|
|
|
Issue:
Volume 5, Number 10
Date:
October 2005
From:
Mark J. Anderson, Stat-Ease,
Inc. (http://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 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:
http://www.grand-illusions.com/simulator/montysim.htm. I first*
saw this statistical puzzle detailed in "The curious incident of
the dog in the night time" by Mark Haddon (paperback by Vintage
Books (May, 2004)—see http://makeashorterlink.com/?M2B03229B).
Evidently, self-proclaimed IQ record-holder Marilyn Vos Savant in
a column called "Ask Marilyn" for Parade magazine laid out this
problem, called the "Monty Hall" after the game show host for
television's "Let's Make a Deal." Imagine that you can pick one
of three doors behind which are two goats and one new automobile.
Before it opens, the host shows that behind another door you see a
goat. You then can keep the door you first chose, or switch to
the other unopened one. What are your odds of winning the car if
you make the switch? (Assume that, from the very start, the game-show host knows what is behind all three doors.) If your
inclination is to say "50/50," do not feel bad—most people give
this wrong answer. Consider that only 1/3rd of the time you will
pick the car, thus 2/3rds of the time you win by picking the last
unopened door after Monty reveals the one with the goat. In other
words, you double your odds of winning by making the switch.
*(Also presented by the math whiz in the American television show "Numb3rs" in an episode from its first season in 2005. Evidently
Hollywood has discovered the power of statistics as noted at http://www.cbs.com/primetime/numb3rs/about.shtml.)
Here's what I cover in the body text of this DOE FAQ Alert (topics
that delve into statistical detail are designated "Expert"):
1. Software Alert: Major new release of Design-Expert® program
2. FAQ: Statistical significance versus practical importance
3. Info Alert: Thought-provoking article on DOE and synergy
4. User feedback: Kudos to making DOE easy
5. Events alert: Fall Technical Conference in Saint Louis
6. Workshop alert: Experiment Design Made Easy in Anaheim, CA
PS. Quote for the month: Einstein's observational study on the
cause and effect relationship for holes in socks—with a
surprising conclusion.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1. Software Alert: Major new release of Design-Expert® program
Stat-Ease announces a major new release—version 7 of
Design-Expert software (DX7). For a free, fully-functional 45-day
trial, click this link: http://www.statease.com/dx7trial.html.
Pricing for new licenses and upgrades can be seen at the Stat-Ease
e-commerce site: http://www.statease.com/prodsoft.html.
Those of you who’ve used previous versions will be impressed with
the many improvements in V7, including:
- Pareto chart of t-values of effects: Quickly see the vital few
effects relative to the trivial many from two-level factorial
experiments.
- Min-Run Res IV (two-level factorial) designs for 5 to 50 factors: Screen main effects with maximum efficiency in terms of
experimental runs.
- On plots of effects simply lasso a box around the ones you
want selected for your model: This is much easier than clicking
each one with your mouse.
- Central composite designs (CCD’s) now available for up to
30 factors and 8 blocks: This represents a significant expansion
in RSM capability.
- CCD’s based on Min-Run Res V fractional-factorial core:Take advantage of a much more efficient design for larger
numbers of factors.
- Box-Behnken designs expanded up to 21 factors: This
popular response surface method (RSM) design previously was
limited to 3,4,5,6,7,9 or 10 factors (note: 8 factors now
possible).
- Crosshairs window: Predict your response at any place in
the response surface plot.
- Full-color contour and 3D surface plots: Graduated or
banded colorization adds life to reports and presentations.
- Magnification feature: Incredible tool for expanding a
mixture graph that is originally a small sliver and difficult to
interpret.
- Mixture-in-mixture designs: Develop sophisticated experiments
for immiscible liquids or multilayer films involving
separate formulations that may interact.
- Add blocks D-optimally: This will be especially useful for
mixture designs, which previously could not be blocked
automatically.
- Points on 3D graphs: See ‘lollipops’ protruding from
surfaces where actual responses were collected.
This is only a partial list from the highlights that you can see
listed in the Getting Started guide to Design-Expert V7 software
at http://www.statease.com/x70ug/DX7-01-Getting%20Started.pdf.
Page down to the Appendix of this document for many more new
features in this landmark upgrade from Stat-Ease, Inc.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
2. FAQ: Statistical significance versus practical importance
-----Original Question-----
From: Michigan
"If an experiment failed to show that a factor had a significant
effect, is the experimenter justified in saying it is not a
factor that needs to be included in the model? In medical study
X, for example, let's say that bedside manner has a real impact,
but not a significant one compared to error or the active drug.
The practical outcome is that a factor that has an important but
not yet detected effect may be dropped or thrifted out' (leading
to poorer outcomes over the long run). However, in the
literature it is frequently recommended that you can use the
cheaper version, or discontinue use, of some chemical that is
'not significant.' What are your thoughts on this?"
Answer:
This is a very brief response to a very weighty question.* Yes,
it is very important to distinguish between an effect of a
magnitude that is practically "important" versus one that is
statistically "significant." In a design with too few runs to
generate adequate power, effects may emerge that are important,
but not significant. In this case, I say "if you feel it's real,
then replicate," (the DOE equivalent of gangsta rapper Snoop Dog
saying "If the ride is more fly, then you must buy.") If the
design provides adequate power to detect even minimally important
effects, then I'd agree in your second scenario that it would be
lucrative to choose the more economical level of factors found
insignificant, for example—the cheaper of two chemicals or the
alternative of not using any, if this is one of the tested levels.
