Here's another set of frequently asked questions (FAQs) about doing design of experiments (DOE), plus alerts to timely information and free software updates. If you missed the previous DOE FAQ Alert, please click on the links at the bottom of this page. If you have a question that needs answering, click the Search tab and enter the key words. This finds not only answers from previous Alerts, but also other documents posted to the Stat-Ease web site.
Feel free to forward this newsletter to your colleagues. They can subscribe by going to http://www.statease.com/doealertreg.html. If this newsletter prompts you to ask your own questions about DOE, please address them via mail to:StatHelp@StatEase.com.
Here's an appetizer to get this Alert off to a good start: Maybe you should not be subscribing to this newsletter because according to an Internet posting by BBC News UK Edition, e-mail is worse than Marijuana for draining your brain (see http://news.bbc.co.uk/1/hi/uk/4471607.stm). The article reports a study showing the distraction of e-mail causing a 10-point fall in subjects' IQmore than twice that found in studies of the impact of smoking marijuana. I'd like to think that DOE FAQ Alert, being so thought-provoking, causes a net increase in IQ. What do you think (or can you even think while reading this)?
Here's what I cover in the body text of this DOE FAQ Alert (topics that delve into statistical detail are designated "Expert"):
1. FAQ: Setting
up a fractional design with multilevel factors, some of
which are categoric
"I am trying to help our process engineers with a DOE. They want to test four factors in the experimenttwo categorical and two numeric. One of the numeric factors will be tested at three levels, and everything else is at two levels, so the total number of combinations will be 24 (3 x 2 x 2 x 2). Do you know of a good design that would allow me to meet these requirements and still be around 15 to 18 runs? I wish it was as simple as a standard 2^(4-1) half-fractional two-level factorial. The one factor (numeric) at three levels is causing the problem. Any advice?"
(Learn more about D-optimal factorial design by attending the three-day computer-intensive workshop "Experiment Design Made Easy." See http://www.statease.com/clas_edme.html for a course description. Link from this page to the course outline and schedule. Then, if you like, enroll online.)
2. Expert-FAQ: How the prediction interval for a modeled response differs from its confidence interval
"I've been looking at the point prediction tool in Design-Expert software. (I think) I understand that the
standard error (SE) mean is calculated as the product of the SE-design and the standard deviation (Stdev) of the design. I am reasonably happy with what this means and how it is calculated. What I'm struggling to understand is the SEprediction. This seems to be the product of the standard deviation and the square root of one plus the standard error design squares (1+SEdesign^2). Is this correct? Where does the 1 come from in this term? I realize that what this calculation is trying to do is to take into account a variance component for the model. Why use 1 rather than the SE values for each of the components selected for the model? Any light you can shed on my confusion would be appreciated."
Answer (from Stat-Ease Consultant
The confidence interval is expected to contain the mean, or "true," value. The prediction interval is constructed to include a single observation. It incorporates uncertainty as to the location of true value, as well as additional uncertainty associated with any single observation. The 1 added to SEdesign in the formula accounts for this extra variation.
If you want more detail, it can be found in most regression analysis text books. However, this is rarely mentioned in DOE textbooks."
Aside from the mathematical details, I would like to add
that the prediction interval can be very helpful for assessing
individual confirmation runs on what is hoped to be optimal
conditions. Naturally results will vary. The prediction
interval provides an expectation on the amount of variation.
(Learn more about advanced tools of DOE by attending the three-day computer-intensive workshop "Response Surface Methods for Process Optimization." See http://www.statease.com/clas_rsm.html for a complete description. Link from this page to the course outline and schedule. Then, if you like, enroll online.)
This news is bit dated, but I just became aware of it thanks to Cliff Yee, President of Northwest Analytical (NWA) in Portland, Oregon.* He said, "Good to hear your witty and interesting newsletters. How are you guys doing in response to PAT Guidelines by the FDA? I am wondering if you are seeing more investment by the Life Science industries in DOE consulting, training and software?" After tracking down the PAT Guidelines at http://www.fda.gov/cder/guidance/6419fnl.doc (Update 3/07: Link no longer available.) (also see http://www.fda.gov/cder/OPS/PAT.htm), I understand why Cliff believes it will generate interest in DOE. The document greatly encourages the use of planned statistical methods to explore how variations in component levels and process conditions will affect pharmaceuticals. If any of you readers can speak to this, please e-mail me.
*PS. NWA just announced a new version of their Quality Analyst software for statistical process control, etc. See http://www.nwasoft.com for details.
4. Book Alert: There is a new edition of "Statistics for Experimenters" by Box, Hunter and Hunter
According to the listing at Amazon (see below), on May 27 Wiley-Interscience published "Statistics for Experimenters: Design, Innovation and Discovery" the second edition of the classic book on DOE by George E. P. Box, J. Stuart Hunter, and the late William G. Hunter. Stat-Ease has a copy on order. According to the publisher, the "Second Edition is thoroughly revised and updated to reflect the changes in techniques and technologies since the publication of the classic First Edition. Among the new topics included are:
- Graphical analysis of variance
What intrigues me most, is the promise of "An appendix featuring quaquaversal quotes from a variety of sources ranging from noted statisticians and scientists to famous philosophers that embellish key concepts and enliven the learning process." I had to look up the word "quaquaversal." (It blew away my spell-checker!) I will not spoil your fun trying to decipher what it means.
For more details on this new edition from Box, Hunter and Hunter, click this link for the listing at Amazon: http://makeashorterlink.com/?K1F42272B.
5. Events alert: Link to a schedule of appearances by Stat-Ease
Click on http://www.statease.com/events.html for a list of appearances by Stat-Ease professionals. We hope to see you sometime in the near future!
6. Workshop alert: See when and where to learn about DOE
See http://www.statease.com/clas_pub.html for schedule and site information on all Stat-Ease workshops open to the public. To enroll, click the "register online" link on our web site or call Stat-Ease at 1.612.378.9449. If spots remain available, bring along several colleagues and take advantage of quantity discounts in tuition, or consider bringing in an expert from Stat-Ease to teach a private class at your site. Call us to get a quote.
I hope you learned something from this issue. Address your general questions and comments to me at: Mark@StatEase.com.
Mark J. Anderson, PE, CQE
PS. Quote for the month: Uplifting advice from the author of "Silent Spring"—a landmark book on the environment:
"Those who dwell among the beauties and mysteries of the Earth
are never alone or weary of life."
Trademarks: Design-Ease, Design-Expert and Stat-Ease are registered trademarks of Stat-Ease, Inc.
Acknowledgements to contributors:
in previous FAQ DOE Alert e-mail newsletters?
Click here to add your name to the FAQ
DOE Alert newsletter list server.
2021 E. Hennepin Avenue, Ste 480
Minneapolis, MN 55413-2726
p: 612.378.9449, f: 612.378.2152