetica to help distinguish them from surrounding text.
Special information, warnings, and limitations are noted in sentences beginning with the bold word
Note.
Reference to menu names (File menu) or menu items (Save command) appear in Helvetica bold
font.
The notation to select a command from a menu is sometimes written as File > New, meaning select
the New command from the File menu.
Words or phrases that are important or have definitions specific to JMP are in italics the first time
they appear.
1 Introducing JMP
Learning JMP Tips & Tricks
Introducing JMP
Step 1: Start JMP
Chapter 1
Step 1: Start JMP

Start a JMP session by double-clicking the JMP application icon. Your initial view of JMP will be a
menu bar, a tool bar, the Tip of the Day window, and the JMP Starter window (Figure 1.5).
Figure 1.5 First View of JMP (Windows)
To start the online tutorial, click Enter Beginners Tutorial. Or, click the Close button to close the
window and follow the tutorials in this book.
Step 2: Open a JMP Data Table

There are several ways to open a data table:
Go to your Sample Data directory in its default location. For example:
In Windows: C:\Program Files\SAS\JMP\8\Support Files English\Sample Data
In Linux: Home\JMP8\Support Files English\Sample Data
In Macintosh: Hard Drive\Library\Application Support\JMP\Support Files English\Sample
Data
Selecting File > New (or clicking the New Data Table button on the JMP Starter window) creates
and displays a data table with an empty data grid. First, add rows and columns, then type in or paste
in new data. For details, see the JMP User Guide.
Selecting File > Open (or clicking the Open Data Table button on the JMP Starter window) presents a file selection window (Figure 1.6) with a list of existing tables. Select a file and click Open.
For details, see the JMP User Guide.
Chapter 1
Introducing JMP
Step 3: Learn About the Data Table

Opening or creating a data table creates a data grid and table information panels, like the ones shown in
Figure 1.7. The counts of table rows and columns appear in the corresponding panels to the left of the
data grid. In the data grid, a row number identifies each row, and each column has a column name.
Rows and columns are sometimes called observations and variables in an analysis.
Figure 1.7 A Data Table
The JMP data table window is a flexible way to prepare data. Using it, you can accomplish a variety of
table management tasks, such as:
Editing the value in any cell
1 Introducing JMP
Figure 1.6 The Open Data File Window
10
Introducing JMP
Chapter 1
Changing a columns width by dragging the column line

Hiding columns temporarily, or deleting columns permanently
Adding rows, or rearranging the order of rows
Adding columns, or rearranging the order of columns
Selecting a subset of rows for analysis and saving that subset for further use
Sorting or combining tables
Specifying the Values Type

The small icon to the left of the column name in the columns panel is an icon that can be clicked. Use
it to declare the modeling type of the values in the column. JMP uses three modeling types to determine how to analyze the columns values:
Continuous (
Ordinal (
Nominal (
) Values are numeric measurements.
Values are ordered categories, which can have either numeric or character values.
)
)
Values are numeric or character classifications.
Modeling types are changeable depending on how you want to look at your data.
For example, a variable like age should be specified continuous to find the mean
(average) age, but nominal or ordinal to find frequency counts for each age value.
The default modeling type is nominal for character values and continuous for
numeric values. To assign a different modeling type to your variables:
1 Click the icon next to the variable name.
2 Select the appropriate modeling type.
Data Table Cursor Forms

As you move the cursor around the data table, it changes forms. Its shape gives you information about
performable actions. The following sections describe the different cursor forms.
Arrow Cursor
The cursor displays as a standard arrow when it is anywhere in the table panels to the left of the data
grid, except when it is on a red triangle icon ( ) or diamond-shaped disclosure button (
on
Windows and
on the Macintosh) or when it is in the upper-left corner of the data gridthe area
where rows and columns are deselected (as shown here).
Chapter 1
Introducing JMP
11
Click to select
column or double-click to edit
column name
Click to deselect rows
I-Beam Cursor
The cursor is an I-beam when it is over text in the data grid or highlighted column names in the data
grid or column panels. To edit text in the data grid:
1
2
3
4
Click the cell you wish to edit. The cell highlights.

Click again next to any character to mark an insertion point.
The I-beam deposits a vertical blinking bar.
Use the keyboard to make changes.
To edit a column name:

1 Click the column name to highlight the column.
2 Press the Enter key to change the I-beam cursor to an insertion point.
3 Use the keyboard to make changes.
Large Cross Cursor
The cursor becomes a large cross when moved into a column or row selection area. When moved over a
column name, you can edit the name. To do so, click the column name and begin typing.
The cross cursor can also be used to select rows and columns. To select a column, click the area above
the column name. See the next section, Selecting Rows and Columns, p. 12, for a detailed explanation of selecting rows and columns.
Double Arrow Cursor
The cursor changes to a double-arrow cursor when positioned on a column boundary or on a panel
splitter. Dragging the double-arrow cursor changes the column width or the panel size.
Drag to change width of column
1 Introducing JMP
Click to deselect column
12
Introducing JMP
Chapter 1
Hand Cursor
The cursor changes to a hand when you move the mouse over a red triangle icon ( ) or diamond-shaped disclosure button (
on Windows and
on the Macintosh).
Click the red triangle to reveal the menu and select a menu icon. Click the disclosure button to open or
close a panel.
Selecting Rows and Columns

Select rows and columns in a JMP data grid by highlighting them, as explained in Table 1.1 and shown
in Figure 1.8. For additional details, see the JMP User Guide.
Table 1.1 Ways to Select Rows and Columns
Action
Instructions
Highlight a row
Click the space that contains the row number.
Highlight a column
Click the background area above the column name. Or, click the column name in the columns panels to the left of the data grid.
Extend a selection of
rows or columns
Shift-click the first and last rows or columns of the desired range.
Make a discontiguous
selection
Ctrl-click (Command-click on the Macintosh) the desired selections.
Figure 1.8 Select Rows and Columns

selected column
selected rows
Chapter 1
Introducing JMP
Step 4: Select an Analysis
13
There are a variety of analyses available through the Analyze and Graph menus in the main menu. An
alternate way to access these analyses is through toolbar buttons and selections in the JMP Starter window. Selecting an analysis in the Analyze or Graph menus produces graphs, charts, plots, and/or tables.
For example, to see a histogram of columns in the data table you have open, select
Analyze > Distribution. Then, complete the window and click OK.
Casting Columns Into Roles

After you select an analysis from the main menu, a window appears that asks you to cast columns into
roles. For example, if you select Analyze > Fit Y by X from the main menu, you will see the window in
Figure 1.9.
Figure 1.9 Fit Y by X Window
The JMP analysis methods are like stages or platforms for variables to dramatize their values. Each analysis requires information about which variables play what roles in an analysis.
The most typical variable roles are:
Y, Response
studied.
Identifies a column as a response or dependent variable whose distribution is to be
X, Factor Identifies a column as an independent, classification, or explanatory variable whose values divide the rows into sample groups.
Weight Identifies a numeric column whose values supply weights for each response.
Freq
By
Identifies a numeric column whose values assign a frequency to each row for the analysis.
Identifies a column that is used to create a report consisting of separate analyses.
Step 5: View the Output Report

After you have cast columns into their roles, JMP provides output reports that include graphics and
text. For more detail than is presented below, see the JMP User Guide.
1 Introducing JMP
Step 4: Select an Analysis
14
Introducing JMP
Chapter 1
Graphs and Charts

JMP reports are usually filled with graphs, charts, plots, and other graphical displays that show your
results. For example, if you select Analyze > Distribution and assign several columns the Y, Response
role in the Distribution window that appears, you create a report containing a graphical display of each
column assigned the Y, Response role.
For the example shown in Figure 1.10, the Distribution command produces graphical displays which
include:
Histograms of both the brand and speed columns.
An outlier box plot of the continuous variable speed.
Figure 1.10 Distribution Histograms and Outlier Box Plot
Histogram
Outlier Box
Plot
Display Options
To enhance the default graphical displays that show your results, JMP provides options that you can
add to them. These options are found by clicking the red triangle icon beside a report name. For example, the red triangle icon next to the histogram name lists available report options (Figure 1.11). For
practice, try selecting different combinations of these options and watch the effect they have on the displays and reports.
Select a column in Select Columns.
Select Y, Columns in Cast Selected Columns into Roles.
Select variables for Weight, Freq, and By.
Click OK.
Chapter 1
Introducing JMP
15
Statistical Tables and Text

In addition to the graphs, charts, plots, and other graphical displays in a report, JMP can also include
text tables in a report. The kinds of tables given depend on whether a variable is continuous or categorical (ordinal or nominal).
When you installed JMP, a folder named Sample Data was also installed. In that folder is a file named
Typing Data.jmp. Using the Distribution command to produce a distribution report creates text tables
for the speed and brand variables (Figure 1.12). In this example:
For a continuous variable, JMP displays a Quantiles table and a Moments table.
For nominal and ordinal variables, JMP displays a Frequency table showing the total sample frequency, category frequencies, and associated probabilities.
Figure 1.12 Statistical Reports
1 Introducing JMP
Figure 1.11 Options for Nominal or Ordinal Variable in a Distribution Analysis
16
Introducing JMP
Step 6: Save the JMP Output Report
Chapter 1
JMP also gives you the ability to change the appearance of these tables. For details, see the JMP User
Guide.
Step 6: Save the JMP Output Report

To save a report just as it appears in the report window, select File > Save.
Another way to save is to duplicate it in a separate window titled Journal. Then you can append other
reports to it or manipulate it in the journal window. To do this, select Edit > Journal. Then, select File
> Save As. For more detail than is presented below, see the JMP User Guide.
Report window
Journal window
A Practice Tutorial
Before you begin the tutorials in the following chapters of this book, complete this brief practice tutorial that is a short guided tour through a JMP analysis. Follow the steps to see a three-dimensional scatterplot.
Open a Data Table

Open the file called Cowboy Hat.jmp to begin a JMP session. When you installed JMP, a folder
named Sample Data was also installed. In that folder is a file named Cowboy Hat.jmp.
Chapter 1
Introducing JMP
A Practice Tutorial
17
Figure 1.13 Cowboy Hat Data Table
This data table has three numeric columns and two row state columns. Columns x and y are x- and
y-coordinates, and z is created using the function
2
z = sin x + y
Select an Analysis
To plot the three columns of information from the Cowboy Hat data table:
Choose the Scatterplot 3D command from the Graph menu.
Select the x, y, and z columns from the column selector list on the left side of the window, and click
Y, Columns, as shown in Figure 1.14.
Figure 1.14 Scatterplot 3D Column Selection Window
These column names now appear in the list on the right side of the window.
Click OK.
The scatterplot 3-D appears. Initially, the data points look like a two-dimensional plot because the z
dimension is projected onto the x-y plane.
1 Introducing JMP
The data table shown in Figure 1.13 appears.
18
Introducing JMP
A Practice Tutorial
Spin the Cowboy Hat

To get a better view of the cowboy hat,
Right-click on the plot and select Settings.
Deselect Walls, Grids, Axes, and Box.
Move the mouse over the plot, press the mouse button, and move the cursor about.
The cowboy hat moves in three dimensions.
To have the plot spin itself:
Chapter 1
Chapter 1
To stop the scatterplot 3-D, click again in the scatterplot 3-D frame.
19
1 Introducing JMP
Press the Shift key, and give the plot a push with the cursor.
Introducing JMP
A Practice Tutorial
Chapter 2
Entering and Plotting Data
This lesson evaluates a new drug developed to lower blood pressure. Data were recorded over a
six-month period for the following treatment groups:
300 mg dose
450 mg dose
placebo
control
Following are the mean monthly blood pressure for each group, recorded in a journal. This lesson
shows how to enter data values into the data table and to create a single neat and informative line chart
that shows the study results.
Objectives
Create rows and columns in a data table, one at a time and in groups.
Enter data into JMP.
Create a chart using the Chart command.
Rescale axes in a plot.
Animate a plot.
Blood Pressure Study

Month
Control
Placebo
300mg
450mg
March
165
166
168
April
May
June
162
164
162
163
159
158
161
165
161
158
163
153
151
July
August
166
163
158
158
160
157
148
150
Contents
Starting a JMP Session . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23
Creating Rows and Columns in a JMP Data Table. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .23
Add Columns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
Set Column Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .25
Add Rows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .25
Entering Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .25
Plotting Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
Document the Report . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
Chapter 2

Starting a JMP Session
23
Open the JMP application icon to begin a JMP session.

Use the New command in the File menu to create an empty data table like the one shown here.
Click File > New > Data Table or open JMP Starter and select New Data Table.
The data values for this project are blood pressure statistics collected over six months and recorded in a
notebook page as shown in Figure 2.1.
Figure 2.1 Notebook of Raw Study Data
Blood Pressure Study

Month
Control
Placebo
300mg
450mg
March
165
166
168
April
May
June
162
164
162
163
159
158
161
165
161
158
163
153
151
July
August
166
163
158
158
160
157
148
150
Creating Rows and Columns in a JMP Data Table

JMP data tables have rows and columns, sometimes called observations and variables in statistical terms.
The raw data in Figure 2.1 are arranged as five columns (treatment groups) and six rows (months
March through August). The first line in the notebook names each column of values. These names can
be used as column names in the new JMP table.
2 Data Tables
Starting a JMP Session
24

Creating Rows and Columns in a JMP Data Table
Chapter 2
Add Columns
First create the number of rows and columns that are needed.
Select Cols > Add Multiple Columns, which prompts for the number of columns to add, where to
add them, and which type of columns to add.
Specify five new columns.

The default column names are Column 1, Column 2, and so on, but these can be changed at any time
in the columns panel or at the top of the column in the data table. So, the next step is to type in meaningful names. Use the names from the data journalMonth, Control, Placebo, 300 mg, and 450 mg.
Edit a column name by clicking it in the data grid or in the columns panel.
Begin typing once the name is highlighted.
Click once to select
Then, begin typing
Chapter 2

Entering Data
25
Columns can have different characteristics. By default, they contain numeric data. However, the values
for month in this example are character values. To change the Month column from numeric to
character:
Highlight the column by clicking either in the area at the top of the column or the area beside its
name in the columns panel.
Select the Cols > Column Info to display the window in Figure 2.2.
Change Month to a character variable, as shown in Figure 2.2, by clicking the box beside Data Type.
The Column Info window is also used to change other column characteristics and to access the JMP
formula editor for computing column values.
Figure 2.2 Change Data Type
Add Rows
Now add new rows to the table:
Choose Rows > Add Rows.
Specify six new rows.
Alternatively, double-click anywhere in the body of the table to automatically fill it with new rows up
through the position of the cursor.
Select File > Save to name the table BP Study.jmp and save it.
The data table is now ready to hold data values. To summarize the table evolution so far, you:
Began with a new untitled table.
Added enough rows and columns to accommodate the raw data.
Tailored the characteristics of the table by giving the table and columns descriptive names.
Changed the data type of the Month column to accept character values.
Entering Data
To enter data into the data table, type values into their appropriate table cells.
Type the values from the study journal (Figure 2.1) into the BP Study.jmp table as shown here.
2 Data Tables
Set Column Characteristics
26

Entering Data
Chapter 2
When entering data into the data table:

Edit the cell value by moving the cursor into a data cell and double-clicking. The cursor becomes a
flashing vertical bar.
Correct a mistake by dragging the text entry bar across the incorrect entry and typing the correction
over it.
Press the Return or Enter key on the numeric keypad or the Tab key to move the highlight one cell to
the right. Press Shift-Tab or Shift-Return/Enter on the numeric keypad to move the highlight one
cell to the left.
Click to highlight
Begin typing
The completed data table
Plotting Data
When working with the Analyze and Graph menu commands, you tell JMP which columns to work
with and what to do with them. This section shows how to plot the months across the horizontal (x)
axis and the columns of blood pressure statistics for each treatment group overlaid on the vertical (y)
axis.
Select Graph > Chart.
The window in Figure 2.3 appears.
Assign x and y roles and choose the type of chart. This example specification is for a bar chart, with data
(as opposed to statistics) as chart points.
Assure that the default choice Vertical is selected from the chart type drop-down list.
Select one continuous variable in the list.
Select the Shift and down arrow keys to select the other continuous variables.
Chapter 2

Entering Data
27
2 Data Tables
Select the four continuous variables in the Select Columns list.

