Issue: Volume 1, Number 6
Date: August, 2001
From: Mark J. Anderson, Stat-Ease, Inc.
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. Feel free to forward this newsletter to your colleagues. They can subscribe by going to: http://www.statease.com/doealertreg.html.
I apologize for being a bit tardy getting out this month's "Alert." I just got back from a biking vacation on Wisconsin's Sparta-Elroy rail-trail, which I highly recommend: See http://www.elroy-sparta-trail.org/ for pictures and details. The neat parts of this trail are the three tunnels. These were literally very cool. Thank goodness for that, because I rode the trail on the hottest and most humid day in many years in these parts. On that same day a Minnesota Viking professional football player died of heat stroke. :(
Now it's time to get back to business - "Statistics Made Easy." Before I get into the meat of this message, I offer an applet from the Rice Virtual Lab in Statistics (RVLS) as an appetizer: http://www.ruf.rice.edu/~lane/stat_sim/conf_interval/index.html. It generates 95% and 99% intervals based on specified sample sizes. You (or your students) might find it enlightening. Click back to the RVLS home page to see what else the author, David Lane, offers for those who don't know a t-test from a p-value, or for anyone needing a quick refresher on statistical concepts.
Here's what I cover in the body text
of this DOE FAQ Alert:
1. FAQ: Why invest in specialized DOE software/workshops
versus general Six Sigma software/training?
2. Reader feedback: Dangers of happenstance data.
3. Events alert: A heads-up on DOE talks and demos.
4. Workshop alert: New "Statistics for Technical
Professionals" debuts in Minneapolis.
PS. Statistics quotes for the month: Subject - Happenstance
data.
PPS. Photo of the month: The answer is: Fear of statistics.
Click on the link provided for the question.
1 - FAQ: Why invest in specialized DOE software/workshops versus general Six Sigma software/training?
-----Original Question-----
From: Several inquirers (paraphrased)
"Our top management recently committed to a Six Sigma
program. They've chosen a vendor who offers training for Master Black Belts,
who then will train Green Belts. Their materials include coverage of DOE, which
they expect to be performed with the aid of a general statistics package. Our
top management wonders why, if given these new Six Sigma resources, they should
support use of specialized DOE software and in-depth training on the associated
statistical methods."
Answer:
DOE is the only statistical tool that's dedicated to making breakthrough improvements, thus making it vital for the success of Six Sigma. To take full advantage of DOE, Master Black Belts (MBBs) must go beyond the broad-and-shallow training/software provided by the pre-fab programs.
Take a close look at how much time is devoted to DOE in your Six Sigma curriculum. Many programs devote only a few days to DOE and cover only factorial design. This might be enough for some people, but not those who work in R&D, process engineering or the like. Some Six Sigma vendors do delve into response surface methods, but in a very limited manner. A good litmus test for the quality of DOE training is whether it covers the topic of propagation of error (POE), a very powerful RSM tool for reducing variation transmitted from input factors. (For details on POE, refer to "DOE FAQ Alert" Volume 1, Number 3, item 3 at http://www.statease.com/news/faqalert3.txt.) If you're in the process industry (food, pharmaceutical, chemical, paint, etc.), a further acid test is whether there's any coverage of mixture design for optimal (and robust!) formulation.
The software used in your Six Sigma
program should also be scrutinized for DOE power. One vendor that I found on
the Internet offers only a few Excel templates to do two-level designs on 2
or 3 factors! However, most of the Six Sigma vendors align themselves with a
general statistics package that ostensibly offers many DOE features. These programs
do an adequate job on simple, standard designs and give correct statistical
analyses so far as they go. They generally don't make it easy to do a proper
diagnosis of residuals. If there's a need for response transformations, these
general stats packages may require some tricky navigation to find the proper
tools. Other features you'd like to make easy for Six Sigma users are:
- Multilinear constraint capabilities
(to avoid infeasible, and possibly dangerous experimental conditions such as
high time at high temperature)
- Design evaluation and repair for botched experiments that become badly aliased
- Multiple response optimization (numerical and graphical) to find the "sweet
spot" meeting all specifications with minimal variation (including POE
as a criteria) at lowest cost.
I could get into many other issues on the quality of graphics, etc., but the bottom line is that software programs offering all of the Six Sigma tools probably don't do any one of them as well or as easy as a devoted package. Some of our more sophisticated clients have put this to the test by sitting their people down and seeing how much more quickly they learn how to use dedicated DOE software (such as Design-Expert) versus various general statistics programs.
Consider the huge loss of potential revenue if people do not perform proper DOE's due to inadequate training or bad software. Don't shy away from going beyond the standard Six Sigma template offered by your vendor - augment it with training and software that's dedicated to DOE. Then you will really see the big paybacks touted by others who've done Six Sigma, such as General Electric, who contracted Stat-Ease to develop specialized training for DOE built around Design-Expert software.
