Multivariate and Probabilistic Analyses of Sensory Science Problems (Institute of Food Technologists Series)

248 Pages·2007·3.13 MB·English

by  Jean-François Meullenet

Multivariate and
Probabilistic Analyses of
Sensory Science Problems
TheIFT Pressseries reflects the mission of the Institute of Food Technologists—advanc-
ing the science and technology of food through the exchange of knowledge. Developed in
partnership with Blackwell Publishing, IFT Pressbooks serve as leading edge handbooks
for industrial application and reference and as essential texts for academic programs.
Crafted through rigorous peer review and meticulous research, IFT Press publications
represent the latest, most significant resources available to food scientists and related
agriculture professionals worldwide.
IFT Book Communications Committee
Ruth M. Patrick
Dennis R. Heldman
Theron W. Downes
Joseph H. Hotchkiss
Marianne H. Gillette
Alina S. Szczesniak
Mark Barrett
Neil H. Mermelstein
Karen Banasiak
IFT Press Editorial Advisory Board
Malcolm C. Bourne
Fergus M. Clydesdale
Dietrich Knorr
Theodore P. Labuza
Thomas J. Montville
S. Suzanne Nielsen
Martin R. Okos
Michael W. Pariza
Barbara J. Petersen
David S. Reid
Sam Saguy
Herbert Stone
Kenneth R. Swartzel
Multivariate and
Probabilistic Analyses of
Sensory Science Problems
Jean-François Meullenet, Rui Xiong,
and Christopher J. Findlay
Jean-François Meullenet, Ph.D., is an Associate Professor in the Department of Food Science at the University
of Arkansas, Fayetteville, Arkansas. Dr. Meullenet conducts research in the area of sensory science and his
expertise encompasses sensory science, rheology, and modeling of food perception.
Rui Xiong,Ph.D., is a Research Scientist with Consumer Insights, Unilever Home & Personal Care, Trumbull,
Connecticut.
Christopher J. Findlay,Ph.D., is President of Compusense, Inc., Guelph, Ontario, Canada. He is the Associate
Editor for sensory evaluation for Food Research International.
©2007 Blackwell Publishing
All rights reserved
Blackwell Publishing Professional
2121 State Avenue, Ames, Iowa 50014, USA
Orders: 1-800-862-6657
Office: 1-515-292-0140
Fax: [phone]
Web site: www.blackwellprofessional.com
Blackwell Publishing Ltd
9600 Garsington Road, Oxford OX4 2DQ, UK
Tel.:[phone]
Blackwell Publishing Asia
550 Swanston Street, Carlton, Victoria 3053, Australia
Tel.:[phone]
Authorization to photocopy items for internal or personal use, or the internal or personal use of specific clients,
is granted by Blackwell Publishing, provided that the base fee is paid directly to the Copyright Clearance Center,
222 Rosewood Drive, Danvers, MA 01923. For those organizations that have been granted a photocopy license
by CCC, a separate system of payments has been arranged. The fee codes for users of the Transactional Report-
ing Service is ISBN-13: 978-0-8138-0178-0/2007.
First edition, 2007
Library of Congress Cataloging-in-Publication Data
Meullenet, J.-F. (Jean-François), 1968–
Multivariate and probabilistic analyses of sensory science problems/Jean-François Meullenet, Rui Xiong,
and Christopher J. Findlay.–1st ed.
p. cm.
Includes bibliographical references and index.
ISBN-13: 978-0-8138-0178-0 (alk. paper)
ISBN-10: 0-8138-0178-8
1. Food–Sensory evaluation–Statistical methods. 2. Multivariate analysis. I. Xiong, Rui.
II. Findlay, Christopher J. III. Title.
TX546.M48 2007
664′.07–dc22
2006036135
The last digit is the print number: 9 8 7 6 5 4 3 2 1
Titles in the IFT Press series
• Accelerating New Food Product Design and Development (Jacqueline H.P. Beckley,
Elizabeth J. Topp, M. Michele Foley, J.C. Huang and Witoon Prinyawiwatkul)
• Biofi lms in the Food Environment (Hans P. Blaschek, Hua Wang, and Meredith E.