Click the Statistics button and select Data from the drop-down list.
Enter Month as the Categories (X).
Select Month from the Select Columns list.
Click Categories.
Click OK.
Figure 2.3 Creating the Bar Chart
Choose chart
type
Choose y-axis
information
JMP displays an overlaid bar chart of the data.
Reorder the Values

By default, the month values in this chart do not appear in a logical order. To further enhance the
report, reorder the months so they are in chronological order rather than alphabetical order.
In the data table, highlight the Month column and select Cols > Column Info.
Select Value Ordering from the Column Properties drop-down menu. Value ordering information
appears on the right, as shown in Figure 2.4. The list in the window contains values in the order
JMP will display them in reports. Use the Move Up and Move Down buttons to change the order
of the months.
28

Entering Data
Chapter 2
Figure 2.4 The Value Ordering Window
Click OK. The properties icon ( ) now appears next to the column name in the data tables column
panel, indicating the column contains a property.
In the analysis report, click the red triangle and select Script > Redo Analysis.
Rescale the Plot Axis
By default, y-axis scaling begins at zero and the overlay chart looks like the one shown here. But, to
present easy-to-read information, the y-axis needs to be rescaled and the chart needs labels.
Double-click the y-axis area, which accesses the Axis Specification window (Figure 2.5).
This window gives you the ability to:
Set the minimum and maximum of the axis scale.
Specify the tick mark increment.
Request minor tick marks.
Request grid lines at major or minor tick marks.
Format numeric axes.
Use either a linear or log-based scale.
Chapter 2

Entering Data
29
In this example, the plotted values range from about 145 to 175.
Enter these figures into the Axis Specification window for Minimum and Maximum.
Change the increment for the tick marks from 50 to 1 by entering a 1 in the Increment box.
Click OK.
Tip: The magnifier tool (
), found in the Tools menu and the cursor tool bar, can also be used to
change the scale of graphs. Drag the magnifier diagonally across the points of interest to see the chart
automatically adjust. Double-click the plot frame to reset the plot to its original scale.
Click the edge of the graph and drag it to the right to increase its width.
Change the name of the axis from Y to Blood Pressure:
Place cursor over Analysis Report until cursor becomes an I-bar.
Click for a text box and enter Blood Pressure.
Document the Report

The chart also needs a title and other documentation to make it easy to interpret. The annotate tool
( ) places text on the report. Refer to Figure 2.6 to see where to place the following steps.
Select the annotate tool.
Click and drag in the report to create a text box.
Release the button and enter the text for the title Comparison of Treatment Groups.
Click outside the annotation to quit editing the text.
Repeat to enter the footnote XYZ Blood Pressure Study 2007.
Note: Double-click any report title bar to edit the text on the bar.
2 Data Tables
Figure 2.5 Axis Specification Window
30

Chapter Summary
Chapter 2
Figure 2.6 Bar Chart with Modified Y-Axis, Titles, and Footnotes
Chapter Summary
A study was done to evaluate the effect of a new drug on blood pressure. To complete this analysis, you:
Used the New Data Table command in the File menu to create a new JMP table.
Created the appropriate number of rows and columns for the data.
Typed the data into the empty data grid.
Used the Chart command in the Graph menu to request a bar chart of blood pressure measures over
time.
Ordered the values in chronological order so they would appear properly in the chart.
Tailored the chart with a specific axis scale and axis name, and added a plot title and footnote with
the annotate tool.
Chapter 3
Summarizing Data
Look Closely at the Data
The hot dog is a questionable item on a school cafeteria menu because of its reputation as an unhealthy
food, possibly classified in the junk food category. Many students feel this is unpatriotic and are upset.
This lesson examines the hot dog as a menu item, but not before looking into the multitude of brands
available. The data shows information about cost, nutritional ingredients of concern, and taste preference for 54 hot dog brands. This information is sufficient to provide a summary of hot dog statistics
and to identify the brands that are:
most nutritious
least costly
best tasting
The taste, cost, and nutritional variables used in this chapter are an enhancement of data from Moore,
D. S., and McCabe G. P., (1989), Introduction to the Practice of Statistics, and Consumer Reports (1986).
The brand names were changed to fictional names, and the taste preference labels correspond to a taste
preference scale.
Objectives
Find and mark subgroups of data
Produce scatterplots using the Fit Y by X command and use them as discovery tools
Label individual points in plots
Produce and plot summary statistics
Contents
Look Before You Leap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
Grouping Data. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .34
Creating Statistics for Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
Charting Statistics from Grouped Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .36
Charting Statistics for Two Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .38
Finding a Subgroup with Multiple Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .39
Comparative Scatterplots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .39
What Has Been Discovered? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41
Finding the Best Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .41
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
Chapter 3
Summarizing Data
Look Before You Leap
33
When you installed JMP, a folder named Sample Data was also installed. In that folder is a file named
Hot Dogs.jmp. Open the Hot Dogs.jmp file to see the data shown in Figure 3.1.
Figure 3.1 Hot Dogs Data Table
The Hot Dogs.jmp table has the following information:

The columns called Type, Calories, Sodium, and Protein/Fat (an index ratio of protein to fat) give
information about nutrition. The Type column has values Meat, Poultry, and Beef.
Cost information is in columns $/oz (dollars per ounce of hot dog) and $/lb Protein (dollars per
pound of hot dog protein).
Three categories of taste are coded Bland, Medium, and Scrumptious in the Taste column.
To get a feel for these data, use the Distribution command.
Select Analyze > Distribution to see the window in Figure 3.2.
Select all the variables except Product Name and click the Y, Columns button.
To select more than one item, highlight the first item, hold down the Shift key and press the down
arrow button until all desired items are selected.
Click OK.
Figure 3.2 Distribution Command
Examine the resulting report to see the distributions and levels of each variable.
3 Summarizing Data
34
Summarizing Data
Grouping Data
Chapter 3
Grouping Data
Of course, health is a primary concern of a school cafeteria. It is interesting to see if type of hot dog
plays a role in healthfulness. In particular:
Which type of hot dog has the fewest calories?
Is the amount of sodium different in the three types of hot dogs?
Which hot dogs have the highest protein content?
Which hot dogs taste good and are healthy?
To address these issues, the data need to be grouped into hot dog type and taste
preference categories with summary statistics computed for each group. The
Summary command in the Tables menu groups data and computes summary
statistics.
The Summary command creates a summary table. This table summarizes columns from the active data table, called its source table. The Hot Dogs.jmp table
is the source table in this example. A summary table has a single row for each
level (value) of a specified variable.
Select Tables > Summary.
Select Type and click the Group button to see the window as shown in Figure 3.3.
Click OK.
Figure 3.3 Summary Window
Chapter 3
Summarizing Data
Grouping Data
35
A summary table is not independent of its source table. It has these characteristics:
When rows are highlighted in the summary table, their corresponding rows highlight in the source
table.
The summary table is not saved when closed. Select File > Save As to specify a name and location
for the table.
Figure 3.4 Summary Table for Type of Hot Dog
Creating Statistics for Groups

Next, expand the summary table with columns of statistics. Summary tables have an additional command in the columns panel called Add Statistics Column. Use this command to add statistical summary columns to the table at any time.
To follow along with this example, do these steps:

Click the red triangle icon and select the Add Statistics Column command from the columns panel
on the left side of the screen.
Select Calories, Sodium, and Protein/Fat in the column selector list of the window, as shown in
Figure 3.5.
Click the Statistics button and select Mean. You should now see new column names in the Statistics list.
Click OK.
3 Summarizing Data
The Hot Dogs By (Type) summary table (Figure 3.4) appears in a new window. The Type column lists
hot dog type and the NRows column gives the frequency of each type in the source table.
36
Summarizing Data
Grouping Data
Chapter 3
Figure 3.5 Summary Window
The new columns of statistics are displayed in the Hot Dogs By (Type) table (top table of Figure 3.6).
Repeat the previous steps to create a second summary table of Hot Dogs by Taste to look at health
factors and hot dog tastiness. The Hot Dogs By (Taste) summary table shows average calories,
sodium, and protein-to-fat ratio for each taste category (bottom table of Figure 3.6).
Figure 3.6 Summary Statistics for Hot Dog Groups
Charting Statistics from Grouped Data

The summary tables in Figure 3.6 show the summary statistics in tabular form, but bar charts are better
for visual comparison. The Chart command on the Graph menu can also summarize data and then create charts of the summarized data.
Make sure the Hot Dogs.jmp table is active.
Assign variable roles as shown in Figure 3.7.
Charts like those below should appear.
Chapter 3
Summarizing Data
Grouping Data
37
Click to
choose a
statistic
Remove
Click to add
highlighted
columns as
the x-variable
Overlay
checkmark
Change to
Horizontal
It appears that poultry hot dogs have fewer calories on average

than the other two hot dog types. Also note that the poultry hot
dogs have slightly more sodium. The most visible difference is that
the protein-to-fat ratio appears much higher in poultry hot dogs.
Repeat the above steps using taste to produce another set of bar
charts.
Select Calories, Sodium and Protein Fat in the Select Columns
list.
Click Statistics > Mean.
Select Taste from the select columns list.
Click Categories, (X), Levels.
Remove overlay checkmark.
Change to horizontal.
It may not be surprising to see from the bar charts that hot dogs
rated as bland tasting have (on average) more calories, more
sodium, and a lower protein-to-fat ratio. It can be seen from the
data table that scrumptious tasting hot dogs have the lowest average calories and sodium content. However, medium tasting hot
dogs have the highest protein-to-fat ratio, and they compare well
with respect to the other nutritional factors.
3 Summarizing Data
Figure 3.7 Charting Data
38
Summarizing Data
Grouping Data
Chapter 3
Charting Statistics for Two Groups

Next, it is useful to know the frequency of the three taste responses for each type of hot dog.
Make sure the Hot Dogs.jmp table is active and use Tables > Summary again and select both Type
and Taste as grouping variables.
Click OK.
This produces the table shown here. There is one row for each taste response within each type of hot
dog. The N Rows column lists the frequency in the source table of each type-taste combination.
Select the Graph > Chart command, with both grouping variables as Categories (X) and the Nrows
column with the y role.
Select Statistics > Data.
Click OK.
This produces the chart shown here. In this
example, there are side-by-side charts that
show the frequency for each taste within each
type of hot dog.
Note: Graph > Chart can also be used to
directly chart data grouped by two variables;
the data doesnt have to be grouped first by
Tables > Summary.
To label each bar with the frequency it represents:
Label > Label by Value is selected by
default. Right-click the bars and select
Label > Show Labels.
The chart shows that the poultry hot dogs
excelled in nutrition factors and that most
people find them medium-tasting. However,
because the sodium content appears slightly
high in some poultry brands, more investigation is needed.
Chapter 3
Summarizing Data
Finding a Subgroup with Multiple Characteristics
39
Continue the search for the ideal hot dog. Add special markers to the summary table, Hot Dogs by
(Type, Taste), that identify each type of hot dog.
In the Hot Dogs by (Type, Taste) summary table:
Shift-click or click and drag over the medium and scrumptious beef rows (2 and 3) to select them.
Use the Markers command in the Rows menu to assign them the Z marker.
Deselect those rows.
Shift-click or drag the medium and scrumptious meat rows (5 and 6) to select them, assign them the
Y marker, and deselect them.
Shift-click or drag the medium and scrumptious poultry rows (8 and 9), assign them the X marker,
and deselect them.
The type-taste summary table now looks like the one shown here, and the corresponding rows in the
Hot Dogs.jmp table are marked likewise.
Comparative Scatterplots
Now, examine the relevant variables with scatterplots to identify specific points (brands). The Fit Y by
X command in the Analyze menu produces scatterplots when both the x and y are continuous numeric
variables.
The following scatterplots graphically show the relationship of cost and the nutritional factors together.
Click the Hot Dogs.jmp source table to make it active.
Select Analyze > Fit Y by X.
Make your selections in the window, giving $/lb Protein the y role and both $/oz and Protein/Fat
the x role.
Click OK.
This produces $/lb Protein by $/oz and a $/lb Protein by Protein/Fat scatterplots.
Click the red triangle icon and select Group By.
Choose Type as the grouping variable from the list of variables in the Grouping window.
3 Summarizing Data
40
Summarizing Data
Chapter 3
Repeat this action for the $/lb Protein by Protein/Fat scatterplot.

For both plots, click the red triangle icon and choose Density Ellipse > .90 to make a density ellipse
visible.
Repeat to complete a similar Fit Y by X analysis with Calories as y and Sodium as x.
These commands produce the $/lb Protein by $/oz, the $/lb Protein by Protein/Fat, and the Calories
by Sodium scatterplots shown in Figure 3.8 and Figure 3.9.
The 0.90 ellipses in the scatterplots show the shape of the bivariate response for each type of hot dog.
The special markers identify the taste and type of each point.
Figure 3.8 Scatterplots Comparing Cost, Taste, and Nutritional Factors
To further identify and highlight points of interest:

Select the brush tool (
) from the Tools menu.
Press the Alt key (Alt-Shift on Linux and Option on Macintosh) and drag the brush in the lower left
quadrant of the Calories by Sodium scatterplot, as shown in Figure 3.9.
These points represent brands with both low sodium and low calories. The highlighted points of these
healthiest brands also highlight in the other scatterplots.
Figure 3.9 Select Low Sodium and Low Calorie Brands
Chapter 3
Summarizing Data
41
The costs of meat and beef brands range from low to high. However, it is not surprising to see the tight
low-cost cluster of poultry brands (X-marked) at the lower left of the $/lb Protein by $/oz scatterplot.
The highlighted points include poultry brands, one meat brand, and one beef brand. The selected beef
point (Z-marked) is in the upper-right corner of the plot, which places it in the most expensive category. The single meat point (Y-marked) is more costly than the poultry brands but less than the beef
brands.
A bigger surprise appears in the $/lb Protein by Protein/Fat scatterplot. As the protein-to-fat ratio
increases, the cost per pound of protein stays about the same. Further, the poultry brands not only cost
the least but also contain the most protein. Most of the selected points are in the three highest protein
categories.
The density ellipses on the Calories by Sodium scatterplot show clearly that the poultry brands have
about the same range of sodium content as the meat and beef brands, but many poultry brands have
fewer calories.
Finding the Best Points