*(You readers are free to weigh in on this far too short answer.
I will publish enlightening comments. For example, see the more
conservative approach suggested by my colleague Pat. —Mark)
From: Pat Whitcomb—Stat-Ease Consultant
"An important consideration in selecting the ultimate level of a
factor is that often we know from first principles and actual
experience that this variable (for example, time or temperature)
must affect the process. If the experimental results show such a
factor to be insignificant, then this only indicates that over
the tested range its effect was small relative to normal
variation. By setting the factor at its mid-point the process
becomes insensitive to variation in that factor, because its
operational level falls within the middle of a plateau-like
response surface. If the factor is set at either extreme it may
be on the edge of the plateau and normal variation may send it
onto a steep downward slope. Unless there is compelling reason
to do otherwise (for example, economically), always set a non-significant factor at its the mid-point to achieve a robust
process."
(Learn more about power 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.)
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
3. Info Alert: Thought-provoking articles on DOE and synergy
The New York Society of Cosmetic Chemists published an intriguing
two-part article titled "Desperately Seeking Synergy" by Joseph
Albanese, which you can view at:
http://www.nyscc.org/news/archive/tech0405.htm (part 1)
http://www.nyscc.org/news/archive/tech0505.htm (part 2).
Joe, with whom I've established a correspondence, spices up this
article on DOE with cartoons and thought-provoking diagrams and
figures. With the spike in oil prices, it's easy to relate to his
re-telling of John Cornell's gasoline blending example of synergy
in mixtures: Wouldn't we all like to squeeze a few more miles per
gallon from the incredibly costly fuel for our automobiles?
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
4. User feedback: Kudos to making DOE easy
From: John a PhD scientist from Maryland
"(You are welcome to pass along my comments in your alert, but
please omit my surname and company. I would be happy to support
your product in that manner.)
Dear Stat-Ease,
Recently, I purchased Design-Expert after examining the
software through a demonstration trial. I am writing to share my
enthusiasm for your software. The tutorial manual is easy to
read, and the examples are very informative. I'm impressed at
how "smart" the software is, allowing flexibility in correcting
selections made in earlier steps and anticipating what the user
needs/is looking for. Your software is leagues ahead of other
DOE packages, for which I never did JMP for joy. I am now able
to focus on my experimental design instead of laboring through
the math myself (which I was doing!). I am truly looking forward
to using Stat-Ease. Thank you!"
Response:
You are welcome! All of us at Stat-Ease really appreciate your
compliments. For your information, I am the author of the User
Guide tutorials. It is a lot of work, but knowing that these are
put to such good use makes the tedium of writing them very
worthwhile.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
5. Events alert: Fall Technical Conference in Saint Louis
Pat Whitcomb will present at talk titled "Using a Pareto Chart
to Select Effects for a Two-Level Factorial DOE" to the Fall Tech
Conference (FTC) held in St. Louis, Missouri on October 20-21. It
details a new method for using a Pareto chart of t-values* so that
relative effect sizes get displayed properly, thus allowing the
addition of t-limits that aid in the selection of the vital few
that are likely to be statistically significant.
*(As noted in item #1, this is a new feature offered in version 7
of Stat-Ease software.)
Click 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: Experiment Design Made Easy in Anaheim, CA
The last Stat-Ease public workshop for 2005 will be a presentation
of Experiment Design Made Easy on December 6-8 in Anaheim,
California. This computer-intensive workshop on the basics of DOE
has been extensively revised to make use of new features in
version 7 of Design-Expert software. Nevertheless, it's main
purpose is to educate on the principles of design and analysis of
experiments, aided by the statistical calculations and graphical
tools provided by the DX7 program.
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.
PLEASE DO NOT SEND ME REQUESTS TO SUBSCRIBE OR
UNSUBSCRIBE - FOLLOW THE INSTRUCTIONS AT THE END OF THIS MESSAGE.
Sincerely,
Mark
Mark J. Anderson, PE, CQE
Principal, Stat-Ease, Inc. (http://www.statease.com)
2021 East Hennepin Avenue, Suite 480
Minneapolis, Minnesota 55413 USA
PS. Quote for the month—Einstein's observational study on the
cause and effect relationship for holes in socks—with a
surprising conclusion:
"When I was young, I found out that the big toe always ends up
making a hole in the sock. So I stopped wearing socks."
—Albert Einstein
(Cited in "Age Doesn't Matter Unless You're a Cheese" by Kathryn
and Ross Petras. I've also heard that it's good to be old,
unless you are a banana. Mark)
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, #3-6 Jun 03, #3-7
Jul 03, #3-8 Aug 03, #3-9
Sep 03 #3-10 Oct 03, #3-11
Nov 03, #3-12 Dec 03, #4-1
Jan 04, #4-2 Feb 04, #4-3
Mar 04, #4-4 Apr 04, #4-5
May 04, #4-6 Jun 04, #4-7
Jul 04, #4-8 Aug 04, #4-9
Sep 04, #4-10 Oct 04, #4-11
Nov 04, #4-12 Dec 04, #5-1
Jan 05, #5-2 Feb 05, #5-3
Mar 05, #5-4 Apr 05, #5-5 May 05, #5-6 Jun 05, #5-7 Jul 05, #5-8 Aug 05, #5-9 Sep 05, #5-10 Oct 05 (see above)
Click here to add your name to the FAQ
DOE Alert newsletter list server.
Statistics Made Easy™
DOE FAQ Alert ©2005 Stat-Ease, Inc.
All rights reserved.
|
|
|
