Issue: Volume 2, Number 8
Date:
August 2002
From:
Mark J. Anderson, Stat-Ease, Inc.

(Update 3/07: This link has been changed to http://www.fort.usgs.gov/research/rarmi/Proj3.asp.)

Here's what I cover in the body text of this DOE FAQ Alert (topics that delve into statistical detail are rated "X"):

1. FAQ: Does it matter how I label factors for a two-level DOE?
2. X-FAQ: How do I change ranges on mixture components?
3. Events alert: Stat-Ease is in New York next week
4. Workshop alert: "Experiment Design Made Easy" is going to San Jose later this month

PS. Quote for the month from Richard Feynman (plus a link to the site on this famous physicist)

1 - FAQ: Does it matter how I label factors for a two-level DOE?

-----Original Question-----
From: Colombia
"When I enter the factors for a 5-factor half-fractional factorial design [2^(5-1)], does the order affect the results? Which factors should I put first? What if I have to block because I will conduct the experiment in two days? How should the order of the factors should be changed?"

On this design, if not blocked, the order you enter in the factors will not matter. However, when you impose blocking, this changes the picture - see the alias structure below (copied from Design-Expert® software's default design choice for a 2^(5-1) with two blocks).

>Design Matrix Evaluation for Factorial Reduced 3FI Model
>Factorial Effects Aliases
>[Est. Terms] Aliased Terms
>[Intercept] = Intercept
>[Block 1] = Block 1 - ABD - CDE
>[Block 2] = Block 2 + ABD + CDE
>[A] = A + BCE
>[B] = B + ACE
>[C] = C + ABE
>[D] = D
>[E] = E + ABC
>[AB] = AB + CE
>[AC] = AC + BE
>[AE] = AE + BC
>[BD] = BD
>[CD] = CD
>[DE] = DE
>[ACD] = ACD + BDE

Now some of the two-factor interactions (2fi) become aliased with each other. Do you think certain factors might interact? If so, name them in a way that provides maximum resolution of the 2fi. In this design, I suggest naming the potentially most powerful factor (in terms of interacting with other factors) as "D" because none of the 2fi's involving this letter will be aliased with any other 2fi. Also, D stands alone as a main effect. (It's actually aliased with the 5fi, which we don't bother reporting during the design building phase since it's so unlikely.)

2 - X-FAQ: How do I change ranges on mixture components?

-----Original Question-----

From:
Mexico
"I ran a mixture experiment on 3 components. Initially variable B had a range from 75 to 120 grams in the mixture. The results were very good so we made 5 new formulations around the optimal zone; with a lower value: A=60 grams. We added the results to the design layout but the software did not accept them because they were out of the initial range. How can we change the range of a variable, after you have results, so we can add new data, outside of the initial range?"

Answer (from Pat Whitcomb, consultant and principal of Stat-Ease):
"You are correct - the only way to change the ranges in a mixture design is to rebuild it. Here is a reasonable procedure to follow:

1. Open the original design file with Design-Expert software.

3. Enter the new component low and high values. Here are your options:

- Making ranges broader is OK. (Hint: I would not extend the levels higher or lower than are necessary to include the new runs. The pseudo values are based on these highs and lows and collinearity problems are minimized when the actual lows and highs are used.)

- Making ranges narrower will exclude some of the existing design points from the new design space.

- The total must remain the same.

4. Build your new design. (Hint: Enter the number of runs you want in the new design.)

5. Run a second copy of Design-Expert software and open the original design again.

6. Copy the component and response columns from the original design. (Hint: Click on the first component in the first run and drag it to the last response in the last run; and copy.)

7. Paste the original design runs into the rebuilt design. (Hint: Both designs should be sorted in run order and displaying actual values.)

8. Right click on the Block column header and choose the Edit Info option. Change the number of blocks from 1 to 2.

9. For the new runs, change to block 2 and type in the new component levels and response values.

[Learn more about developing good mixture designs by attending the "Mixture Design for Optimal Formulations" workshop (see http://www.statease.com/clas_mix.html ).]

3 - Events alert: Stat-Ease is in New York next week

Stat-Ease is going to New York City to exhibit at ASA's (American Statistical Society) Joint Statistical Meeting (JSM) on August 11-15. See http://www.amstat.org/meetings/jsm/2002/ for details from ASA. Stat-Ease consultant, Shari Kraber, will host a roundtable luncheon at the JSM. The topic is "How to Avoid DOE Disaster." A summary article on the topic will appear in our September "Stat-Teaser" newsletter.

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

4 - Workshop alert: "Experiment Design Made Easy" is going to San Jose later this month

Consider attending the "Experiment Design Made Easy" workshop on August 20-22 in San Jose, California. For schedule and site information on this and all Stat-Ease workshops open to the public, see http://www.statease.com/clas_pub.html. To enroll, click the "register online" link at 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.

Mark@StatEase.com

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

PS. The quote for the month is from Richard Feynman (plus a link to the site on this famous physicist).

"The test of all knowledge is experiment."
-Richard Feynman (see http://www.crunchygods.com/weirdo/feynman/ for more about Feynman.)

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 (see above)