Now there is sufficient information to identify several hot dog brands as possible cafeteria menu items.
Click the Hot Dogs.jmp table to make it active. Note that the icon beside the Product Name column in the Column panel means that is has been designated as a Label column.
The poultry (X-marked) brands are acceptably economical, and some of them have high protein content. Few meat or beef brands compared well.
Select the Arrow tool from the Tools menu.
Click inside the Calories by Sodium scatterplot to deselect all points.
Shift-click (hold down Alt-Shift and click on Linux) to highlight the two poultry brands (X-marked)
in the data table with the least calories and lowest sodium content.
Shift-click to highlight the lone meat point (Y-marked) in the data table that has the least sodium of
all brands, is low in calories, has a moderate protein count, and is average in price.
Select Rows > Label/Unlabel to display the brand names of highlighted points (Thin Jack Veal,
Calorie-less Turkey, and Estate Chicken), as shown in Figure 3.10.
3 Summarizing Data
What Has Been Discovered?
42
Summarizing Data
Chapter Summary
Chapter 3
Figure 3.10 Labeling Ideal Points
As a final step, use Analyze > Fit Y by X to look again at the two scatterplots that compare costs.
Assign $/lb Protein as Y.
Assign both $/oz and Protein/Fat as X.
Click OK.
The plot in Figure 3.11 shows that the Estate Chicken brand is the most economical of the three
labeled brands (showing $/oz as continuous). The plot to the right indicates that the Calorie-less Turkey brand is in the group with the highest proportion of protein (showing Protein/Fat as nominal).
Figure 3.11 Winning Hot Dog Brands.
Chapter Summary
This lesson examined different hot dog brands for a cafeteria menu. A JMP table has data for 54 brands
of hot dog showing type of hot dog, taste preference, nutritional factors, and cost factors.
To find the ideal hot dog, we did the following:
Chapter 3
Summarizing Data
Chapter Summary
43
Used Graph > Chart to chart summary statistics and identify the subset of hot dog brands that are
both the most nutritious and the best tasting.
Assigned different markers to each type of hot dog.
Used Analyze > Fit Y by X to see scatterplots that compare cost factors and nutritional factors.
Selected the points representing the lowest cost, most nutritious, and used the Label/Unlabel command in the Rows menu to identify the Calorie-less Turkey brand as a possible cafeteria hot dog.
See the JMP User Guide for details about the Summary command. For scatterplot and bar chart examples, see the JMP Statistics and Graphics Guide.
3 Summarizing Data
Created a summary table that group the data by hot dog type and by taste preference within each
hot dog type.
Chapter 4
Histograms, moments, quantiles, and proportions
The students in a local school are participating in a health study. This lesson summarizes basic information about the students for the school systems health care specialists. The data collected include age,
sex, weight, and height.
To document the sample of participating students and identify any students with unusual characteristics who may need special attention, we will need to view summaries of the data. This lesson produces
reports with graphs and short, straightforward explanations.
Objectives
Use the distribution analysis to explore several variables at once.
Produce reports of moments, quantiles, frequencies, and proportions.
Use the formula editor to compute a columns value.
Create a subset of a data table.
Contents
Look Before You Leap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Displaying Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
Understanding Histograms of Nominal and Ordinal Variables . . . . . . . . . . . . . . . . . . . . . . . . . . 48
Understanding Histograms of Continuous Values . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
Learning About Report Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Reports for Continuous Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
Frequency Table for Ordinal or Nominal Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .52
Adding a Computed Column. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .52
Creating Subsets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .56
Chapter 4
47
The first step in this analysis is to become familiar with the data in the Big Class.jmp file. Looking at
the information in the JMP data table helps us decide which summary charts and tables to use in the
health report.
Open the Big Class.jmp data table to see the data table shown in Figure 4.1.
Figure 4.1 Big Class.jmp Data Table
The file contains the name, age, sex, height, and weight for each student participating in the health
study. The data table is in order by age, and sex is ordered within each age group.
Even though there are only five columns of information, these variables address the following questions:
How many boys and how many girls are there?
How old are they?
What is the average height and weight of the students?
Are there any students drastically younger or older than the average age?
Are there any students whose height or weight might signal the need for special attention?
Displaying Distributions
To summarize the data:
Select the Distribution command from the Analyze menu.
In the window that appears, select the age and sex columns as Y, Columns.
4 Distributions
48
Chapter 4
Click OK.
The frequencies table that appears shows that the class of 40 contains 18 girls and 22 boys.
Understanding Histograms of Nominal and Ordinal Variables

After selecting Analyze > Distribution and completing the window, you will see a window that displays
histograms for analysis variables. The histogram for ordinal or nominal variables like age and sex has a
bar for each level (value) of the variable.
Figure 4.2 Histogram of the age and sex Variables
Chapter 4
49
In histograms of ordinal and nominal variables, you can display a mosaic plot by clicking
the red triangle icon in the variables title bar and selecting Mosiac Plot. A mosaic plot
(shown in the figure to the right) visualizes the proportion of each ordinal or nominal
level within the sample. It has a section for each level of the variable, where the size of the
section is proportional to the corresponding groups size. Think of a mosaic plot as a bar
chart with its bars stacked end to end.
Understanding Histograms of Continuous Values

If your variables are continuous, the histogram looks slightly different than if the variables are nominal
or ordinal. To see the difference, look at the distribution of the continuous variables height and weight
in the sample of students.
Click the Big Class.jmp data table to make it the active window.
Again choose the Distribution command from the Analyze menu.
Designate the height and weight columns as Y, Columns variables.
Click OK.
Histograms are displayed for height and weight, as shown in Figure 4.3.
Figure 4.3 Histograms of height and weight
Both height and weight appear to have approximately normal (bell-shaped) distributions, but notice
the extremely high weight value. It will be examined more closely later.
It is important to present data in the best possible form. Sometimes it is worthwhile to experiment with
the shape of a histogram by changing the number of bars or altering their arrangement on the axis.
To adjust the histogram bars:
Select the hand from the graph cursor tool bar.
Position the hand on the bars and press the mouse button to grab the plot.
4 Distributions
Displaying Mosaic Plots
50
Chapter 4
Move the hand to the left to increase the bar width and combine intervals (see Graphs and Charts,
p. 14). The number of bars decreases as the bar size increases.
Move the hand to the right to decrease the bar width, showing more bars.
Move the hand up or down to change the boundaries of the bins. The height of each bar adjusts
according to the new number of observations within each bin.
Using Outlier Box Plots
Available by default in histograms with continuous variables, the outlier box plot (see Figure 4.4) is a
schematic that shows the sample distribution and allows identification of points with extreme values,
sometimes called outliers. You can display and hide an outlier box plot by clicking the red triangle icon
in the variables title bar and selecting Outlier Box Plot.
The ends of the box are the 25th and 75th quantiles, also called the quartiles. The difference between
the quartiles is the interquartile range. The line across the middle of the box identifies the median sample value.
The lines extending from each end of the box are sometimes called whiskers. The whiskers extend from
the ends of the box to the outermost data points that fall within the distance computed as quartile
1.5*(interquartile range). Points beyond the whiskers indicate extreme values that are possible
outliers. To label a point, click the point to highlight it, and then select Rows > Label/Unlabel.
The red bracket along the edge of the box identifies the shortest half, which is the most dense 50% of
the observations.
Figure 4.4 Outlier Box Plot
25th percentile
possible outliers
Interquartile
range
75th percentile
shortest half
Using Quantile Box Plots

In histograms whose variables are continuous, you display a quantile box plot by clicking the triangle
icon in the variables title bar and selecting Quantile Box Plot.
A quantile box plot shows the location of preselected percentiles, sometimes called quantiles, on the
response axis. The median shows as a line in the body of the box. The ends of the box locate the 25th
and 75th quantiles. The number of other quantile lines depends on the available space. The accompanying text report lists the data values for each of the standard quantiles. The box also contains a means
diamond. The two diamond points within the box identify the 95% confidence interval of the mean.
The line that passes through the two diamond points spanning the box identifies the sample mean.
Chapter 4
Learning About Report Tables
51
Figure 4.5 Quantile Box Plot and Quantiles Table
means diamond
Learning About Report Tables

JMP produces tables of statistical summaries along with graphs. The tables that JMP produces depend
on whether a variable is continuous, ordinal, or nominal. Click the red triangle icon and select Display
Options to reveal tables that are not available by default.
The diamond-shaped disclosure button (
on Windows and Linux and
at the top-left of each report opens and closes it.
Reports for Continuous Variables

The figure to the right is the table JMP gives along with histograms for the
continuous variable height.
The Quantiles table displays the maximum value, minimum value, and other
values for selected quantiles.
The Moments table displays the mean, standard deviation, and other summary statistics.
Note: You can also obtain more moments by clicking the red triangle icon on
the variables title bar and choosing Display Options > More Moments.
on the Macintosh)
4 Distributions
Looking at the quantile box plot and means rectangle together helps see if data are distributed normally,
as shown in Figure 4.5. If data are distributed normally (bell shaped), then the 50th quantile and the
mean are the same and other quantiles show symmetrically above and below them.
52
Adding a Computed Column
Chapter 4
Frequency Table for Ordinal or Nominal Variables

The report for nominal and ordinal variables has a different table
from those produced for continuous variables. The table JMP
gives along with histograms for the nominal or ordinal (categorical) variables sex and age has frequency tables that show these
items:
Level lists each value of the response variable.
Count lists the number of rows found for each level of a
response variable.
Prob lists the probability of occurrence for each level of a response variable. The probability is
computed as the count divided by the total frequency of the variable, shown at the bottom of the
table.
The following two statistics are not displayed by default. Right-click (Ctrl-click on the Macintosh) the
frequency table and select the Columns menu to reveal them.
StdErr Prob lists the standard error of the probabilities.
Cum Prob
contains the cumulative sum of the column of probabilities.

Be on the alert for any unusual subjects, such as students who have extreme height or weight values. A
good indicator of extreme values is the ratio of weight to height. To examine the ratio of weight to
height, create a new column called ratio, computed as weight divided by height. To do this:
Click the Big Class.jmp data table to make it the active window.
Select Cols > New Column.
To create the new column of weight-to-height ratios, complete the New Column window as in
Figure 4.6.
Type the new name, ratio, in the Column Name area.
The default data type is Numeric and is correct as is.
The modeling type is Continuous and is correct as is.
Click the drop-down menu beside Format and set the format for ratio in the data grid to Fixed Dec
with two decimal places.
Click the Column Properties button and select Formula, as shown in Figure 4.6, to compute values for the new column.
Chapter 4
53
Construct the formula that calculates values for the ratio column as follows:
Highlight the empty term in the formula and select weight from the list of column names in the
upper-left corner of the formula editor.
Press the divide () key on the formula editor keypad.
With the empty denominator term highlighted, select height from the list of column names.
When the formula is complete, click Apply or OK on the formula editor, or just close its window.
The new column called ratio is now in the Big Class data table as shown here. Its values are the computed weight-to-height ratio for each student.
4 Distributions
Figure 4.6 New Column Window
54
Chapter 4
Now look at the distribution of the ratio variable.

Select Analyze > Distribution and assign the new column (ratio) to the Y, Columns role.
Click OK.
One way to identify subjects that have extreme values is to highlight histogram bars for the highest and
lowest values. To highlight more than one bar, press the Shift key and click the desired bars.
Select the bars for the two lowest and one highest bar (Figure 4.7).
Figure 4.7 Histogram of Ratio with Bars Highlighted at Extreme Values
The highlighted bars in the histogram represent a ratio either greater than or equal to 2.25 or less than
1.5. The corresponding points automatically highlight in the data table and in all other reports generated from the Big Class data table.
Chapter 4
55
4 Distributions
Creating Subsets
Looking in the Big Class data table allows examination of the selected rows, but scrolling through a
large data grid can be tedious. For the final report to the health researchers, include a separate list containing only the highlighted studentsthose with extreme values. To do this, use Tables menu commands to create new data tables or modify existing tables.
Select Tables > Subset or click the Big Class red triangle icon and select Tables > Subset.
Click OK to accept the default choices presented in the window.
This creates a new data table that has only the selected rows and columns from the active data table.
The new data table, shown in Figure 4.8, contains only the students that have extreme weight-to-height
ratios. By default, the table is named Subset of Big Class. Change the name by clicking the existing
name (Subset of Big Class) in the panel located on the top left side of the window. The table can be
saved, exported for use in another application, or printed.
Figure 4.8 Data Table Containing a Selected Subset
56
Chapter Summary
Chapter 4
Chapter Summary
In this chapter, the demographic and vital data of students participating in a health study were summarized. The profile was completed using the Distribution command and the data management features
of the JMP data table.
The Distribution command displayed histograms and box plots or stacked (mosaic) bar charts for each
variable assigned the role of response variable (y). Using display and text report options to look more
closely at the data, the following actions were completed:
Adjusted the number of bars and the scale of the histograms
Produced supporting statistical reports showing moments and quantiles of numeric variables and
frequencies and proportions of nominal and ordinal variables
Created a new column in the data table computed as a function of existing columns
Highlighted histogram bars to identify a subset of rows in the data table
Created a new data table from a subset of highlighted rows
Graphs and text reports can be printed directly from JMP. Graphs and reports can be copied to a JMP
journal or into other applications to complete a report for the school system health care specialists.
See the chapter Univariate Analysis in the JMP Statistics and Graphics Guide for more information
about distributions.
Chapter 5
Testing Differences
The company has decided to replace all computer keyboards with the brand that produces the fastest
accurate typing. The employees participated in a study to help decide what kind of keyboards to buy.
The company selected three different brands of keyboards to test. These keyboards were randomly
assigned to three groups of employees with comparable typing skills. The employees completed typing
tests and recorded their words-per-minute scores.
This lesson finds out if the typing scores are significantly better on any one brand of keyboard than on
the others.
Objectives
Use the Fit Y by X command to produce plots and analyses appropriate for a one-way analysis of
variance.
Use JMPs interactive capabilities to examine differences among groups.
Produce text reports to display differences among groups.
Contents
Look Before You Leap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .59
Graphical Display of Grouped Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .59
Choose Variable Roles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Show Points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Fit Means Option . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .61
Fit Quantiles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .63
Comparison Circles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
Quantify Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .65
Analysis of Variance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .65
Mean Estimates and Statistical Comparisons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
Chapter 5

59
The first step is to become familiar with the data. The typing test scores are in a JMP file so that they
can be reviewed and the kind of analysis determined.
When you installed JMP, a folder named Sample Data was also installed. In that folder is a file
named Typing Data.jmp. Open the file Typing Data.jmp.
The Typing Data table appears in the form of a data grid, as shown here.
The data table has columns named brand and speed. The modeling type for each column shows to the
left of each column name in the columns panel. The character variable brand has nominal ( ) values
and the numeric variable speed has continuous ( ) values.
There are 17 rows that represent typing scores for 17 employees. However, the number of participants
in the groups differs because some of the scheduled participants did not show up for the study. Perhaps
other statistics for the groups differ also. In particular:
Is the mean (average) typing speed the same for each brand?
Do any one of the three brands of keyboard stand out from the others?
Does it make a difference as to which brand the employees use?
Graphical Display of Grouped Data

Comparing the mean typing scores of each keyboard brand involves analyzing two variables, so use the
Fit Y by X command from the Analyze menu.
Selecting Fit Y by X allows you to perform:
Categorical analysis when both x and y have nominal or ordinal values
Analysis of variance when x is nominal or ordinal and y has continuous values, as in the example
shown here
Logistic regression when x is continuous and y has nominal or ordinal values
Regression analysis when both x and y have continuous values.
5 Group Means
60

Chapter 5
Choose Variable Roles

To discover if typing speed is related to (dependent on) a brand of keyboard, follow these steps:
Choose Analyze > Fit Y by X.
Select brand as X, Factor and speed as Y, Response (see Figure 5.1).
Click OK.
The plot shown in Figure 5.2 appears.
Figure 5.1 The Fit Y by X Launch Dialog
Selecting Fit Y by X and completing the window produces a statistical analysis appropriate for the variable roles (x and y) and the modeling type (continuous and nominal or ordinal) of each variable.
Y, Response identifies a response (dependent) variable.
X, Factor identifies a classification (independent) variable.
The next step is to choose an analysis that investigates if there is a statistical difference between the
group mean values.
Show Points
Each of the typing test scores is plotted for each brand of keyboard. Note that the distance between tick
marks on the brand axis is proportional to the sample size of each group. The mean typing score for the
total sample is shown as a horizontal line across the plot.
Chapter 5

It is easy to see at a glance that most participants who used the SPEEDYTYPE machines typed faster
than the others.
Fit Means Option

Now look at more graphical information about the distribution of typing scores.
Click the red triangle icon on the title bar and select
Means/Anova.
This produces the appropriate analysis of variance reports.
It automatically turns on mean diamonds, which draws a
95% means diamond for each group, as shown in
Figure 5.3. You can also turn the Mean Diamonds option
on and off from the Display Options submenu. Click the
red triangle icon on the title bar and select the Display
Options submenu. Mean Diamonds will have a check
mark beside it if it is turned on.
5 Group Means
Figure 5.2 Oneway Analysis for speed by brand
61
62

Chapter 5
Figure 5.3 Example of Means Diamonds
The illustration in Figure 5.4 illustrates the means diamond.

