"An excellent book for any practicing engineers who are interested
in using Taguchi's approach in optimizing and characterizing a
process. Dr. Roy's step-by-step method outlined in the book does
not require the readers to have any prior statistical knowledge to
understand Taguchi's approach to Design of Experiments. The book
also walks you through the computer program on the CD-ROM that
comes with it." - Marcus Khoo, Quality Manager, Agilent
Technologies, Inc.
"Per student evaluations (including my own!), Ranjit Roy is one
of the best instructors that anyone is likely to meet. This is
especially remarkable considering that Dr. Roy specializes in a
very difficult subject: making Design of Experiments (DoE) easy to
understand and accessible to everyone. His software, Qualitek 4,
based upon his experience in doing real experiments and teaching
other engineers to do experiments, established the benchmark in how
to make this type of software useful and user friendly. While
traditional classes in designed experiments spend considerable time
teaching engineers how to do the mathematical analysis of these
studies (yawn), Ranjit likes to teach engineers how to do
experiments using software that automatically designs the
experiment, analyzes the data, and prints reports (with graphics)
that can immediately be shared in management or team meetings. Now
Ranjit has taken about fifteen years of experience in teaching
these types of classes, and has made it available to everyone in a
recent book published by John Wiley & Sons, Wiley-Interscience
Professional/Trade Division. This book is written for people who
want to learn a method for designing and analyzing simple
experiments with the goal of helping the reader apply these
techniques in their own work environment. I believe this book will
be particularly helpful to engineers (Black Belts or Green Belts)
trying to learn the technique of designed experiments as taught in
the Six Sigma process. This book is a "must have" for anyone who
currently uses, or who wants to learn to use, Ranjit's software.
Others will want to purchase this book for the very plentiful and
helpful case studies taken from real life experiences. Dr. Roy
provides resources for three key elements of the learning process:
brief discussions of applicable theory, step-by-step application of
the concepts, and fully functional software to use with the
stories/examples in the book. Explanations of theory are simple and
elegant. For example, in explaining to the reader the concept of
"Degrees Of Freedom", Ranjit writes, "Degrees Of Freedom (DOF) is
an indication of the amount of information contained in a data set.
For example, if there are three numbers in a data set, the DOF=3-1,
which indicates the amount of additional information that can be
derived by taking the differences among the three numbers. Looked
at in another way, if there are three people, you will need two
comparisons to determine who is tallest." No wonder engineers like
to take Dr. Roy's classes! Dr. Roy provides sixteen steps to learn
designed experiments and apply the methods to any problem. These
steps include topics such as: Dr. Taguchi's philosophy, the
definition of quality and how to measure it, the logic behind
experimental design analysis methods, the use of orthogonal arrays,
two-level experiments, three-level experiments, Analysis of
Variance (ANOVA), interactions, mixed level factors, combination
designs, strategies for robust design, signal-to-noise ratios,
experiments with multiple criteria, how to quantify improvement,
experiment preparation and planning, and even more case studies (a
special chapter of added examples in addition to the plentiful
examples provided in each of the above mentioned topics). Each step
is a chapter, and each chapter: starts with a learning objective,
provides learning by going through examples using software, and
then ends with: a "question and answer" dialogue based upon
Ranjit's teaching experience, a chapter summary, additional
exercises, and software solutions for these exercises. Readers will
especially appreciate material in the book that is original with
Ranjit himself: the development of overall evaluation criteria for
studies with multiple responses, and the use of a "severity index"
to quantify the strength of an interaction. The real heart of this
book is the combination of software with examples. The book takes
the reader through the software frame by frame, showing you where
to click, what the output looks like, and how to interpret this
output. The software provided with the book is fully functional
with the data from the examples used in the book. However, when it
comes time for the reader to do their own experiments, they are
only allowed to use the portion of the software that deals with
two-level experiments in an L8 orthogonal array (unless, of course,
you purchase the complete software package). The software requires
Windows 3.1 or higher with 32 MB RAM, 10 MB of hard drive space, a
CD-ROM drive, and Microsoft PowerPoint and Word (to view additional
documents). Ranjit makes it clear that this book is intended for
beginners who want to get started. He only covers Taguchi's static
robustness studies (even though the software is capable of dynamic
studies), and tells the reader that, "As you gain application
experience, you will find it necessary to refer to other books to
develop clearer concepts of the mathematical treatments." Readers
check assumptions in their analysis by doing a confirmation
experiment to see if the result of setting factors at the perceived
best levels falls within a confidence interval provided by the
software. No book is perfect. Critics may say that some of the
definitions are weak (for example, "system design" is defined as
the "design of a product or process using special Taguchi
techniques"), that some areas are too simplified (not enough
emphasis on assumptions used in the analysis, no explanation of the
distinction between data replication and repetition, no mention
that the Cpk analysis is only applicable if the process is stable),
and that in certain situations the software is weak (Dr. Roy warns
the reader that the software will not correctly calculate
confidence intervals if the Degrees Of Freedom are less than three,
and that in some situations the software will only allow you to use
an 80% confidence interval). No doubt, the biggest criticism of the
book is that the reader will learn how to use many features of the
software that they will not be able to later apply unless they
purchase the full software package. Nevertheless, a beginner can
learn designed experiments with this book, and can make a great
deal of progress using L8 arrays (in fact, the technique of
designed experiments is so powerful that readers may be able to
purchase software for their entire company based upon the savings
associated with the application of one good L8 experiment!). So, if
you have ever thought about doing experiments but have never quite
felt comfortable enough to get started ? get this book and get
going. We have Ranjit to thank for taking the fear out of the
process and providing engineers with simple explanations, examples,
and software to quickly build the self-confidence needed to get
real results." -Larry R. Smith, Quality and Reliability Manager,
Ford Motor Company