Agle)
• Food Irradiation Research and Technology (Christopher H. Sommers and Xuetong
Fan)
• Food Risk and Crisis Communication(Anthony O. Flood and Christine M. Bruhn)
• Foodborne Pathogens in the Food Processing Environment: Sources, Detection and
Control(Sadhana Ravishankar and Vijay K. Juneja)
• High Pressure Processing of Foods (Christopher J. Doona, C. Patrick Dunne, and
Florence E. Feeherry)
• Hydrocolloids in Food Processing (Thomas R. Laaman)
• Microbiology and Technology of Fermented Foods(Robert W. Hutkins)
• Multivariate and Probabilistic Analyses of Sensory Science Problems (Jean-François
Meullenet, Rui Xiong, and Christopher Findlay)
• Nondestructive Testing of Food Quality (Joseph Irudayaraj and Christoph Reh)
• Nonthermal Processing Technologies for Food (Howard Q. Zhang, Gustavo V. Barbosa-
Canovas, V.M. Balasubramaniam, Editors; C. Patrick Dunne, Daniel F. Farkas, James
T.C. Yuan, Associate Editors)
• Nutraceuticals, Glycemic Health and Diabetes (Vijai K. Pasupuleti and James W.
Anderson)
• Packaging for Nonthermal Processing of Food(J.H. Han)
• Preharvest and Postharvest Food Safety: Contemporary Issues and Future Directions
(Ross C. Beier, Suresh D. Pillai, and Timothy D. Phillips, Editors; Richard L. Ziprin,
Associate Editor)
• Processing and Nutrition of Fats and Oils (Ernesto Hernandez, Monjur Hossen, and
Afaf Kamal-Eldin)
• Regulation of Functional Foods and Nutraceuticals: A Global Perspective (Clare M.
Hasler)
• Sensory and Consumer Research in Food Product Design and Development (Howard
R. Moskowitz, Jacqueline H. Beckley, and Anna V.A. Resurreccion)
• Thermal Processing of Foods: Control and Automation (K.P. Sandeep)
• Water Activity in Foods: Fundamentals and Applications(Gustavo V. Barbosa-Canovas,
Anthony J. Fontana Jr., Shelly J. Schmidt, and Theodore P. Labuza)
• Whey Processing, Functionality and Health Benefi ts (Charles Onwulata and Peter
Huth)
Table of Contents
Introduction, 3
Chapter 1. A Description of Sample Data Sets Used in Further Chapters, 9
1.1. A Description of Example Data Sets, 9
References, 25
Chapter 2. Panelist and Panel Performance: A Multivariate Experience, 27
2.1. The Multivariate Nature of Sensory Evaluation, 27
2.2. Univariate Approaches to Panelist Assessment, 29
2.3. Multivariate Techniques for Panelist Performance, 32
2.4. Panel Evaluation through Multivariate Techniques, 43
2.5. Conclusions, 46
References, 47
Chapter 3. A Nontechnical Description of Preference Mapping, 49
3.1. Introduction, 49
3.2. Internal Preference Mapping, 49
3.3. External Preference Mapping (PREFMAP), 58
3.4. Conclusions, 66
References, 67
Chapter 4. Deterministic Extensions to Preference Mapping Techniques, 69
4.1. Introduction, 69
4.2. Application and Models Available, 69
4.3. Conclusions, 89
References, 94
Chapter 5. Multidimensional Scaling and Unfolding and the Application
of Probabilistic Unfolding to Model Preference Data, 95
5.1. Introduction, 95
5.2. Multidimensional Scaling (MDS) and Unfolding, 96
5.3. Probabilistic Approach to Unfolding and Identifying the Drivers of Liking, 98
5.4. Examples, 100
References, 109
vii
viii Contents
Chapter 6. Consumer Segmentation Techniques, 111
6.1. Introduction, 111
6.2. Methods Available, 111
6.3. Segmentation Methods Using Hierarchical Cluster Analysis, 113
References, 126
Chapter 7. Ordinal Logistic Regression Models in Consumer Research, 129
7.1. Introduction, 129
7.2. Limitations of Ordinary Least Square Regression, 129
7.3. Odds, Odds Ratio, and Logit, 130
7.4. Binary Logistic Regression, 133
7.5. Ordinal Logistic Regression Models, 144
7.6. Proportional Odds Model (POM), 144
7.7. Conclusions, 160
References, 160
Chapter 8. Risk Assessment in Sensory and Consumer Science, 163
8.1. Introduction, 163
8.2. Concepts of Quantitative Risk Assessment, 164
8.3. A Case Study: Cheese Sticks Appetizers, 166
8.4. Conclusions, 176
References, 176
Chapter 9. Application of MARS to Preference Mapping, 179