The means diamond has a line drawn at the mean (average) value of words-per-minute for each
brand of keyboard.
The upper and lower points of the means diamond span a 95% confidence interval computed from
the sample values for each machine.
The width of each diamond spans the distance on the horizontal axis proportional to the group size.
Overlap lines within each diamond are drawn at ( 2CI 2 ) . For groups with equal sample sizes, the
marks that appear not to overlap indicate that two group means could be significant at the 95% confidence interval.
Figure 5.4 Means Diamond with X Axis Proportional Option Turned On (Left) and Off (Right)
overlap marks
95% confidence interval
group mean
total response sample mean
The mean scores of the REGAL and WORD-O-MATIC keyboards appear to be nearly the same, but note
that the SPEEDYTYPE mean is much higher (Figure 5.3).
Chapter 5

63
The next logical step is to check the distribution of points within each group. This gives a better idea of
the spread of the values and shows the distance of extreme values from the center of the data.
Click the red triangle icon and select Quantiles.
When you select the Quantiles command, JMP automatically overlays a quantile box plot on each
group of typing scores, as shown in Figure 5.5. JMP also displays the report in Figure 5.5, which lists
the standard percentiles for each keyboard. The median (50th percentile) is the typing speed that
divides the sample in half. This means that 50% of the employees had speeds greater than the median,
and the other half had lower speeds.
Figure 5.5 Fit Quantiles Option
Figure 5.6 illustrates the quantile box plot. The median, or 50th quantile, shows as a line in the body of
the box. The top and bottom of the box represent the 75th and 25th quantiles, also called the upper
and lower quantiles. The box encompasses the interquantile range of the sample data. The 10th and
90th quantiles show as lines above and below each box.
Looking at the quantile box plot and the means diamond together helps show if data are distributed
normally within a group. If data are normally distributed (bell shaped), the 50th percentile and the
mean are the same and the other quantiles are arranged symmetrically above and below the median.
Figure 5.6 Quantiles Box Plot
sample mean
group mean
90th percentile
75th percentile
50th percentile
25th percentile
10th percentile
5 Group Means
Fit Quantiles
64

Chapter 5
The quantile box plots (Figure 5.5) show a difference in variation of scores across the three groups. The
scores in the REGAL group cluster tightly around the mean score but the WORD-O-MATIC scores show
much more variation. However, even with this variation among the groups, the SPEEDYTYPE brand
still appears to promote the best performance.
Comparison Circles
To complete the typing data inspection:
Click the red triangle icon and choose
Compare Means > All Pairs, Tukey HSD.
This option produces statistical reports (discussed later) and automatically draws a set of
comparison circles to the right of the plot that provides a graphical test of whether the mean typing
scores are statistically different. Comparison circles for the three word-processor groups are
shown in Figure 5.7.
The center of each circle is aligned with the mean
of the group it represents. For the Students t-test,
the diameter of each circle spans the 95% confidence interval for each group. Whenever two circles intersect, the confidence intervals of the two
means overlap, suggesting that the means may not be significantly different. Whenever two circles do
not intersect, the group means they represent are significantly different.
Click the SPEEDYTYPE comparison circle.
This graphically illustrates that the SPEEDYTYPE machine is statistically better than the other
machines. The comparison circles highlight to show the statistical magnitude of the difference between
typing scores. Circles for groups that are statistically the same have the same color.
Figure 5.7 Comparison Circles
The comparison circle for the SPEEDYTYPE brand does not intersect with either of the other two. The
REGAL and WORD-O-MATIC brands are statistically slower than SPEEDYTYPE but do not appear dif-
Chapter 5

Quantify Results
65
Quantify Results
Now, examine the report beneath the plot that consists of several tables. The Summary of Fit table,
shown in Figure 5.8, summarizes the typing data distribution with these statistics:
Rsquare (R2) quantifies the proportion of total variation in the typing scores resulting from different keyboards rather than from different people.
Rsquare Adj
parameters.
adjusts R2 to make it more comparable over models with different numbers of
Root Mean Square Error (RMSE) is a measure of the variation in the typing scores that can be
attributed to different people rather than to different machines.
Mean of Response
Observations
is the mean (average) of all the typing scores.
is the total number of scores recorded.
Figure 5.8 Analysis of Variance Report for speend by brand
Analysis of Variance
When you select the Means/Anova command from the red triangle icon in the title bar, JMP gives you
a standard analysis of variance table. If there are only two group levels, the report also includes a t-test
table.
Note that the value of the F-probability (Prob>F) for the Analysis of Variance is 0.0004. This implies
that differences as great as seen in this typing trial are expected only four times in 10,000 similar trials if
the keyboards did not really promote different typing performances.
The Analysis of Variance table has the following information:
Source
lists the sources of variation: brand, Error, and C. Total.
DF is the degrees of freedom associated with the three sources of variation.

Sum of Squares
(SS for short) identifies the sources of variation in the typing scores.
C. Total is the corrected total SS. It divides (partitions) into the SS attributable to brand and the
SS for Error. The brand SS is the variation in the typing scores explained by the analysis of variance
model, that hypothesizes the keyboards are different. The Error SS is the remaining or unexplained
variation.
5 Group Means
ferent from each other. A later section, Mean Estimates and Statistical Comparisons, p. 66, discusses
the multiple comparison tests the comparison circles represent.
66

Quantify Results
Mean Square
F Ratio
Chapter 5
is the sum of squares divided by its associated degrees of freedom.
is the model mean square divided by the error mean square.
Prob > F is the probability of obtaining a greater F-value if the mean typing scores for the keyboards differed only because different people were typing on them rather than because the keyboards promoted different scores in any way.
Mean Estimates and Statistical Comparisons

To see the list of means for each group, look at the Means for Oneway Anova table. This table summarizes the scores for each brand and reveals what level of performance to expect.
The Means for Oneway Anova table shows the following information:
Level
lists the name of each group.
Number
is the number of scores in each group.
Mean is the mean of each group.

Std Error
is the standard error of each group mean.
Lower 95%
is the lower 95% confidence interval for the group means.
Upper 95%
is the upper 95% confidence interval for the group means.
When you select the Compare Means command from the red triangle icon in the title bar, JMP gives
several multiple comparison options to statistically compare pairs of groups. This example uses the All
Pairs, Tukey HSD option, which performs a statistical means comparison for the three pairs of means
using the Tukey-Kramer HSD (honestly significant difference) test (Tukey 1953, Kramer 1956). This
means comparison method compares the actual difference between group means with the difference
that would be significantly different. The difference needed for statistical significance is called the LSD
(least significant difference).
The graphical results show as the comparison circles previously seen in Figure 5.7. The circles centers
represent the actual difference in the group means. The corresponding report is the Means Comparisons table (Figure 5.9), which shows the actual absolute difference between each mean and the LSD.
The top half of the report gives information based on a Students t comparison of each pair. The bot-
Chapter 5

Chapter Summary
67
Figure 5.9 Means Comparisons Table for Tukey-Kramer HSD
Chapter Summary
In this chapter, the difference in mean typing scores for three brands of keyboard was summarized using
the Fit Y by X command in the Analyze menu. This command was also used to:
Plot the typing scores for the three brands of keyboard.
Overlay a means diamond on each group of typing scores to compare the means of each group.
Overlay a quantile box plot on each group of typing scores to compare the shape of the distribution
of scores in each group.
Produce comparison circles to visualize the difference in mean typing scores.
Compute and display a one-way analysis of variance table, which confirmed that at least one pair of
means is statistically different.
Display a table of the group means and standard errors.
Display a table showing the multiple comparison statistical test results for group means.
Using the selection tool ( ) from the Tools menu, the graphs or tables can be copied and prepared in
a report for presentation. The analysis concludes that, in this typing trial, the SPEEDYTYPE keyboard
produced significantly higher scores than either of the other two brands.
See the chapter Oneway Layout of the JMP Statistics and Graphics Guide for a complete discussion of
one-way analysis of variance.
5 Group Means
tom half shows the results of the Tukey-Kramer multiple comparison tests. Pairs with a positive value
are significantly different. The Means Comparison table confirms the visual results in Figure 5.7.
Chapter 6
Comparing Proportions
Survey data are frequently categorical data rather than measurement data. Analysis of categorical data
begins by simply counting the number of responses in categories and subcategories. Counting is easy,
but interpreting the relationship between categories based on counts is more complex. It requires computing probabilities and evaluating the likelihood of these probabilities compared to expectations.
For example, an American automobile manufacturerfeeling the pinch of competition from foreign
auto salesneeds a market analysis before proceeding with a multimillion-dollar advertising campaign.
A random sample of people is surveyed. The auto manufacturer wants to know each participants age,
sex, marital status, and auto information. The auto information consists of the manufacturing country,
the cars size, and the cars type, and whether it is a family, work, or sporty car. This information may
provide the advertising experts with direction for the upcoming advertising campaign.
Who buys what?
Objectives
Use the Fit Y by X command to compare two variables consisting of categorical data.
Use the formula editor to re-code a categorical variable as a numeric variable.
Produce and examine graphs and statistics appropriate for the comparison of proportions such as
Chi-squared tests and mosaic plots.
Contents
Open a Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .71
Address the Research Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .71
Modify the Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .71
Contingency Table Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .75
Cast Variables Into Roles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .75
Contingency Table Mosaic Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .75
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Chapter 6

71
The first step is to become familiar with the data. Begin by reviewing the data to determine the best
way to proceed with the market analysis.
Open a Data Table

When you installed JMP, a folder
named Sample Data was also
installed. In that folder is a file
named Car Poll.jmp. Open Car
Poll.jmp, as shown here.
The Car Poll data were collected
from a random sample of people in a
specific geographic area. The columns panel shows that Age is a
numeric variable and is assigned the
continuous ( ) modeling type. The
other five columns are character variables with nominal (
) modeling types.
Address the Research Question

The basic research question asks,Is the response probability for country of manufacture, size of car, or
type of car a function of the age, sex, or marital status of the owner?
Look at the data table to see what specific relationships lend insight into this question. You are interested in the relationships between the following automobile characteristics and demographics:
manufacturing country by age
manufacturing country by sex
manufacturing country by marital status
size of car by age
size of car by sex
size of car by marital status
type of car by age
type of car by sex
type of car by marital status
Modify the Data Table

You can sometimes obtain better summary information from age groups rather than specific ages. In
fact, dividing people into two age groups is often the basis for a valuable broad analysis. So, lets find the
median agethe age that divides the sample into two equal age groups.
6 Categorical Data
72

Chapter 6
The distribution of a variable and its corresponding quantiles display a good way to form sample
groups. Use the distribution of the age column to find a reasonable value of age that divides the sample
into two groups.
Select Analyze > Distribution.
When the Distribution window appears, select age as the analysis column (Y, Column) and click
OK.
JMP displays a histogram with an accompanying outlier box plot, Quantiles table, and Moments table.
The Quantiles table, shown here, identifies 30 as the median age.
The next step is to create a new column whose values identify whether a subjects age is greater than 30,
or is less than or equal to 30.
Select Cols > New Column to display the New Column window, shown in Figure 6.1, which is
used to define column characteristics.
Data Type, Modeling Type, and Format options define the new columns characteristics. Enter charac-
teristics for the new column as follows:

Type the new name (call it age group) in the Column Name text box.
Because the new column has grouping values instead of measurements, select Character from the
box beside Data Type.
Figure 6.1 New Column Window
Click the Column Properties button and select Formula.

You are presented with the formula editor window shown in Figure 6.2.
Chapter 6

73
column selector list
6 Categorical Data
Figure 6.2 Formula Editor Window
function selector list
Suppose 0 represents ages greater than the median (30) and 1 represents the ages less than or equal to
the median. To create a formula that divides the sample into two groups, follow these steps:
Click Conditional in the function selector list and select the If function.
Highlight the expression term, denoted expr.

Choose a<=b from the Comparison functions.
74

Chapter 6
.
Highlight the left side of the comparison clause and click age in the Table Columns list.
Double-click the right side of the comparison clause to obtain a text entry box.
Enter 30 for the numeric comparison.
Double-click the term denoted then clause.
Enter 1 (in double quotes because this column is a character variable).
Double-click the term denoted else clause.
Enter 0 (with double quotes).
The complete equation should look like the one shown here.
Click Apply, OK, or the formula editors close box to fill the new column with calculated values.
Tip: Instead of using the buttons in the formula editor, you can double-click the outermost nesting box
to create a single text entry box and enter if(age<=30, 1, 0). Then, press Enter (or Return) or click
outside the text box, and the formula appears in formatted form.
Chapter 6

Contingency Table Reports
75
The nominal age grouping variable shows the relationship of age to the
other nominal variables using contingency tables. To look at combinations
of two variables:
Choose Analyze > Fit Y by X.
JMP does the statistical analysis appropriate for a variables modeling types
and role assignments.
Cast Variables Into Roles

Assign analysis roles to variables by choosing an analysis from the Analyze menu and making selections
in the window that appears. In this investigation, the country, size, and type columns are dependent
response (y) variables; sex, marital status, and age group are independent (x) variables. This example
shows how to complete the Fit Y by X window (Figure 6.3).
Select the three y variables (country, size, and type).
Click the Y, Response button.
Assign the x variables (sex, marital status, and age group) by selecting them and clicking the X,
Factor button.
Click OK when finished.
Figure 6.3 The Fit Y by X Window
Contingency Table Mosaic Plots

If both x and y have either nominal or ordinal values, JMP displays a mosaic plot with accompanying
text reports for each combination of columns assigned x and y modeling roles.
6 Categorical Data
76

Chapter 6
A mosaic chart has side-by-side divided bars for each level of its x variable. The bars are divided into
segments proportional to each discrete level (value) of the y variable. The mosaic chart in Figure 6.4
shows the relationship of marital status to the manufacturing country.
The width of each bar is proportional to the sample size. When the lines dividing the bars align horizontally, the response proportions are the same. When the lines are far apart, the response rates of the
samples might be statistically different.
Figure 6.4 Mosaic Plot Axes
response rates
proportion of married people
with Japanese cars
width of x-axis is proportional

to number in sample
Scroll to see each x variable as it relates to manufacturing country.
Sex and country do not appear to have any relationship at all. The proportion of automobiles from the
three manufacturing countries is about the same for each sex.
The country by age group mosaic plot shows that the proportion of American car owners 30 years or
over is only slightly greater than the proportion of American car owners under age 30.
The most significant relationship is seen between marital status and country. The mosaic plot, shown
previously in Figure 6.4, and its supporting Tests table (Figure 6.5), suggest that married people are
more likely than single people to own American cars.
The Likelihood Ratio and Pearson Chi-squared tests evaluate the relationship between an automobiles
country of manufacture and the marital status of owner. If no relationship exists between country and
marital status, a smaller Chi-squared value than the one computed in this survey would occur only
seven times in 100 similar surveys.
Figure 6.5 Table of Statistical Tests for Marital Status By Country
Chapter 6

