Issue: Volume 3, Number 5
Date: May 2003
From: Mark J. Anderson, Stat-Ease, Inc. (www.statease.com)

Dear Experimenter,

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 previous DOE FAQ Alerts, please click on the links at the bottom of this page. 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 ask to your own questions about DOE, please address them to [email protected].

Here's some appetizers to get this Alert off to a good start.

For a sneak preview of the future for computer (PC) interfaces,see http://gtresearchnews.gatech.edu/newsrelease/infocanvas.htm. It shows "InfoCanvas" - a program developed by researchers at the Georgia Institute of Technology that creates an artistic representation of information. You might be surprised at what can be packed into a eye-pleasing beach scene!

On the hardware front, University of Toronto researchers offer a new tool called "ShapeTape" for inputting oddball curves. See http://www.newsandevents.utoronto.ca/bin4/030415d.asp for a picture and some details on this device which looks a lot like what in pre-PC days we engineers called a "snake". Back then we just twisted it into the shape we wanted and drew along the edge on to graph paper. How did we ever get by with such crude technology?

Oh, and one more thing - a literal "heads-up": North American readers should watch (if skies clear!) for a total lunar eclipse on Thursday, May 15 starting at 8 PM Eastern Daylight Time. See http://www.skynewsmagazine.com/pages/lunareclipse.html (back issues not available) for details.

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: Run order versus standard order: Why randomize?
2. X-FAQ: Impossibly negative factor settings generated by a response surface methods (RSM) design
3. Info alert: A case study in the April issue of "Powder Coating" magazine documents an innovative DOE resulting in significant defect reduction (links provided)
4. Sim alert: Link to a really cool applet showing graphically how simplex optimization works
5. Events alert: Stat-Ease is giving a talk and exhibiting at the Annual Quality Conference in Kansas City
6. Workshop alert: "DOE Simplified," a 1-day overview, is coming up later this month in Minneapolis, then the 3-day workshop, "Experiment Design Made Easy," follows in June, as well as several other classes on statistics

PS. Quote for the month—regarding statisticians being more like pathologists than physicians when asked to help with experiments after they've been done.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

1—FAQ: Run order versus standard order: Why randomize?

-----Original Question-----
From: Michigan

"When setting up a DOE using Stat-Ease software, I was not able to sort the data in the order below.

Run A B C
1 - - -
2 + - -
3 - + -
4 + + -
5 - - +
6 + - +
7 - + +
8 + + +

The program puts the runs in random order by default. How do I do over-ride this?"

Answer:
In my opinion, you really should run your DOE in random order, but to answer your question:

1. Right-click the Run column header and select "Sort by Standard Order".
2. Again right-click the Run column header, but this time "Re-order as currently displayed".

Years ago at a conference on DOE, I mentioned offhand that naturally I had randomized the run order of an experiment, even though my talk showed it in the standard order (such as that shown above). I was shocked when an argument broke out about the need for randomization. I've always taken this as a given because otherwise time-related lurking factors will bias the results. In my defense, I defer to George Box's article "Must We Randomize Our Experiment?" published in Quality Engineering, 1990, Vol. 2, No. 4, pp. 497-502. The basis for this article, done in December 1989, is posted as the 47th download at http://www.engr.wisc.edu/centers/cqpi/reports.html. You may want to browse other reports at this site maintained by the University of Wisconsin's Center for Quality and Productivity Improvement (CQPI), founded in 1985 by Professor Box and the late Professor William Hunter.

(Learn more about proper setup of factorial designs by attending the 3-day computer-intensive workshop "Experiment Design Made Easy." See http://www.statease.com/clasedme.html for a complete description. Link from this page to the course outline and schedule. Then, if you like, enroll online.)