(See http://www.statease.com/clasrdrv.html
for course content of the "Robust Design, DOE Tools for Reducing Variability"
workshop, which provides a plethora of tools for achieving Six Sigma objectives.
This computer-intensive class requires proficiency in RSM which can be gained
by attending Stat-Ease's three-day computer-intensive workshop called "Response
Surface Methods for Process Optimization". See http://www.statease.com/clas_rsm.html
for a description and links to the RSM course outline and schedule.)
2 - Reader feedback: Dangers of happenstance data
-----Original Comment-----
From: UK reader of DOE FAQ Alert Volume 1, Number 5, Item 1 (See http://www.statease.com/news/faqalert5.html.)
"OK, DX6.04 can handle it [happenstance data via
an option under Response Surface design called "Historical Data."],
but you might want to warn people of the issues of using historical production
or happenstance data."
Answer:
All of you readers, consider yourself warned! Based on personal experience in the chemical industry, my feeling is that it rarely, if ever proves beneficial to spend time going back over historical data, other than to establish a starting point for a proper DOE. As my partner Tryg says: "Trying to glean useful information out of happenstance data is akin to resurrecting a gourmet meal from a garbage can." (Also see the related quote in the postscript to this e-mail.) For the definitive discussion on this topic, see section 14.7 (page 487) entitled, "Hazards of Fitting Regression Equations to Happenstance Data," in Box, Hunter and Hunters' "Statistics for Experimenters," which can be purchased via our e-commerce site at http://www.statease.com/prodbook.html.
[Learn
more about developing good predictive models by attending the "Response
Surface Methods for Process Optimization" workshop (see http://www.statease.com/clas_rsm.html).]
3 - Events alert: A heads-up on DOE talks and demos.
Stat-Ease marketing director Heidi Hansel and statistical consultant Doug Hubbell are displaying our DOE software and workshop materials this week at the Joint Statistical Meetings of the American Statistical Association (ASA) in Atlanta. This is a broad conference with only a small percentage of talks of any interest to industrial practitioners, but some of these are very good, such as two by General Electric on use of DOE (featuring use of Design-Expert software) for their Six Sigma program. (Go to http://www.amstat.org/meetings/jsm/2001/onlineprogram/ (oldest archive is 2005) and search by author for two presentations by staff at GE Corporate Research and Development (CRD): Angela Neff and Christopher Stanard.)
We will also exhibit at the Fall
Technical Conference (FTC) in Toronto on October 18-19. I highly recommend the
FTC for those of you advanced practitioners who want to broaden your knowledge
of applied statistical tools, including DOE. You can get all the details from
the co-sponsors' sites:
- American Society of Quality's (ASQ) Statistic's Division at http://www.asqstatdiv.org/
(click on the FTC link under Hot News)
- ASA's Section on Physical and Engineering Sciences (SPES) at http://www.amstat.org/sections/spes/
(click "Fall Technical")
(Note: ASQ's Chemical and Process Industries Division is also a sponsor for
the Fall Technical Conference.)
4 - Workshop alert: New "Statistics for Technical Professionals"
Our new 2-day "Statistics for Technical Professionals" workshop debuts October 30-31 in Minneapolis. Despite the conjunction of this presentation with Halloween, we do not intend to scare attendees with statistics. In fact we hope to do just the opposite: Show engineers, scientists and quality professionals how to gear up their stats knowledge to achieve Six Sigma objectives or other quality improvement initiatives. For all the details, see http://www.statease.com/clas_stp.html. To enroll, just hit the "register online" link on this page. We hope to see you this Fall. Come as you are - no costume is required.
Here's a heads-up for those of you
in the Northwestern US: I will present another session of the 1-day "DOE
Simplified" in Seattle on Sept. 13. My last presentation in Philadelphia
sold out, so don't delay enrolling in this fun and informative overview of the
power of DOE tools. See http://www.statease.com/does.html
for details on the class content.
See http://www.statease.com/clas_pub.html
for schedule and site information on all Stat-Ease workshops open to the public.
To enroll, call Stat-Ease at 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 questions and comments to me at:
PS. Statistics quote for the month:
"It
has been said that data collection is like garbage collection: BEFORE you collect
it you should have in mind what you are going to do with it."
- Russell Fox, Max Gorbuny and Robert Hooker, authors of "The Science of
Science"
PPS. Photo of the month: The answer
is: Fear of statistics. See: http://nationalgeographic.com/ngm/tellus/0108 (this page no longer exists)
for the question.
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
Sincerely,
Mark
Mark J. Anderson, PE, CQE
Principal, Stat-Ease, Inc. (http://www.statease.com)
Minneapolis, Minnesota USA
Click here to add your name to the FAQ DOE Alert newsletter list server.