77
Scroll the report to see the relationship between size of car and each x variable (sex, marital status,
and age).
The three mosaic plots indicate no relationship between car size and gender, marital status, or age
group. This is seen numerically by looking at the Contingency Tables and the Tests tables beneath each
of the mosaic plots (see Figure 6.6).
Note that by default, Col% and Row% also appear in the Contingency Tables. Right-click (Ctrl-click
on the Macintosh) the table to access the Columns menu to turn columns on and off.
The Chi-squared values support the hypothesis that the purchase of large, medium, and small cars is
not significantly different across the sex, marital status, and age group factor levels. The Chi-squared
probabilities range from 0.06 to 0.30, so you should expect smaller Chi-squared values to occur six to
30 times in 100 similar surveys.
It probably makes no difference what size cars appear in advertisements.
Figure 6.6 Tables for Relationships with Size of Car
The market survey categorizes cars based on both size and type, where a cars type is work, sporty, or
family.
Scroll to see the plots that show the relationship between type of car and the three x variables.
The mosaic plots in Figure 6.7 show that the type of car varies for levels of marital status and age
group. As perhaps expected, many of the cars owned by married people are family automobiles, while
the largest proportion of cars owned by single people are sporty cars.
6 Categorical Data
These statistical results reveal that American auto manufacturers might want to direct advertising plans
toward married couples.
78

Chapter Summary
Chapter 6
Figure 6.7 Reports for Type of Car and Marital Status and Age Group
So, American automobile manufacturers may choose to focus advertisements toward married couples
buying family-type automobiles.
It follows logically that a relationship between age group and type of car also exists because older people
are more likely to be married. The graph to the right in Figure 6.7 shows graphically that the proportion of people over 30 years old who own family cars is much greater than those under 30. The small
Chi-squared values support the significant difference in proportions. The Chi-squared values of 0.0005
mean that proportions as varied as these are expected to occur only five times in 1,000 similar surveys.
Chapter Summary
This chapter looked at relationships between categorical variables obtained from a survey. The survey
recorded age, sex, marital status, and information about the type of automobile owned by a random
sample of people in the same geographical area. The auto information included manufacturing country,
size, and type of car. Car types were classified as work, sporty, and family. The question Is the size of
car, type of car, or manufacturing country related to the age, gender, or marital status of the owner?
was investigated.
The Fit Y by X command produced nine mosaic charts with supporting statistical summaries that
show:
No relationship between either sex or age and manufacturing country.
Chapter 6

Chapter Summary
79
No relationship between sex, age, or marital status and size of car.

No relationship between sex and type of car.
Significant relationships between marital status and type of car. As might be expected, married people over 30 years old were more likely to own family type cars than younger, single people.
The chapter Contingency Tables Analysis in the JMP Statistics and Graphics Guide discusses analyzing
categorical data in more detail.
For more information about using the formula editor, see the chapter Using the Formula Editor in
the JMP User Guide.
6 Categorical Data
A significant relationship between marital status and manufacturing country with married people
more likely to own American cars than single people.
Chapter 7
Visualizing Relationships
This lesson demonstrates the interactive regression capabilities of JMP.
The data is from Eppright et al (1972) as reported in Eubank (1988, p. 272). The study subjects are
young males. The variables in the data table are age (in months) and the ratio of weight to height. A
third variable classifies the subjects into two groups based on age. The goal is to describe and model the
growth pattern of subjects for the age range given in the data table.
Objectives
Use the Fit Y by X command to fit least-squares lines to continuous data.
Fit polynomial curves and cubic splines to the data set and explore their goodness of fit.
Journal and save analysis results.
Use the Group By command to fit different lines to certain groups of data.
Contents
Open a JMP File . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .83
Select an Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .83
Choose Variable Roles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .83
Fitting Models to Continuous Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
Fitting the Mean . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .85
Fitting a Line. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .85
Understanding the Summary of Fit Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
Understanding the Analysis of Variance Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Understanding the Parameter Estimates Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
Excluding Points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .88
Journaling JMP Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .88
Examining a Polynomial Fit (Linear Regression) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Fitting a Spline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .91
Fitting By Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .93
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Chapter 7

83
The first step is to become familiar with the data. Begin by reviewing the data to determine the best
way to proceed with the regression.
Open a JMP File

When you installed JMP, a folder named Sample Data
was also installed. In that folder is a file named
Growth.jmp. Open Growth.jmp.
A partial listing of the Growth.jmp data table is shown here.
There are 2 columns and 72 rows. The ratio column contains the average weight-to-height ratio for each age group
in the study. The age groups range from 0.5 to 71.5
months.
The modeling type for each column is shown to the left of
the variable name in the columns panel. Both columns have
a continuous modeling type ( ), as needed for a regression
analysis in JMP.
The purpose of the analysis is to determine if the ratio values are related to (or dependent on) the age
values.
Select an Analysis
To fit regression curves:
The Fit Y by X analysis does four kinds of analyses, depending on the modeling type of the variable:
Regression analysis when both x and y have continuous values, as in this example.
Categorical analysis when both x and y have nominal or ordinal values.
Analysis of variance when x is nominal and y has continuous values.
Logistic regression when x is continuous and y has nominal or ordinal values.
Choose Variable Roles

The Fit Y by X command first displays the Fit Y by X window. Y identifies a response or dependent
variable and x identifies a classification or independent variable. To choose variable roles:
Highlight ratio and click Y, Response.
Highlight age and click X, Factor, as shown in Figure 7.1.
Click OK.
7 Regression
84

Fitting Models to Continuous Data
Chapter 7
Figure 7.1 The Fit Y By X Window
Now investigate if the ratio of weight to height is a function of age.

The scatterplot shown in Figure 7.2 is the result of the Fit Y by X analysis. It is easy to see that the
growth pattern is not random. A straight line regression is a good baseline fit to compare with other
regression curves.
Figure 7.2 Scatterplot of ratio by age
Chapter 7

Fitting
Options
Begin with a simple line and try different techniques after

inspecting the initial straight line regression fit.
Fitting the Mean

Click the red triangle icon and select Fit Mean.
This is the baseline fit that hypothesizes that there is no relationship between x and y. All other fits compare to this fit. Since the Fit Mean table is closed by default:
Click the disclosure button by Fit Mean (
dows/Linux and
on the Macintosh).
on Win-
This displays a report that shows:

The sample mean (arithmetic average) of the response
variable.
The standard deviation of the response variable.
The standard error of the response mean.
The error sum of squares for the simple mean model.
Fitting a Line
To fit a simple regression line through the data points:
Click the red triangle icon in the title bar and select Fit Line.
The regression line minimizes the sum of squared distances from each point to the line of fit. Because
of this property, it is sometimes referred to as the line of best fit.
7 Regression
When clicked, the red triangle icon on the scatterplot title bar
reveals a variety of fitting commands and additional display
options. Options include Show Points, fitting commands, and
other features. The Show Points command alternately hides or
displays the points in the plot. Fitting options can be as simple
as fitting a straight line or involved as drawing density ellipses.
Fitting options can be used repeatedly to overlay different fits
on the same scatterplot.
85
86

Chapter 7
Red triangle icons

Each time a fit is selected from the red triangle icon, the regression equation and another red triangle
icon for that fit show beneath the scatterplot, as shown here.
Click the red triangle icon to reveal commands that show confidence curves and give the ability to save
predicted and residual values as new data table columns. The Save Predicteds command saves the
prediction equation for the fit with the new column of predicted values. The fit can be removed from
the scatterplot at any time with the Remove Fit command.
Understanding the Summary of Fit Table

Clicking the red triangle icon in the title bar and selecting Fit Line produces a Summary of Fit table,
which summarizes the linear fit.
Rsquare (R2) quantifies the proportion of total variation in the growth ratios accounted for by fitting the regression line.
Rsquare Adj
parameters.
adjusts R2 to make it more comparable over models with different numbers of
Root Mean Square Error (RMSE) is a measure of the variation in the ratio values that is attributable to different people rather than to different ages.
Mean of Response
Observations
is the arithmetic mean (average) of the ratio values.
is the total number of nonmissing values.
Note: The first line of the report is the regression equation, which is editable.
Chapter 7

87
In addition to producing a Summary of Fit table, clicking the red triangle icon and selecting Fit Line
produces an Analysis of Variance table.
The elements of the table give an indication of how well the straight line fits the data points:
Source
DF
identifies the sources of variation in the growth ratio values (Model, Error, and C. Total).
records the associated degrees of freedom for each source of variation.
Sum of Squares (SS for short) quantifies the variation associated with each variation source. The
C. Total SS is the corrected total SS computed from all the ratio values. It divides (partitions) into
the SS for Model and SS for Error. The Model SS is the amount of the total variation in the ratio
scores explained by fitting a straight line to the data. The Error SS is the remaining or unexplained
variation.
Mean Square lists the Sum of Squares divided by its associated degrees of freedom (DF) for
Model and Error.
is the regression (Model) mean square divided by the Error mean square.
F Ratio
Prob > F is the probability of a greater F-value occurring if the ratio values differed only because
of different subjects rather than because the subjects are different ages.
In this example, the significance of the F-value is 0.0001, which strongly indicates that the linear fit to
the weight/height growth pattern is significantly better than the horizontal line that fits the sample
mean to the data.
Understanding the Parameter Estimates Table

In addition to producing a Line of Fit table and an Analysis of Variance table, clicking the red triangle
icon and selecting Fit Line produces a Parameter Estimates table.
Term
lists the parameter terms in the regression model.
Estimate
Std Error
t Ratio
lists estimates of the coefficients in the regression line equation.

lists estimates of the standard error of the parameters.
is the parameter estimate divided by its standard error.
Prob > |t| is the probability of a greater absolute t-value occurring by chance alone if the parameter has no effect in the model.
The significant F-ratio in the Analysis of Variance table tells the student that the regression line fits significantly better than the horizontal line at the mean (the simple mean model). However, while the
7 Regression
Understanding the Analysis of Variance Table
88

Chapter 7
regression line looks like a good fit for age groups above seven months, it does not describe the data well
for ages younger than seven months.
Excluding Points
Because the low-age points are the trouble spots for the linear fit, remove them from the analysis and
try fitting the model to the remaining values.
To highlight these outliers and exclude them from the analysis:
Select the lasso tool from Tools menu or toolbar.
Drag the lasso around the points to be excluded.
Select Rows > Exclude/Unexclude to exclude the selected points.
Right-click the selected points, select Row Markers, and select X to assign the X marker to the
excluded points.
Click the red triangle icon in the title bar and choose the Fit Line command again to see the results
of excluding the low-age points.
The scatterplot shown here has both regression lines. The low-age points still show on the plot but are
not included in the second regression lines computation.
fitted line after excluding points
fitted line for all points
Journaling JMP Results

After completing this part of the exploratory regression analysis:
Choose Edit > Journal.
The first time the Journal command is selected during a JMP session, a journal window opens and fills
with the graphs and tables from the report window.
Chapter 7

89
Choose Save As from the File

menu to save the journal.
The window similar to the one
shown here prompts for a file name,
and appends .jrn to the file name to
identify the file type.
Leaving a journal file open causes
each subsequent use of the Journal
command to append results in the
active window at the end of the journal contents.
Name the journal Regression
Results.
On Windows, change the Save as
type to RTF Files (*.RTF) and
click Save.
On the Macintosh, change the Format type to RTF document and click Save.
On Linux, change the Save as type option to *.rtf (Rich Text File) and click Finished.
Navigate to the files directory on your system and open the file.
The file should open in your default word processor as shown here. Note that the graphics are saved as
graphics, and the reports are saved as text tables.
7 Regression
The open journal file contains all reports from the active report window. Plots can be resized, opened,
or closed, as can outlines. This allows for printing of certain parts of the report.
90

Chapter 7
Examining a Polynomial Fit (Linear Regression)

Now, examine a polynomial fit for comparison to the linear fit. A linear regression is simply a polynomial of degree 1.
Click the Bivariate Fit of ratio By age report window to assure it is active.
To remove the row exclusions and markers, select Rows > Clear Row States.
Click the red triangle icon in the title bar, select Fit Polynomial > 2,quadratic, which allows the fit
to have curvature.
Remove the line of fit that excluded the lower age groups by clicking the second Linear Fit (modified) regression lines red triangle icon and selecting Remove Fit.
Remove the Fit Mean results so that only the polynomial fit and the line fit to all the data points
remain. Select Remove Fit from the Fit Mean red triangle icon.
Click the red triangle icon in the title bar and select Fit Polynomial > 3,cubic to overlay a polynomial curve of degree 3 on the scatterplot.
Again select the Edit > Journal command to append these results to the existing journal.
Chapter 7

91
The tables in Figure 7.3 show the R2 value from the Summary of Fit tables for the linear fit, the second
degree polynomial fit, and the third degree fit. As polynomial terms are added to the model, the regression curve appears to fit the data better.
Fitting a Spline
Even the polynomial fit of degree 3 does not quite reach the outlying points of the very young subjects.
A free-form function that acts as if it smooths the data, such as a smoothing spline, may be better.
Use the Remove Fit command on both polynomial fits, so that only the
first linear regression line shows on the scatterplot.
Click the red triangle icon on the title bar and select Fit Spline three times,
with lambda values of 10, 1,000, and 100,000.
Lambda is a tuning factor that determines the flexibility of the spline. The
Fit Spline command submenu (shown to the left in Figure 7.4) lists lambda
values. The three new fits are overlaid on the scatterplot.
7 Regression
Figure 7.3 Comparison of Linear Fit and Polynomial Fits of Degree 2 and 3
92

Chapter 7
Figure 7.4 Comparison of Spline Fits
linear fit
= 100,000
= 1,000
= 10
By inspecting the plot, see that the lambda = 10 curve is too flexible and therefore local error has too
great an effect on it. The lambda = 100,000 curve is too stiff. It is so straight that it does not reach
down to model the lower ages closely. However, the lambda of 1,000 curve fits well. Its shape is not
influenced by local errors, and it appears to fit the data smoothly.
If a report of these results is needed, journal these results.
Select Edit > Journal.
The Journal command appends the scatterplot with spline fits and
text reports to the open journal file. After journaling the final analyses, the following draft notes about the spline-fitting technique
can be added at the bottom of the journal window.
Select the annotate tool from the tool bar.
Click and drag a large box at the bottom of the report.
Add the following text to the box.
"This fitting technique applies a cubic polynomial to the interval
between points; the polynomial is joined such that the curve meets at the
same point with the same slope to form a continuous and smooth curve. A
small enough lambda could make such a curve go through every point, which
would model the error, not the mean. A moderate lambda value forces the
curve to be smoother, i.e., less curved. This is accomplished by adding a
curvature penalty to the optimization that minimizes the sum of squares
error."
By comparing various regression fits, notice that both the polynomial fits and the spline fit with moderate flexibility best describe the data. These models show that infants grow most rapidly during the first
months of life and that growth rate decreases significantly at approximately 12 months.
Chapter 7

