Response Surface Methodology: Process and Product Optimization Using Designed Experiments – An interview with co-author C. Anderson-Cook

Featuring a substantial revision, Wiley was proud to publish earlier this year the fourth edition of Response Surface Methodology: Process and Product Optimization Using Designed Experiments presents updated coverage on the underlying theory and applications of response surface methodology (RSM). Providing the assumptions and conditions necessary to successfully apply RSM in modern applications, the new edition covers classical and modern response surface designs in order to present a clear connection between the designs and analyses in RSM.

With multiple revised sections with new topics and expanded coverage, Response Surface Methodology: Process and Product Optimization Using Designed Experiments, Fourth Edition includes:

  • Many updates on topics such as optimal designs, optimization techniques, robust parameter design, methods for design evaluation, computer-generated designs, multiple response optimization, and non-normal responses
  • Additional coverage on topics such as experiments with computer models, definitive screening designs, and data measured with error
  • Expanded integration of examples and experiments, which present up-to-date software applications, such as JMP®, SAS, and Design-Expert®, throughout
  • An extensive references section to help readers stay up-to-date with leading research in the field of RSM

An ideal textbook for upper-undergraduate and graduate-level courses in statistics, engineering, and chemical/physical sciences, Response Surface Methodology: Process and Product Optimization Using Designed Experiments, Fourth Edition is also a useful reference for applied statisticians and engineers in disciplines such as quality, process, and chemistry.

Alison Oliver talks to co-author Christine Anderson-Cook about the writing process and her own career.

 

1. Congratulations on the publication of the fourth edition of Response Surface Methodology: Process and Product Optimization Using Designed Experiments which provides the assumptions and conditions methods and tools necessary to successfully apply RSM in modern applications, and covers classical and modern response surface designs in order to present a clear connection between the designs and analyses in RSM. How did the writing process begin for the first edition?

Even before the first edition of the book, the lead author, Raymond Myers, had written his “little green book” which was self-published and outlined the fundamentals of a brand new area in statistics. Ray is truly one of the founding fathers of Response Surface Methodology, and the material in all of the previous editions, as well as this current edition builds on the fundamentals that were presented 40 years ago. The official first edition of the book appeared in 1996, with revisions in 2002 and 2009, before the current edition came out this year.

I was honored to be invited to join co-authors Raymond Myers and Douglas Montgomery for the 3rd and 4th editions. The book is very practical and gives engineers and scientists tools for collecting data, identifying important relationships between inputs and outputs, and using those relationships to optimize the performance of their product or process. The popularity of RSM methods can be seen with the rise in the number of citations of the book over the years, which now totals over 11000.

2. What were the primary objectives you had in mind during the writing process?

The goal of the book is to provide practitioners with strategies and tools for solving important problems. The area of RSM is growing increasingly popular as we are living in a golden age of design of experiments. People in many disciplines are realizing the importance of having high quality data to answer their important questions, and the ability to create a customized design that matches the needs of your experiment is become more and more straightforward. The book walks through a process of collecting and analysing the right day for a variety of different classes of problems.

3. For those unfamiliar with the original editions, please could you let us know what the reader can expect from this new edition?

The heart of the book has remained the same between different editions: It focuses on the sequential collection and analysis of data that begins with a screening experiment to assess which input factors are most influential in changing the response (or responses). Once some of the less influential factors have been eliminated, the experimenter can identify the best ranges of the input variables to study that hones in on the most desirable values of the response. In the next stage, typically a low-order polynomial is used to characterize the relationship between important input factors and the response with a response surface model. This estimated model can then be used to optimize by identifying best combinations of the input factors that yield ideal values of the response(s). Choosing the right design to match the intended analysis at each stage of the process can accelerate learning about the process and use valuable experimentation resources efficiently.

4. The edition also includes many updates on topics such as optimal designs, optimization techniques, robust parameter design, methods for design evaluation, computer-generated designs, multiple response optimization, and non-normal responses. Please could you give us a sneak preview of one of the updates provided?