9.1. Introduction, 179
9.2. MARS Basics, 179
9.3. Setting Control Parameters and Refining Models, 187
9.4. Example of Application of MARS, 188
9.5. A Comparison with PLS Regression, 201
References, 205
Chapter 10. Analysis of Just About Right Data, 207
10.1. Introduction, 207
10.2. Basics of Penalty Analysis, 208
10.3. Boot Strapping Penalty Analysis, 210
10.4. Use of MARS to Model JAR Data, 212
10.5. A Proportional Odds/Hazards Approach to Diagnostic Data Analysis, 215
10.6 Use of Dummy Variables to Model JAR Data, 220
References, 233
Index,237
Multivariate and Probabilistic Analyses of Sensory Science Problems
Jean-FrançoisMeullenet, Rui Xiong, Christopher J. Findlay
Copyright © 2007 by Blackwell Publishing
Multivariate and
Probabilistic Analyses of
Sensory Science Problems
Multivariate and Probabilistic Analyses of Sensory Science Problems
Jean-FrançoisMeullenet, Rui Xiong, Christopher J. Findlay
Copyright © 2007 by Blackwell Publishing
Introduction
Multivariate analysis of sensory and consumer science data is common practice today in
industry. This stems from the fact that with advances in computing and software develop-
ment these analyses can be performed with ease and also because the data to be analyzed
can be more complex today than in the early days of the sensory evaluation discipline.
This book was written with the sensory practitioner in mind. Many in industry who are
in sensory evaluation leadership roles deal with this data on a routine basis and have had,
in many cases, to educate themselves about these advanced techniques while on the job.
Our intent in writing this book was to provide nontechnical descriptions of multivariate
techniques that are commonly and less commonly used in sensory and consumer science.
The authors are not statisticians by training, which we hope helped in keeping this book
at a level most sensory professionals will find comfortable. We apologize to the statisti-
cians and sensometricians for not providing great details about statistical theories associ-
ated with these methods. However, in a few instances, we felt that some statistical details
were necessary, especially when the techniques described were not widely published. We
present techniques that answer specific sensory questions related to sensory and consumer
testing, but we have omitted any discussions about discrimination testing as multivariate
analyses are not typically used with this type of data. Table 1 summarizes the subjects
this text will deal with and the corresponding sections of the book that provide answers
to these questions.
Panelists and Panel Performance
Common answers provided by sensory tests dealing with descriptive analysis include
quantifying sensory differences among products. This is often the case in product devel-
opment when, for example, ingredients need to be substituted or the sensory profiles of
competing products need to be assessed. Aspects of multivariate analysis dealing with
the quantification of product differences are discussed in Chapter 2. Although descriptive
analysis panels are very useful to quantify product sensory differences, they need to reach
a certain level of performance to yield reliable data. Panelists and panel performance as
a whole are dictated by the panelists’ ability to discriminate among products when dif-
ferences exist and by their ability to reproduce their assessment of the product. Both of
these aspects of panel performance are important and dictate the outcome of any particular
test. Several multivariate techniques have been developed to assess panelists and panel
performance.
The use of MANOVA (multivariate analysis of variance) is discussed to evaluate
panelists’ ability to discriminate among products and panel homogeneity. To measure
panel homogeneity, the use of Principal Component Analysis (PCA) and Generalized
Procrustes Analysis (GPA) of individual attributes also is discussed. There are also
3