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

2—X-FAQ: Impossibly negative factor settings generated by a response surface methods (RSM) design

-----Original Question-----
From: Malaysia

"I am trying to develop an optimization process for my research using RSM with five factors. But somehow the results showed negative values for certain parameters. I am confused with this. Could you give me the actual value based on these parameters:

A. Mole ratio: 1:1 to 1:10
B. Enzyme: 2% to 10%
C. Water content: 2% to 10%
D. Temperature: 40C to 75C
E. Time: 6h to 48h"

Answer:
My guess is that you selected the top-listed RSM design, a central composite (CCD), which puts points (typically drawn with stars in textbooks on the topic) outside of your specified plus/minus 1 coded unit limits at distance alpha to provide better leverage for the modeling. If so, for the full CCD (the default design with 50 runs), Design-Expert® software puts the star points at an alpha distance of 2.37841 (for the statistical property of rotatability), which causes the factors to range as follows:

A. -5.20286 to 16.2029
B. -3.51366 to 15.5137
C. -3.51366 to 15.5137
D. 15.8778 to 99.1222
E.-22.9467 to 76.9467

In this case you are asked to do impossible things such as putting in negative percents of various materials and doing it in negative time (company managers like this—get it done yesterday!).

Here are some feasible design options:

1. Choose the Central Composite, but click the "Factor ranges in terms of alpha" option and then re-enter your original low and high levels (noted in your question). Consider the following enhancements to this CCD: Change "Type" to "1/2 Fraction" which reduces the runs to 32 and shortens the alpha to 2. The disadvantage to this design is that the well-modeled area, defined by coded ranges of -1 to +1, shrinks to a small fraction of the experimental region, presumably of interest. If this is bothersome, try option 2 or 3 noted below.

2. Choose the Central Composite, but click the Options button and press "Face Centered" for an alpha value of 1. Then enter your original low and high levels. Again, consider changing "Type" to "1/2 Fraction." Now the star points do not go outside of your original limits.

3. Choose the Box-Behnken design, which is restricted to 3 levels. For 5 factors you get 46 runs. Enter your original low and high levels - the software will fill in the center value.

(Learn more about RSM designs by attending the "Response Surface Methods for Process Optimization" workshop. For a description, see http://www.statease.com/clas_rsm.html. Link from this page to the course outline and schedule. You can enroll online by linking to the Stat-Ease e-commerce page for workshops.)

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

3—Info alert: A case study in the April issue of "Powder Coating" magazine documents an innovative DOE resulting in significant defect reduction (links are provided)

Click http://www.statease.com/pubs/morton.pdf to see how Julia O'Neill and her team at Morton Powder coatings designed and analyzed a split-plot, two-level factorial experiment aimed at trouble-shooting a client problem with a product defect. The solution involved a combination of process changes and formulation variables. The published article can be downloaded for US $7.50 from http://www.pcoating.com/ag_20030401.636111.htm (Update 3/07: this link has changed to http://www.pcoating.com/getarticle.cfm/articleid=1269).

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

4—Sim alert: Link to a really cool applet showing graphically how simplex optimization works

We make use of a variable-sized simplex algorithm in Design-Expert software's numerical optimization tools. It's a robust way to find maximum desirability for multiple responses. To see how this algorithm works and play around with it graphically, visit http://www.chem.uoa.gr/Applets/AppletSimplex/Appl_Simplex2.html.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

5—Events alert: Stat-Ease is giving a talk and exhibiting at the Annual Quality Conference in Kansas City

I hope to see many of you readers at the upcoming American Society of Quality's (ASQ) 57th Annual Quality Congress (AQC) in Kansas City. (See http://aqc.asq.org/ for details on how to register for the exhibit and conference.) Please stop by our booth (#115) to talk with me and other representatives of Stat-Ease. We'd like to see you! I will be presenting T310, "How to Use Graphs to Diagnose and Deal with Bad Experimental Data", on Tuesday May 20 from 2:00 PM - 03:15 PM. The material, co-authored by fellow Stat-Ease consultant and principal Pat Whitcomb, is rated "Basic." See http://www.statease.com/pubs/andersonaqc.pdf http://www.statease.com/pubs/baddata.pdf for content.

May will be a big month for conferences providing technical talks on DOE. Here are two more in addition to AQC:

—Conference on New Directions in Experimental Design to be held in Chicago, Illinois on May 14-17, 2003. The focus of the conference will be on design of experiments in the pharmaceutical and related industries. See http://www.math.uic.edu/~kjryan/dae2003.html (Update 3/07: link no longer available) for more details.