Fitting By Groups
93
Excluding Points, p. 88 in this chapter, shows how to overlay a linear fit for the whole sample with a
linear fit for children over the age of seven months. Carry this idea one step further with overlay fits to
compare children under the age of one year with children over one year.
In the Growth.jmp data table, create a new column called group to act as a grouping variable.
Right-click in the new column area of the data table and select New Column from the resulting
menu. Write the column name and click OK.
Right-click in the Group column (Command-click on the Macintosh) and select Formula.
Enter the formula shown in Figure 7.5.
Click Conditional in the function selector list and select the If function.
The expression term, denoted expr, is highlighted.
Choose a < b from the Comparison functions.
The left side of the comparison clause is highlighted. Click age in the column selection list.
Enter 12 for the numeric comparison.
Double-click the term denoted then clause.
Enter Babies (in double quotes because this column is a character variable).
Double-click the term denoted else clause.
Enter Toddlers (with double quotes).
Click Apply and then OK.
This assigns the value Babies to each child less than 12 months old, and Toddlers to children who are
12 months or older.
Figure 7.5 Computed Age Grouping Variable
Click the Bivariate report to make it the active window.
7 Regression
Fitting By Groups
94

Chapter Summary
Chapter 7
Clear the Smoothing Spline fits still showing, such as those seen in Figure 7.4, using each fits
Remove Fit command. Click the red triangle icon for all three smoothing spline fits and select
Remove Fit for each one.
Click the red triangle icon and select Group By to display the window shown here.
Select group, the newly created grouping variable, and click OK.
Choose the Fit Line command.
With a grouping variable (group) in effect, the overlaid regression lines shown in Figure 7.6 appear
automatically. The points that correspond to each regression give a dramatic visualization of the steep
growth rate for babies during the first year of life compared to the more moderate growth rate of toddlers and small children age one to five years.
Figure 7.6 Regression Lines for Levels of a Grouping Variable
Chapter Summary
To analyze some bivariate data, the Fit Y by X command was used to examine a variety of regression
model fits. The task was to model and describe the growth pattern of subjects over a range of ages. You
measured growth using the ratio of weight to height and accomplished this task by:
Fitting mean to use as a baseline comparison to other regression models and evaluate the fit using
statistical text reports.
Chapter 7

Chapter Summary
95
Excluding outliers and again fitting a straight line to compare the R2 values given by the Summary
of Fit tables for both lines.
Fitting second and third degree polynomials to see if they model the growth pattern more
realistically.
Fitting smoothing splines with lambda values of 10, 1,000, and 100,000 and comparing them with
each other and with the linear fit.
Clicking the red triangle icon and selecting the grouping facility (Group By) to compare growth
rates of babies under the age of one year with toddlers from age one to five years.
Using the Journal command to append each of these regression reports and graphs to a journal file.
7 Regression
Fitting a straight line as a first guess for a model.
Chapter 8
Designed Modeling
This lesson examines two treatments of popcorn. The plain, everyday type has been around for years,
but researchers claim to have discovered a special treatment of corn kernels. This new process supposedly increases the popcorn yield as measured by popcorn volume from a given measure of kernels.
Is this true? If so, how much is the increase? Are these increases the same for all groups of conditions?
The special treatment raises the cost of the popcorn, so the increase in yield must be significant enough
to warrant the higher costs.
The popcorn data used in this chapter and for examples in the JMP User Guide and the JMP Statistics
and Graphics Guide are artificial, but the experiment was inspired by experimental data reported in Box,
Hunter, and Hunter (1978).
Objectives
Learn techniques to analyze a designed factorial experiment using the Fit Model command.
Evaluate and interpret effects using JMPs interactive graphical techniques.
Examine supporting text reports.
Evaluate the significance of interaction effects using interaction plots.
Save a models predicted values for each observation.
Contents
Look Before You Leap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
Open a Data Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
What Questions Can Be Answered?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
The Fit Model Window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .100
Graphical Display: Leverage Plots. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
Quantify Results: Statistical Reports. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .102
Analysis of Variance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
Summary Reports For The Whole Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Summary Reports for Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .107
Chapter 8
99
The popcorn yield data are the result of a designed experiment. The same amounts of different kinds of
corn were methodically popped under different conditions. First, look at the data to review the results
of the popcorn experiment.
Open a Data Table

When you installed JMP, a folder named
Sample Data was also installed. In that
folder is a file named Popcorn.jmp. Open
Popcorn.jmp.
The Popcorn data table displays in spreadsheet form as shown here.
For the experiment, the corn was popped
under controlled conditions. Plain popcorn
and specially-treated gourmet popcorn were
each popped in large or small amounts of oil
and in large or small batches. Two trials were done for both kinds of corn under all popping conditions.
This experimental design is called a factorial design. The experiment has three factors, usually called
main effects, which are:
Kind of popcorn (plain or gourmet)
Amount of cooking oil (little or lots)
Cooking batch size (large or small)
What Questions Can Be Answered?

The appropriate statistical analysis for a factorial design addresses the following questions about the
main effects:
Is there an overall difference in yield between plain and gourmet popcorn?
Is there an overall difference in yield between cooking in lots of oil instead of a small amount of oil?
What is the difference in yield between cooking several small batches instead of one large batch?
Analysis of a factorial experiment also provides information about the interaction between the main
effects as addressed by the following questions:
Does the amount of cooking oil have the same effect on both types of popcorn? In other words, is
there an interaction effect between popcorn type and amount of cooking oil used?
Is there an interaction effect between batch size and type of popcorn?
Is there an interaction effect between batch size and amount of oil used?
Are there interaction effects among the three main effects?
8 Factorial Analysis
100
The Fit Model Window
Chapter 8
The Fit Model Window

Select Analyze > Fit Model.
The Fit Model command lets you specify and analyze complex models like the factorial design in this
experiment.
The Fit Model command displays the Fit Model window shown in Figure 8.1. This window is used to
define the type of model, the model response variable, and model effects.
To specify the factorial model:
Select yield from the Select Columns list.
Click the Y button.
Select popcorn, oil amt, and batch from the Select Columns list.
Click the Macros button and select Full Factorial. This adds all main effects and interactions
(crossed effects) to the Construct Model Effects list (Figure 8.1).
Further tailor the model by adding effects or removing unwanted effects with the Add and Remove
buttons. In this case, remove the three-way interaction term by highlighting popcorn*oil amt*batch
and clicking Remove.
Figure 8.1 Fit Model Window
Select Effect Leverage from the box beside Emphasis at the top right of the Fit Model window.
Click Run Model to estimate the model parameters and view the results.
Chapter 8
Graphical Display: Leverage Plots
101
The Fit Model command graphically displays the whole model and each model effect as the
leverage plots shown in Figure 8.2 through Figure 8.5. It is possible to tell at a glance whether the factorial model explains the popcorn data and which factors are most influential.
The Whole Model plot to the left in Figure 8.2 shows actual yield by predicted yield values with a
regression line and 95% confidence curves. The regression line and the 95% confidence curves cross
the sample mean (the horizontal line), which show that the whole factorial model (all effects together)
explains a significant proportion of the variation in popcorn yield.
There is also a significant difference in yield between the two types of popcorn, as shown in the
right-hand leverage plot for the popcorn main effect. The small p-values beneath the plots quantify the
significant model fit and popcorn effect.
Figure 8.2 Leverage Plots of Actual by Predicted and of Popcorn Effect
In Figure 8.3, the confidence curves for oil amt and the popcorn*oil amt interaction do not cross the
horizontal mean line (rather, they encompass the mean line). This shows that neither of these factors
significantly affected popcorn yield.
Figure 8.3 Leverage Plots for the Oil Amt and Its Interaction with Popcorn
The leverage plots in Figure 8.4 show that the batch size effect (batch) and the interaction between
popcorn type and batch size (popcorn*batch) are significant effects. This means that the size of the
Graphical Display: Leverage Plots
102
Quantify Results: Statistical Reports
Chapter 8
batch makes a difference in the popcorn yield. Furthermore, the significant interaction means that
batch size affects each type of popcorn differently.
Figure 8.4 Leverage Plots for Batch and Its Interaction with Popcorn
The two leverage plots shown in Figure 8.5 show that there is no significant interaction between
amount of oil and batch size.
Figure 8.5 Leverage Plots for Other Interaction Effects
For more information about interpretation of leverage plots, see the chapters Understanding JMP
Analyses and Standard Least Squares: Introduction and the appendix Statistical Details of the JMP
Statistics and Graphics Guide.
Quantify Results: Statistical Reports

Because oil amt and its interactions with other effects are not significant, fit the popcorn data again
without these effects. The new model should have the significant factors (type of popcorn, batch size),
and their interaction term. This approach condenses the statistical reports that show estimates of yield
under the different conditions of interest. Use the same Fit Model command as before.
Click the Fit Model window to make it the active window.
Chapter 8
103
Use the following method to specify the two-factor

model:
From the full factorial model, select unwanted
effects listed in the Construct Model Effects box
and click Remove.
Click Run Model.
The whole model leverage plot in Figure 8.6 shows that the two-factor model describes the popcorn
experiment well. Examine the tables that accompany the whole model leverage plot.
The Analysis of Variance table (in Figure 8.6) that accompanies the whole model leverage plot quantifies the analysis results. It lists the partitioning of the total variation of the sample into components. The
ratio of the Mean Square components forms an F-statistic that evaluates the effectiveness of the model
fit. If the probability associated with the F-ratio is small, then the analysis of variance model fits better
statistically than the simple model that contains only the overall response mean.
If the window is closed, click the red triangle icon on the report and select Script > Redo Analysis
to open a new Fit Model window.
104
Chapter 8
Figure 8.6 Analysis of Variance for the Two-Factor Whole Model
The Analysis of Variance table shows these quantities:

Source
DF
identifies the sources of variation in the popcorn yield values (Model, Error, and C. Total).
records the degrees of freedom for each source of variation.
Sum of Squares (SS for short) quantifies the variation in yield. C. Total is the corrected total SS.
It is divided (partitioned) into the SS for Model and SS for Error. The SS for Model is the variation
in the yield explained by the analysis of variance model, which hypothesizes that the model factors
have a significant effect. The SS for Error is the remaining or unexplained variation.
Mean Square
F Ratio
is a sum of squares divided by its associated degrees of freedom (DF).
is the model mean square divided by the error mean square.
Prob > F is the probability of a greater F ratio occurring if the variation in popcorn yield resulted
from chance alone rather than from the model effects.
In this example, the p-value (Prob > F) is 0.0001. JMP indicates a significant p-value by placing an
asterisk beside it. The low value of this p-value implies that the difference found in the popcorn yield
produced by this experiment is expected only 1 time in 10,000 similar trials if the model factors do not
affect the popcorn yield.
Chapter 8
105
Other tables in the Fit Model report provide statistical summaries. The Summary of Fit table shows the numeric summaries of
the response for the factorial model:
Rsquare (R2) of 0.809 tells the scientist that the two-factor model explains nearly 81% of the variation in the data.
Rsquare Adj adjusts R2 to make it more comparable over
models with different numbers of parameters.
Root Mean Square Error (sometimes called the RMSE) is
a measure of the variation in the yield scores that can be
attributed to random error rather than differences in the
models factors.
Mean of Response
scores.
Observations
is the mean (average) of the yield
is the total number of recorded scores.
The F-test probabilities in the Effect Test table tell the scientist
that all model effects explain a significant proportion of the total variation. JMP indicates a significant
F-value by placing an asterisk beside it. There is also a table that gives the parameter estimates for the
model.
Summary Reports for Effects

Now look at the summary tables for each effect in the model. The tables for the main effects are shown
here. The Least Squares Means table lists the least squares means and standard errors for each level of
the model factors, without considering the interaction between them. In this balanced example, the
least squares means are simply the sample means of each factor level.
The nature of the interaction is important in the interpretation of the popcorn experiment. To examine
the significant popcorn*batch interaction,
Click the red triangle icon from the Response yield title bar and select
Factor Profiling > Interaction Plots.
This command plots the least squares means for each combination of effect levels, as shown in
Figure 8.7.
Summary Reports For The Whole Model
106
Figure 8.7 Interaction Plots
The Least Squares Means table for the popcorn*batch effect

tells the whole story. Batch size makes no difference for the plain
brand popcorn, but popping in small batches increases the yield
in the new gourmet brand.
Because the factorial model with two-factors is a good prediction model, save the prediction formula.
Click the red triangle icon in the Response yield title bar
and select Save Columns > Prediction Formula, as shown in Figure 8.8.
Figure 8.8 Prediction Formula for Popcorn Yield
Chapter 8
Chapter 8
Chapter Summary
107
The prediction formula, shown at the bottom of Figure 8.8, becomes part of the column information.
To see this formula:
Highlight the new column name (Pred Formula yield).
Select Formula from the Cols menu.
The prediction formula can be copied to the clipboard using standard cut and paste techniques.
Results show that popcorn should be packaged:
in small packages so that the yield will be good.
in family size packages with smaller packets inside.
in family size packages with popping instructions that clearly state the best batch size for good
results.
Chapter Summary
In this chapter, a designed experiment evaluated the difference in yield between two types of popcorn.
A three-factor factorial experimental design was the basis for popcorn popping trials. The results were
analyzed by using the Analyze > Fit Model command. The following results were found:
The leverage plots for the factorial analysis of three factors showed one main effect and its associated
interactions to be insignificant.
A more compact two-factor analysis with interaction adequately described the variation in yield for
the popcorn trials.
The interaction between the two main effects was significant. The Least Squares Means table for the
interaction showed how the two types of popcorn behaved under different popping conditions.
The new, more expensive gourmet popcorn had better yield than the plain everyday type only if
popped in small batches.
This command creates a new column in the Popcorn data table called Pred Formula yield that contains the predicted values for each experimental condition.
Chapter 9
Exploring Data
Finding Exceptions
Exploration is the search to find something newthe endeavor to make some discovery. For data analysis, exploratory study is often the most fruitful part of the analytical process because it is the most
open to serendipity. Something noticed in a data set can be the seed of an important advance.
There are two important aspects of exploration:
What is the pattern or shape of the data?
Are there points unusually far away from the bulk of the data (outliers)?
When exploring data composed of many variables, the great challenge is dealing with this high dimensionality. There can be many variables that have interesting relationships, but its hard to visualize the
relationship of more than a few variables at a time.
Objectives
Use graphical techniques to search for outliers in one, two, three, and higher dimensions.
Perform a principal components analysis and examine it graphically.
Examine outliers graphically using Mahalanobis distance.
Contents
Solubility Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
One-Dimensional Views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
Two-Dimensional Views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
Three-Dimensional Views . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
Principal Components and Biplots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
Multivariate Distance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Chapter 9
Exploring Data
Solubility Data
111
This lesson examines compounds for those with unusual solubility patterns in various solvents. When
you installed JMP, a folder named Sample Data was also installed. In that folder is a file named
Solubility.jmp. Data from an experiment by Koehler, Grigorus, and Dunn (1988) are in the
Solubility.jmp file.
Open Solubility.jmp.
There are 72 compounds tested with six solvents, in columns called 1-Octanol, Ether, Chloroform,
Benzene, Carbon Tetrachloride, and Hexane.
The Labels column in the table should serve as a label variable (Figure 9.1) so when you plot them, the
compound names instead of row numbers identify points. Although this is already done for you in
Solubility.jmp, you should know that to assign the label role to columns, you should select the columns,
then select Cols > Label/Unlabel or click the red triangle in the columns panel, make sure Labels is
highlighted, and select Label/Unlabel from the resulting menu.
Figure 9.1 Solubility Data Table
Use this menu to