The research area of response surface methodology is evolving quickly with many new advances in the last few years. The latest edition incorporates a lot of the ground-breaking work that allows greater flexibility and tailored solutions in RSM. It is now possible to create a design and analysis strategy that is an ideal match for the goals of a study, and to understand the strengths and weaknesses of a design before the experiment is actually run. This can help avoid costly mistakes and build in provisions for handling some of the unexpected things that can happen with data collection.

An emerging area of importance is how to handle multiple objectives or multiple responses. The multiple objectives might come into play if we are planning a designed experiment, but we are not sure of what the final model will look like. Choosing a design that performs well for the experimenters best guess of the model, but also can accommodate a slightly different model can be a better choice than focusing exclusively on a single model as being correct. For multiple responses, we often collect data on several responses that might be of different levels of importance to us. Getting excellent performance on a primary response could be complemented with small tradeoffs to also get improved performance on several secondary responses.

5. If there is one piece of information or advice that you would want your reader to take away and remember after reading your book, what would that be?

I think there are a couple of key messages in the book:

(a) It can be very beneficial to think about the relationship between several inputs and several responses in outputs in functional form or as a surface. This visualization can lead to deeper understanding of those relationships and provide a framework for optimization.
(b) Choosing the right designed experiment should be based on what you believe the response surface connecting inputs and outputs looks like. Simpler relationships can be explored with simpler designs, but more complicated relationships often need more levels of each factor to understand the patterns.
(c) Uncertainty plays a role in all aspects of screening, characterization and optimization. It is important to summarize response surfaces with their associated uncertainty intervals to understand what new observations are likely in future data collection.

6. Who should read the book and why?

The book is well suited for engineers and scientists who will be collecting data to answer specific questions about their products or processes. It also is an ideal textbook for graduate students in statistics and engineering, as having the right strategy for data collection and analysis can help with learning and understanding of processes.

7. Why is this book of particular interest now?

As I mentioned earlier, I think we are living in a golden age of experimentation – it has become easier to collect data with experiments in recent years, and business has realized the importance of disciplined approaches for understanding and modeling their processes and products. The competitive advantage of being efficient with data collection and its analysis can be huge.

8. Were there areas of the book that you found more challenging to write, and if so, why?

Staying current with all of the new developments has been an ongoing challenge, but it is also exciting to incorporate these advances into the book so that the readers can benefit from methods and approaches that are on the leading edge of RSM.

9. What is it about the area of experimental design that fascinates you?

In my regular job, I design experiments for researchers in a wide variety of disciplines. It feels very powerful to be able to take a set of statistical tools, like RSM, and apply them to solve many different problems. It feels very rewarding to help a researcher be efficient with their data collection, and then it is great to help facilitate those “aha” moments where suddenly complex, previously unknown relationships can be effectively characterized and understood.

10. What will be your next book-length undertaking?

I am not sure that I will be tackling another book for a while, but I have just recently completed another projects with Dr. Lu Lu. We co-edited a book, Statistical Roundtables: Insights and Best Practices that is a compilation of the American Society for Quality columns that appear in Quality Progress. Those columns are written by some leading experts in quality and statistics, and we organized the best of 16 years of columns into topics areas to make extracting the pearls of wisdom more accessible.

11. Please could you tell us about your educational background and what inspired you to pursue your career in your respective disciplines?

When I was growing up, I would always tell people that I wanted to be a mathematician. I am not sure what a 10 year old thinks a mathematician does, but I have always loved mathematics and solving problems. When I was an undergraduate at the University of Waterloo in Canada, I realized that solving practical problems was something that I was also really drawn to – so statistics can be the perfect blend of solving high impact problems using mathematical tools. I initially thought that I wanted to be a teacher, so I have a Bachelor of Education in addition to my Bachelor of Mathematics, Masters and Ph.D. in statistics. I was a faculty member in the Department of Statistics at Virginia Tech for 8 years before moving to New Mexico in 2004. Currently in my work at Los Alamos National Laboratory, I lead projects in the modelling of reliability of complex systems, detection and classification of malware, and non-proliferation.