—The 2003 Quality & Productivity Research Conference will be held 21-23 May 2003 in the IBM Thomas J. Watson Research Center, Yorktown Heights, NY. It's jointly sponsored by IBM and Rutgers University. The theme is "Data + Statistics = Competitive Advantage." Prof. John A.Cornell will present a full-day pre-conference tutorial "Experimenting with Mixtures" on May 20. Two special events are planned: a dinner party honoring our special guest (Prof. P.W. John) and student's awards (May 21), and a visit to the IBM Industry Solutions Lab (May 22). The conference program and related materials are available on the web site: http://www.research.ibm.com/stat/qprc.

Click on http://www.statease.com/events.html for a list of where Stat-Ease consultants will be giving talks and doing DOE demos. We hope to see you sometime in the near future!

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

6—Workshop alert: "DOE Simplified," a 1-day overview, is coming up later this month in Minneapolis, then the 3-day workshop, "Experiment Design Made Easy," follows in June, as well as several other classes on statistics

We offer a variety of learning opportunities over the next few months, all at the Stat-Ease training facility:

—Our "DOE Simplified" 1-day presentation on May 29, 2003 offers the perfect introduction to DOE for everyone from technicians to engineers to managers. It provides an overview of DOE concepts and illustrates an array of tools and design types. Although this workshop is not a substitute for hands-on computer-intensive training, it provides an educational starting point. See http://www.statease.com/does.html for details.

—To gain a working knowledge of DOE, attend the computer-intensive Experiment Design Made Easy at the Stat-Ease training facility in Minneapolis on June 10-12. For course content, see http://www.statease.com/clasedme.html.

—On June 24 and 25 we present "Statistics for Technical Professionals". This Excel-intensive workshop offers many cool tools for quality. See http://www.statease.com/clas_stp.html for details.

Also, it's not too late to sign up for a presentation of "Mixture Design for Optimal Formulations" May 13 through 15. It's a great time to come to Minneapolis and get educated on statistics, too! Information on this workshop is posted at: http://www.statease.com/clas_mix.html.

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:

[email protected]

Sincerely,

Mark

Mark J. Anderson, PE, CQE
Principal, Stat-Ease, Inc. (http://www.statease.com)
Minneapolis, Minnesota USA

PS. Quote for the month—regarding statisticians being more like pathologists than physicians when asked to help with experiments after they've been done:

"To consult a statistician after an experiment is finished is often merely to ask him to conduct a post mortem examination. He can perhaps say what the experiment died of."
- R.A. Fisher, Presidential Address to the First Indian Statistical Congress, 1938

Trademarks: Design-Ease, Design-Expert and Stat-Ease are registered trademarks of Stat-Ease, Inc.

Acknowledgements to contributors:

—Students of Stat-Ease training and users of Stat-Ease software
—Fellow Stat-Ease consultants Pat Whitcomb and Shari Kraber (see http://www.statease.com/consult.html for resumes)
—Statistical advisor to Stat-Ease: Dr. Gary Oehlert (http://www.statease.com/garyoehl.html)
—Stat-Ease programmers, especially Tryg Helseth (http://www.statease.com/pgmstaff.html)
—Heidi Hansel, Stat-Ease marketing director, and all the remaining staff

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Interested in previous FAQ DOE Alert e-mail newsletters?
To view a past issue, choose it below.

#1 Mar 01
, #2 Apr 01, #3 May 01, #4 Jun 01, #5 Jul 01 , #6 Aug 01, #7 Sep 01, #8 Oct 01, #9 Nov 01, #10 Dec 01, #2-1 Jan 02, #2-2 Feb 02, #2-3 Mar 02, #2-4 Apr 02, #2-5 May 02, #2-6 Jun 02, #2-7 Jul 02, #2-8 Aug 02, #2-9 Sep 02, #2-10 Oct 02, #2-11 Nov 02, #2-12 Dec 02, #3-1 Jan 03, #3-2 Feb 03, #3-3 Mar 03, #3-4 Apr 03, #3-5 May 03 (see above)

Click here to add your name to the FAQ DOE Alert newsletter list server.

Statistics Made Easy™

DOE FAQ Alert ©2003 Stat-Ease, Inc.
All rights reserved.

 


Software      Training      Consulting      Publications      Order Online      Contact Us       Search

Stat-Ease, Inc.
2021 E. Hennepin Avenue, Ste 480
Minneapolis, MN 55413-2726
e-mail: [email protected]
p: 612.378.9449, f: 612.378.2152