assign Label role
to a column
There are six solvent variables, but there are no six-dimensional graphics. However, it is possible to look
at six one-dimensional graphs, 15 two-dimensional graphs, and 20 three-dimensional plots. Using
principal components, a representation of higher dimensions can be displayed.
One-Dimensional Views
The Distribution command helps you summarize data one column at a time. It does not show any relationships between variables, but the shape of the individual distributions helps identify the
one-dimensional outliers.
To begin exploring the solubility data:
Choose Analyze > Distribution.
9 Exploring Data
Solubility Data
112
Exploring Data
Solubility Data
Chapter 9
Select the six solubility columns and click the Y, Columns button.
Click OK.
Their histograms, resized and trimmed of other output, are shown in Figure 9.2.
Click any histogram bar.
That bar, and all other representations of that data, are highlighted in all related windows. To see how
outlying values are distributed in the other histograms:
Shift-click the outlying bars in each histogram.
This identifies the outlying rows in each single dimension.
Use the Rows > Markers palette to assign the X marker to these selected rows.
The markers appear in the data table and in subsequent plots.
Figure 9.2 One-Dimensional Views
To create a new data table that contains only the outlying rows:
Use the Tables > Subset command as shown here.
Click OK to accept the default settings.
Scroll through the new subset table to see the compound names of the one-dimensional outliers.
Two-Dimensional Views
Return to Solubility.jmp.
Select Analyze > Multivariate Methods > Multivariate.
Highlight all the continuous columns in the table and click the Y, Columns button.
Click OK.
This displays a correlation matrix and a scatterplot matrix of all 30 two-dimensional scatterplots
(Figure 9.3).
The one-dimensional outliers appear as Xs in each scatterplot. Note in the scatterplot matrix that many
of the variables appear to be correlated, as evidenced by the diagonal flattening of the normal bivariate
density ellipses. There appear to be two groups of variables that correlate among themselves but are not
very correlated with variables in the other group.
Chapter 9
Exploring Data
Solubility Data
113
The variables Ether and 1-Octanol appear to make up one group, and the other group consists of the
remaining four variables. These two groups are outlined on the scatterplot matrix shown in Figure 9.3.
Scan these plots looking for outliers (points that fall outside the bivariate ellipses) of a two-dimensional
nature and identify them with square markers using the following steps.
Double-click on Selected in the rows panel in the Solubility.jmp data table to clear your current
selection.
Shift-click each outlier.
Select Rows >Markers and select the square marker from the palette.
Now, both one- and two-dimensional outliers are identified.
Three-Dimensional Views
To see points in three dimensions:
Double-click on Selected in the rows panel in the Solubility.jmp data table to clear the row selection.
Select Graph > Scatterplot 3D, which opens the Scatterplot 3D window.
Add all six continuous variables to the Y, Columns list.
Click OK.
After the plot appears, change the drop-down lists below the plot to any combination of the three
variables.
9 Exploring Data
Figure 9.3 Two-Dimensional View
114
Exploring Data
Solubility Data
Chapter 9
The goal is to look for points away from the point cloud for each combination of three variables. To aid
in this search:
Rotate and examine each three-dimensional plot by dragging the plot with the mouse.
Hover over points to identify outliers.
Figure 9.4 shows two three-dimensional outlying points in the view of Ether by 1-Octanol by Benzene
that hadnt been apparent before. To label them:
Shift-click these points.
Select Rows > Label/Unlabel.
Their labels, METHYLACETATE and ACETONE, appear on the plot.
Figure 9.4 Spotting Outliers in a Three-Dimensional View
Principal Components and Biplots

Because many of the variables in the Solubility.jmp table are highly correlated, there is not a lot of scatter in six dimensions. The scatter is oriented in some directions but is flattened in other directions.
Chapter 9
Exploring Data
Solubility Data
115
Remove the labels from METHYLACETATE and

ACETONE by selecting Rows > Label/ Unlabel.
Select Analyze > Multivariate Methods >
Principal Components.
Add only two highly correlated variables, in this
case Ether and 1-Octanol.
Highlight Ether and 1-Octanol, select Y,
Columns and click OK.
Click the red triangle icon in the Principal Components title bar and select Spin Principal
Components.
The results are shown in Figure 9.5. Note that because the data are highly correlated, the scatter in the
points runs in a narrow ellipse whose principal axis is oriented in the direction marked Prin1.
Figure 9.5 Two Correlated Variables with Principal Components
To see the greatest variation of the data in one dimension,

Rotate the axis so that the first principal component, Prin1, is horizontal.
Principal components capture the most variation possible in the smallest number of dimensions.
Use principal components to explore all six dimensions with the following steps:
Select Analyze > Multivariate Methods > Principal Components.
Add all six continuous variables to the Y, Columns list.
Click OK.
9 Exploring Data
To illustrate this:
116
Exploring Data
Multivariate Distance
Chapter 9
Click the red triangle icon in the Principal Components/Factor Analysis title bar and select
Eigenvectors.
The result is the Principal Components table in Figure 9.6. The cumulative percent row (Cum Percent)
shows that the first three principal components account for 97.8% of the six-dimensional variation.
Figure 9.6 Principal Components Report
Multivariate Distance
The basic concept of distance in several dimensions relates to the
correlation of the variables. For example, in a Multivariate scatterplot cell for Benzene by Chloroform (Figure 9.3),
HYDROQUINONE is located away from the point cloud. This
compound is not particularly unusual in either the x or y direction alone, but it is a two-dimensional outlier because of its
unusual distance from the strong linear relationship between the
two variables. The ellipse is a 95% density contour for a bivariate
normal distribution with the means, standard deviations, and
correlation estimated from the data. The concept of distance
that takes into account the multivariate normal density contours
is called Mahalanobis distance.
Though only three dimensions can be visualized at a time, the
Mahalanobis distance can be calculated for any number of dimensions. To produce a plot of the Mahalanobis distance:
Select Outlier Analysis > Mahalanobis Distances from the menu accessed by the red triangle at
the top of the multivariate report.
Figure 9.7 shows the Mahalanobis distance by the row number for each data point. To label these
points:
Select the brush tool (
) from the tools palette.
While holding down the Shift key, drag the brush over the points labeled in Figure 9.7. These are
the five points with the greatest Mahalanobis distances.
Select Rows > Label/Unlabel.
Chapter 9
Exploring Data
Chapter Summary
117
Chapter Summary
In this example, commands from the Analyze and Graph menus were used for data exploration to
locate and identify unusual points. The data were first examined in one dimension using the
Distribution command and then in two dimensions using the Multivariate command to look for
unusual points in histograms and scatterplots.
Next, the Principal Components command was used to plot three columns at a time. This technique
was used to summarize six dimensions and to plot principal component rays. The Principal Components table showed that the first three principal components accounted for more than 97% of the total
variation.
Finally, the Outlier Analysis command in the Multivariate report produced the Mahalanobis outlier
distance plot, which summarizes the points in six dimensions. The multivariate outliers were highlighted and labeled in this multi-dimensional space.
See the chapter Correlations and Multivariate Techniques, in the JMP Statistics and Graphics Guide,
for documentation and examples of multivariate analyses. Three-Dimensional Scatterplots in the
JMP Statistics and Graphics Guide documents the 3D plot.
9 Exploring Data
Figure 9.7 Mahalanobis Distance Plot to See Multivariate Outliers
Chapter 10
Multiple Regression
Examining Multiple Explanations
Multiple regression is the technique of fitting or predicting a response by a linear combination of several regressor variables. The fitting principle is like simple linear regression, but the space of the fit is in
three or more dimensions, making it more difficult to visualize. With multiple regressors, there are
more opportunities to model the data well, but the process is more complicated.
This chapter begins with an example of a two-regressor fit that includes three dimensional graphics for
visualization. The example is then extended to include six regressors (but unfortunately no
seven-dimensional graphics to go with it).
Objectives
Illustrate the concept of a fitting plane using graphical techniques.
Combine data tables using the Concatenate command.
Explore a three-dimensional version of a leverage plot.
Contents
Aerobic Fitness Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
Fitting Plane. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Fit Planes to Test Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
Whole Model Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
More and More Regressors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
Interpreting Leverage Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
Collinearity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Chapter Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 130
Chapter 10
Multiple Regression
Aerobic Fitness Data
121
Aerobic fitness can be evaluated using a special test that measures the oxygen uptake of a person while
running on a treadmill for a prescribed distance. However, it would be more economical to evaluate fitness with a formula that predicts oxygen uptake with simpler measurements.
To identify such an equation, runtime and fitness measurements were taken for 31 participants who ran
1.5 miles. The participants ages were also recorded.
When you installed JMP, a folder named Sample Data was also installed. In that folder is a file
named Fitness.jmp. Open Fitness.jmp.
The data are shown in Figure 10.1. For purposes of illustration, certain values of MaxPulse and
RunPulse have been changed from data reported by Rawlings (1988, p. 124).
Figure 10.1 Partial Listing of the Fitness.jmp Data File
Investigate Age and Runtime as predictors of oxygen uptake using the Fit Model platform.
To examine a multiple regression model with two effects,
Highlight Oxy in the Select Columns list and click Y.
Highlight both Age and Runtime.
Click Add to specify them as model effects.
You should now see the completed dialog shown in Figure 10.2.
122
Multiple Regression
Chapter 10
Figure 10.2 Completed Fit Model Dialog for Multiple Regression with Two Effects
Click Run Model.

You should now see the tables shown in Figure 10.3. These statistical reports are appropriate for a
response variable and factor variables that have continuous values.
Figure 10.3 Statistical Text Reports
Clicking the red triangle icon and selecting Save Columns displays a list of save commands. To save
predicted values and the prediction equation for this model:
Click the red triangle icon and select Save Columns > Prediction Formula.
Chapter 10
Multiple Regression
Fitting Plane
123
Right-click the Pred Formula Oxy column name and select Formula.
The Formula window opens and displays the formula
88.4356809 + -0.1509571 * Age + -3.1987736 * Runtime
This formula defines a plane of fit for Oxy as a function of Age and Runtime.
Click Cancel to close the window and return to the data table window.
Fitting Plane
JMP can show relationships between Oxy, Runtime, and Age in three dimensions with a surface plot.
Select Graph > Surface Plot.
Add Oxy and Predicted Formula Oxy as Columns, and click OK.
Change the Style drop down menu to Needles, and view the plot, as shown in Figure 10.4.
Figure 10.4 Initial View of the Surface Plot of Oxy, Age, and Runtime
Click and drag to rotate the plot so it looks like that in Figure 10.5.
This command creates a new column in the Fitness data table called Pred Formula Oxy. Its values are
the calculated predicted values for the model. To see the columns formula:
124
Multiple Regression
Fitting Plane
Chapter 10
Figure 10.5 Observed Points using Age, Oxy, and Runtime with the Predicted Plane of Fit
Observed
Oxy values
Plane of
predicted
Oxy values
Fit Planes to Test Effects

The example in the previous section showed a plane fit to the whole model. You can also look at
hypothesis tests for each regressor to test whether the regressors parameter is significantly different from
zero.
One way to view this test is to evaluate the difference between the current fit and the fit that occurs if
the regressor variable is removed from the model.
For example, remove the Runtime variable from the model by following these steps:
First, make sure Fitness.jmp is the active data table.
Select Analyze > Fit Model.
In the Fit Model dialog, select Oxy in the list of variables and click Y.
Select Age in the list of variables and click Add to add Age as a model effect. Then click Run
Model.
Click the red triangle icon in the Response Oxy report and select Save Columns > Prediction
Formula.
The new predicted column (labeled Pred Formula Oxy 2) is calculated using the formula
62.4229492 + -0.3156031*Age
To compare this fitted line with the plane in the previous example,
Select Graph > Surface Plot.
Add Oxy, Pred Formula Oxy, and Pred Formula Oxy 2 as Columns and click OK.
In the Point Response column drop-down menu, select Oxy for both Pred Formula Oxy and Pred
Formula Oxy 2.
Chapter 10
Multiple Regression
Fitting Plane
125
In the Surface drop-down menu, select Both Sides for Pred Formula Oxy and Pred Formula
Oxy 2.
Both this grid and the one in Figure 10.5 represent least squares regression planes, but this plane has a
slope of zero in the orientation of the Runtime axis. Figure 10.6 shows the plot from an angle.
Figure 10.6 Three-Dimensional Plot with Regression Planes
Observed Values (x)
Oxy by Age and
Runtime regression
plane
Oxy by Age
regression plane
Click and drag to rotate the plot.

A rotated view is shown in Figure 10.7. Notice this subset (Age-only model) regression showing as a
line instead of a plane. The view is edge-on for Runtime, which eliminates it from the visual model.
Figure 10.7 Comparison of Three-Dimensional Views
Runtime effect
is edge-on
Open the Ranges outline by clicking the diamond-shaped disclosure button (Figure 10.8).
In the Style drop-down menu, select Needles for both Pred Formula Oxy and Pred Formula
Oxy 2.
126
Multiple Regression
Fitting Plane
Chapter 10
Figure 10.8 Changing the Ranges
Click the Oxy button and change the minimum to 32, the maximum to 60, and the increment to 5.
Click the Age button and change the minimum to 32, the maximum to 60, and the increment to 5.
Your plot shows the bivariate regression plane edge-on and represents the linear combination of the
effects fit by the plane.
Whole Model Tests

The leverage plot in Figure 10.9 shows joint test of the Age, Weight, Runtime, RunPulse, and
MaxPulse effects in the model. This plot compares the full model with the model containing the intercept that fits the overall response mean only. This leverage plot is formed by plotting the actual
observed values on the y-axis and the values predicted by the whole model on the x-axis. The residual
for the subset model is the distance from a point to the horizontal line drawn at the sample mean.
Figure 10.9 Leverage Plot for the Whole Model (Age and Runtime)
More and More Regressors

Its easy to visualize two regressors predicting a response by using fitting planes. But how can this be
done with more regressors when the analysis requires more than three dimensions? In actuality, the fitting, testing, and leverage plot analyses still work for more regressors.
Select Analyze > Fit Model. Fill the window with Oxy as Y and add Age, Weight, Runtime,
RunPulse, and MaxPulse as model effects.
Click Run Model to run the model.
In this case, the prediction formula is
Oxy = 101.3 - 0.2123 Age - 0.0732 Weight - 2.688 Runtime - 0.3703
RunPulse + 0.3055 MaxPulse
Chapter 10
Multiple Regression
Fitting Plane
127
The Age variable seems significant, but Weight does not. The Runtime variable seems highly significant. Both RunPulse and MaxPulse also seem significant, but MaxPulse is less significant than
RunPulse.
Figure 10.10 Statistical Tables for Multiple Regression
Interpreting Leverage Plots

The leverage plots for this example multiple regression model allow visualization of the contribution of
each effect. First, look at the whole-model leverage plot, shown to the right, of observed versus predicted values. This plot illustrates the test for the whole set of regressors.
The Analysis of Variance table in Figure 10.10 shows a highly significant F corresponding to this plot.
The confidence curves show the strong relationship because they cross the horizontal line.
Now examine the leverage plots for the regressors. Each plot illustrates the residuals as they are and as
they would be if that regressor were removed from the model.
The confidence curves in the leverage plot for Age in Figure 10.11 show that Age is borderline significant because the curves barely cross the horizontal line of the mean. Note that the significance of the
Age effect is 0.03 in the text reports (Figure 10.10), which is only slightly different from the 0.05 confidence curves drawn by JMP.
The leverage plot for Weight shows that the effect is not significant. The confidence curves do not cross
the horizontal line of the mean.
Look at the significance of each regressor with t-ratios in the Parameter Estimates table or F-ratios in
the Effects Tests table (see Figure 10.10). Because each effect has only one parameter, the F-ratios are
the squares of the t-ratios, and have the same significance probabilities.
128
Multiple Regression
Fitting Plane
Chapter 10
Figure 10.11 Leverage Plots for the Age and Weight Effects
The leverage plot for Runtime shows that Runtime is the most significant of all the regressors. The
Runtime leverage line and its confidence curves cross the horizontal mean at a steep angle.
The leverage plots for RunPulse and MaxPulse shown in Figure 10.12 are similar. Each is somewhat
shrunken on the x-axis. This indicates that other variables are related in a strong, linear fashion to these
two regressors, which means the two effects are strongly correlated with each other.
Figure 10.12 Leverage Plots for the RunPulse and MaxPulse Effects
Chapter 10
Multiple Regression
Fitting Plane
129
When two or more regressors have a strong correlation, they are said to be collinear. These regression
points occupy a narrow band showing their linear relationship.When a plane is fit representing collinear regressors, the plane fits the points well in the direction where they are widely scattered. However, in the direction where the scatter is very narrow, the fit is weak and the plane is unstable.
In text reports, this phenomenon translates into high standard errors for the parameter estimates and
potentially high values for the parameter estimates themselves. This occurs because a small random
error in the narrow direction can have a huge effect on the slope of the corresponding fitting plane. An
indication of collinearity in leverage plots is when the points tend to collapse toward the center of the
plot in the x direction.
The Longley.jmp example shows collinearity geometrically in the strongly correlated regressors, X1 and
X2. To examine these regressors, examine Figure 10.13, which shows rotated views of the regression
planes. They illustrate a regression of X1 on Y, X2 on Y and both on Y. Most of the points are near the
intersection of the three planes. All three planes fit the data well, but their vastly different slopes show
that the hold is unstable.
Geometrically, collinearity between two regressors means that the points they represent do not spread
out in x space enough to provide stable support for a plane. Instead, the points cluster around the center causing the plane to be unstable. The regressors act as substitutes for each other to define one direction redundantly. This is cured by dropping one of the collinear regressors from the model. In this case,
drop either X1or X2 from the model, since both measure essentially the same thing.
Figure 10.13 Comparison of RunPulse and MaxPulse Effects
Collinearity
130
Multiple Regression
Chapter Summary
Chapter 10
Chapter Summary
Multiple regression uses the same fitting principle as simple regression, but accounting for significance
is more subtle. Each regressor opens a new dimension for fitting a hyperplane, and its significance is
tested by how much the fit suffers in its absence. When regressors correlate to each other, they are said
to be collinear, and they define directions where the fitting hyperplane is not well supported.
References
Becker, R.A., and Cleveland, W.S. (1987), Brushing Scatterplots, Technometrics, 29, 2.
Belsley, D.A., Kuh, E., and Welsch, R.E. (1980), Regression Diagnostics, New York: John Wiley & Sons.
Box, G.E.P., Hunter, W.G., and Hunter, J.S. (1978), Statistics for Experimenters, New York: John Wiley
& Sons, Inc.
Daniel C. and Wood, F. (1980), Fitting Equations to Data, Revised Edition, New York: John Wiley &
Sons, Inc.
Draper, N. and Smith, H. (1981), Applied Regression Analysis, 2nd Edition, New York: John Wiley &
Sons, Inc.
Eppright, E.S., Fox, H.M., Fryer, B.A., Lamkin, G.H., Vivian, V.M., and Fuller, E.S. (1972),
Nutrition of Infants and Preschool Children in the North Central Region of the United States of
America, World Review of Nutrition and Dietetics, 14.
Eubank, R. L., (1988), Spline Smoothing and Nonparametric Regression, New York: Marcel Dekker.
Gabriel, K.R. (1982), Biplot, Encyclopedia of Statistical Sciences, Volume 1, Kotz and Johnson editors,
New York: John Wiley & Sons, Inc.
Hartigan J.A. and B. Kleiner (1981), Mosaics for Contingency Tables, Proceedings of the 13th
Symposium on the Interface between Computer Science and Statistics, W. F. Eddy editor, New York:
Springer.
Hawkins, D.M., (1974), The Detection of Errors in Multivariate Data Using Koehler, Grigorus,
Dunn (1988), The Relationship Between Chemical Structure and the Logarithm of the Partition,
QSAR, 7.
Koehler, M.G., Grigorus, S., and Dunn, J.D. (1988), The Relationship Between Chemical Structure
and the Logarithm of the Partition Coefficient, Quantitative Structure Activity Relationships, 7.
Leven, J. R., Serlin, R. C., and Webne-Behrman, L. (1989), Analysis of Variance Through Simple
Correlation, American Statistician, 43.
Mosteller, F. and Tukey, J.W. (1977), Data Analysis and Regression, Reading Mass: Addison-Wesley.
Rawlings, J. O., Pantula, S.G., and Dickey, D.A. (1998), Applied Regression Analysis: A Research Tool2nd ed., New York, NY: Springer-Verlag New York Inc.
Sall, J. P. (1990), Leverage Plots for General Linear Hypotheses, American Statistician, 308-315.
SAS Institute (1987), SAS/Stat Guide for Personal Computers, Version 6 Edition, Cary NC: SAS Institute
Inc.
Snedecor, G.W. and Cochran, W.G. (1967), Statistical Methods, Ames Iowa: Iowa State University
Press.
132
References
Winer, B.J. (1971), Statistical Principals in Experimental Design, 2nd Edition, New York: McGraw-Hill,
Inc.
Index
JMP Introductory Guide
Symbols
? tool 3, 5
A
Add Columns 24
Add Rows 25
Add Statistics Column 35
All Pairs, Tukeys HSD 64

analysis methods see platforms
Analysis of Variance see Fit Y by X, Fit Model
Analysis of Variance table 65, 87, 104
analysis role 13
analysis type see modeling type 10
analyze categorical data 6979
Analyze menu 13
Annotate cursor 29
arrow cursor 10
assign role 26
B
bar chart 27
beginners tutorial 3, 8
box plot 50
BP Study.jmp 25
C
C. Total 65
calculator example 53
Car Poll.jmp 71
categorical analysis see Fit Y by X, Fit Model
categorical data 6979
categorical type 10
character column 10
Charts 26, 36
Chi-Square 77
classification variable 13
collinear 129
Column Info 25
column name 24
columns 9
Compare Means 57, 64, 6667
Comparison Circles 64
confidence curves 101, 127
confidence interval 50, 62
construct formula 5253
Construct Model Effects 100, 103
continuous 10, 15, 60, 83
Count 52
Cowboy Hat.jmp 16
create subset 5254
crossed effects 100
Cum Prob 52
cursors in data table 10
curve-fitting 83
D
data grid 9
data table
create 2130
density contour 116
designed experiment 99
DF 65, 87, 104
disclosure control 51
discrete data see nominal, ordinal
Display Options 14
DisplayBox scripting index 5
distance 116
Distribution 72, 111
example 14
documentation overview 7
double-arrow cursor 11
drug experiment example 2130
E
enter data 26
Estimate 87
Example button 4
134
Index
Example button in Index Help 4

Exclude/Include 88, 90
explanatory variable 13
explore data 109117
extend selection 12
extreme values 52, 63
F
F Ratio 66, 87, 103104
F Statistic 103
Factor role 13
factorial analysis example 97107
fit by groups 93
Fit Line 85, 94
Fit Model 100
fit plane 124
Fit Polynomial 90
Fit Y by X 39, 59, 75
Fit Spline 91
Fitness.jmp 121
formula 52
example 53
prediction 107, 123
F-probability see Prob > F
Freq role 13
frequencies see Distribution
frequency table 15, 52
G
Graph menu 13
Group By 9495
grouped charts see Charts
grouped fitting 93
grouping data 34, 37
grouping variable 13
Growth.jmp 83
H
hand cursor 12, 49
Help
using online help 3, 5
high dimensionality 109
highlight see Select
histogram see Distribution
honestly significant difference see Tukey-Kramer
HSD
hotdog example 3143
Hotdogs.jmp 33
I
I-beam cursor 11
independent variable 13
Index tab on JMP Starter 4
interaction 99, 105
interquartile range 50
J
JMP Starter window 8
Journal 88, 92
journaling analysis results 88
JSL Operators menu item 5
L
Label/Unlabel 50, 111
lambda 91
large cross cursor 11
Launch button 4
Launch button in Help Index 4
Least Squares Means table 105106
Level 52, 66
leverage plot 102, 126
Likelihood Ratio 76
linked table 35
local error 92
logistic regression see Fit Y by X, Fit Model
Longley.jmp 129
M
magnifier tool 29
Mahalanobis distance 116
main effect 99100
Markers 88, 112
Mean 66
Mean of Response 65, 86, 105
Mean Square 66, 87, 104
Means Diamonds 50, 6162
Means for Oneway Anova table 66
Means, Anova/t-test 61
median 50
menus
tips 6
modeling type 10, 60
modify data table 7174
moments see Distribution
mosaic plot see Distribution
multiple comparison see Compare Means
multiple regression 119, 121130
135
Index
N
New Column 52
computing values 53
New Data Table 8
nominal 10, 15, 60, 75, 83
normal distribution 49, 51
notation used in manuals 7
Number 66
numeric column 10
O
object scripting index 5
Observations 65, 86, 105
Open Data Table 8
operators 5
ordinal 10, 15, 60, 83
outlier 50, 88, 109, 116
outlier box plot see Distribution
P
Parameter Estimates table 87
partitioning 103
pattern in data 109
Pearson Chi Square 76
percentile see quantile
plane fit 124
platforms 13
pointer cursor 12
Polynomial fit 90
post hoc see Compare Means
Prediction Formula 106
Prob 52
Prob > |t| 87
Prob > F 66, 87, 104
Q
Quantile Box Plot 50, 63
Quantile Box Plot see Distribution
question mark (?) tool 3, 5
Quick Reference Guide 7
R
R2 91
references 131
regression analysis 83
regression example 83
regression line 101
regression see Fit Y by X, Fit Model
Remove Fit 86, 91
report options 14
rescale axis 28
resize plot 89
Response role 13
role 13, 26, 75
Root Mean Square Error (RMSE) 65, 86, 105
rows 9
Rsquare 65, 86, 105
Rsquare Adj 65, 86, 105
Run Script
button on JMP Starter 4
S
Save As 89
Save Predicteds 86
Save Prediction Formula 122
saving a JMP session 16

Scatterplot 3D
example 1619
Scatterplot 3D 113
scatterplot matrix 112
scatterplot see Fit Y by X, Multivariate, Overlay
Plot
Script button 4
select rows and columns 12
selection tool 67
Shift-Tab 26
shortest half 50
Show Points 85
smoothing 91
solubility study 109117
Solubility.jmp 111
Source 65, 87, 104
Spin Principal Components 115
spline 91
start JMP 23
statistical index 4
statistical summaries see Distribution
Std Error 66, 87
StdErrProb 52
Subset 5255, 112
Sum of Squares 65, 87, 104
summarizing data 3143
Summary 34
Summary of Fit table 65, 86, 91, 105
Index
Multivariate 112
136
Index
survey data 6979
T
Tab 26
tension 91
Term 87
three-dimensional plots 119130
tick mark 60
Tip of the day 7
Tools 29
Topic Help button 4
t-ratio 87
t-test 87
tutorial 3, 8
tutorial examples
data table 2130
drug experiment 2130
exploratory study 109117
multiple regression 119130
popcorn experiment 97107
regression analysis 83
summarizing data 3143
survey data 6979
typing study 5767
tutorials
learning JMP 3
Typing Data.jmp 59
typing study 5767
W-Z
Weight role 13
weight-height ratio example 83
whiskers 50
Whole-model plot 101
X, Factor role 13
Y, Response role 13
Your Turn
We welcome your feedback.
If you have comments about this book, please send them to
yourturn@sas.com. Include the full title and page
numbers (if applicable).
If you have comments about the software, please send them

to suggest@sas.com.
SAS Publishing gives you the tools to flourish

in any environment with SAS!
Whether you are new to the workforce or an experienced professional, you need to distinguish yourself in this rapidly
changing and competitive job market. SAS Publishing provides you with a wide range of resources to help you set
yourself apart.
SAS Press Series

Need to learn the basics? Struggling with a programming problem? Youll find the expert answers that you
need in example-rich books from the SAS Press Series. Written by experienced SAS professionals from
around the world, these books deliver real-world insights on a broad range of topics for all skill levels.
support.sas.com/saspress
SAS Documentation
To successfully implement applications using SAS software, companies in every industry and on every
continent all turn to the one source for accurate, timely, and reliable informationSAS documentation. We
currently produce the following types of reference documentation: online help that is built into the software,
tutorials that are integrated into the product, reference documentation delivered in HTML and PDFfree on
the Web, and hard-copy books.
support.sas.com/publishing
SAS Learning Edition 4.1

Get a workplace advantage, perform analytics in less time, and prepare for the SAS Base Programming
exam and SAS Advanced Programming exam with SAS Learning Edition 4.1. This inexpensive, intuitive
personal learning version of SAS includes Base SAS 9.1.3, SAS/STAT , SAS/GRAPH , SAS/QC , SAS/ETS ,
and SAS Enterprise Guide 4.1. Whether you are a professor, student, or business professional, this is a
great way to learn SAS.
support.sas.com/LE
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 trademarks of their respective companies. 2008 SAS Institute Inc. All rights reserved. 474059US.0108

Working With JMP

Uploaded by

Copyright:

Available Formats

Working With JMP

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Working With JMP

Uploaded by

Copyright:

Available Formats

Release 8

JMP, A Business Unit of SAS

Creating a JMP Data Table

Set Column Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Comparing Group Means

Analyzing Categorical Data

Regression and Curve Fitting

Quantify Results: Statistical Reports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Credits and Acknowledgments

RECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF

Scintilla is Copyright 1998-2003 by Neil Hodgson <neilh@scintilla.org>. NEIL HODGSON DISCLAIMS

Before you begin using JMP, you should be familiar with:

Learning About JMP

Searching in the Help

What You Need to Know

Windows and Linux

Macintosh Help Menu

Learning About Statistical and JSL Terms

Using the Context-Sensitive Help

Figure 1.1 Help Is Available

Using This Book in Combination with Other Included Books

Conventions Used in this Book

Learning JMP Tips & Tricks

Step 1: Start JMP

Step 2: Open a JMP Data Table

Step 3: Learn About the Data Table

Figure 1.6 The Open Data File Window

Changing a columns width by dragging the column line

Specifying the Values Type

) Values are numeric measurements.

Values are numeric or character classifications.

Data Table Cursor Forms

Click the cell you wish to edit. The cell highlights.

To edit a column name:

Drag to change width of column

Click to deselect column

Selecting Rows and Columns

Click the space that contains the row number.

Ctrl-click (Command-click on the Macintosh) the desired selections.

Figure 1.8 Select Rows and Columns

Casting Columns Into Roles

Identifies a column as a response or dependent variable whose distribution is to be

Step 5: View the Output Report

Step 4: Select an Analysis

Graphs and Charts

Statistical Tables and Text

Figure 1.11 Options for Nominal or Ordinal Variable in a Distribution Analysis

Step 6: Save the JMP Output Report

Open a Data Table

Figure 1.13 Cowboy Hat Data Table

The data table shown in Figure 1.13 appears.

Spin the Cowboy Hat

To have the plot spin itself:

Blood Pressure Study

Creating a JMP Data Table

 Open the JMP application icon to begin a JMP session.

Blood Pressure Study

Creating Rows and Columns in a JMP Data Table

Starting a JMP Session

Open the JMP application icon to begin a JMP session.

Specify five new columns.

Select the four continuous variables in the Select Columns list.

Repeat this action for the $/lb Protein by Protein/Fat scatterplot.