Stat-Ease | The DOE FAQ Alert Vol. 13 No. 3

PS. Quote for the month: The Winner’s Curse plagues scientists who run inadequately powered experiments.

1: Newsletter alert: The April issue of the Stat-Teaser tells how “UK Boffins Pull Off Brilliant DOE on Beer” and details “The Perils of Parts and the Failure of Fillers as Excuses to Dodge Mixture Design”

Many of you have received, or soon will, a printed copy of the latest Stat-Teaser, but others, by choice or because you reside outside of North America, will get your only view of the April issue at this link. It features a report by me on an educational experiment by the technical staff at PRISMTC of Cambridge, England. Through methodical testing and statistical analysis, they mastered how to pour beer so it develops just the right head of foam. Brilliant!

This Stat-Teaser also provides a white paper by me that explains why using mixture design, not factorials or response surface methods (RSM), works best for optimizing formulations. Tools provided by Design-Expert® software make design and analysis of mixture experiments very easy and greatly informative on how components can be combined in new ways to vastly improve product performance.

2: FAQ: Why do some standard errors come up blank when I make a point prediction?

From an R&D Engineer:
“In the point prediction some of the outputs have standard errors [SEs] of the mean and prediction, while others are completely blank. Is that because those responses were transformed? If so, I don’t recall seeing it before. And if there is no standard error of the mean or of the prediction, how is it calculating CI [confidence interval] and PI [prediction interval] for those outputs?”

Yes, these are responses that were transformed. Change Display Options to Response in Transformed Scale to see all SEs. Now it becomes clear that the intervals, by necessity, are calculated in the transformed scale. Return to original scale to see the CI and TI in more convenient metrics, but do not be alarmed to then see the standard error values blank out. —Mark

(Learn more about transformations by attending the two-day computer-intensive workshop Experiment Design Made Easy. Click on the title for a description of this class and link from this page to the course outline and schedule. Then, if you like, enroll online.)

Original Question:

From a Quality Consultant:
“Could you please clarify for me how power is calculated in fractional-factorial designs. Your program Help says that ‘Power for unreplicated factorials can only be estimated if you designate some effects as error.’ I assume this will also apply to fractional designs, which are not replicated. I am not sure how we are to designate the effects as error. For example, I would have thought that the eventual power will be determined by the number of degrees of freedom for error, which will apply when I eventually choose the effects. However, the number of effects I select (which determines the number of df [degrees of freedom] available for error) doesn’t seem to affect the power calculation.”

Answer:

From Stat-Ease Consultant Wayne Adams:
“In the Design Evaluation branch of Stat-Ease software, power, by default, is calculated for the model for which you designed your experiment—not the model you select when analyzing the actual results. However, you can re-evaluate power for your selected model to produce a ‘post power’ analysis; that is, assuming some degrees of freedom remain for estimating error.

Design Evaluation for Post-Power
Design-Expert screen shot showing model selected in Design Evaluation for ‘post-power’ calculation

When you build a factorial design, Stat-Ease software generates a power calculation based on signal and noise estimates that you enter in. At this stage the program defaults to the linear (main factor effects) model, because meta-analysis of published peer-reviewed reports reveal that in any given experiment the number of significant effects can be approximated by the number of factors tested. This makes the linear model convenient as the default because it always contains k terms.

Main Effects Model When Calculating Power Main effects model used by default when calculating power before running an experiment

If you should ever want to calculate power on your own, check out this white paper, upon which we based our methods. Also, search our web site for other articles on power.”

(Learn more about power by attending the two-day computer-intensive workshop Experiment Design Made Easy. Click on the title for a description of this class and link from this page to the course outline and schedule. Then, if you like, enroll online.)

4: Info alert: Stat-Ease is featured in an article on DOE by Chemical & Engineering News

C&E News in their April 1st issue touted DOE in their article on how this “Statistical Method Makes A Comeback.” It quotes me and users of Design-Expert software. :-) To see it, follow this link.

On Tuesday, June 25 at 11 AM CDT* Stat-Ease Consultant Wayne Adams will provide a briefing on how to achieve Quality by Design (QbD) Space for Pharmaceuticals and Beyond. This free webinar, presented at an intermediate level statistically, will be essential for anyone subject to regulation by the Food and Drug Administration (FDA). However, the QbD methods go far beyond regulated industries—they serve well for anyone aiming for robust manufacturing at the highest quality levels.

To accommodate those outside of the Americas, this vital briefing on QbD will be repeated twice more at times shown below. If you cannot schedule any of these webinars, take advantage of the recording to be posted after the series of live events.

Space is limited. Reserve your Webinar seat now at by clicking one of the links below:

If this is your first Stat-Ease webinar, see these suggestions on how to be prepared.

Stat-Ease webinars vary somewhat in length depending on the presenter and the particular session: Plan for 45 minutes to 1.5 hours, with 1 hour being the target median. When developing these one-hour educational sessions, our presenters often draw valuable material from Stat-Ease DOE workshops.

Again, attendance may be limited, so sign up soon via the link above. Direct any questions you may have to our Communications Specialist, Karen Dulski, via [email protected]. However, if this relates to audiovisual issues, please first research help provided online by GotoWebinar.

*(To determine the time in your zone of the world, try using this link. We are based in Minneapolis, which appears on the city list that you must manipulate to calculate the time correctly.)

Stat-Ease Consultant Pat Whitcomb will reveal “Practical DOE 'Tricks of the Trade'” at the Quality & Productivity Research Conference (QPRC) in Schenectady, NY on June 5-7. If you attend, stop and talk with Pat at our display table. Details on QPRC can be found here.

Stat-Ease Sales & Marketing Director Heidi Hansel Wolfe will exhibit our software to the American Chemical Society's Great Lakes Regional Meeting, La Crosse, WI on June 5-8. To register click this link.

All classes listed below will be held at the Stat-Ease training center in Minneapolis unless otherwise noted. If possible, enroll at least 4 weeks prior to the date so your place can be assured. Also, take advantage of a $395 discount when you take two complementary workshops that are offered on consecutive days.

See this web page for complete 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 Elicia at 612-746-2038. 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. Once you achieve a critical mass of about 6 students, it becomes very economical to sponsor a private workshop, which is most convenient and effective for your staff. For a quote, e-mail [email protected].

I hope you learned something from this issue. Address your general questions and comments to me at: [email protected].
Please do not send me requests to subscribe or unsubscribe—follow the instructions at the end of this message. Sincerely,

PS. Quote for the month—The Winner’s Curse plagues scientists who run inadequately powered experiments:

"
The winner’s curse refers to the phenomenon whereby the ‘lucky’ scientist who makes a discovery is cursed by finding an inflated estimate of that effect. The winner’s curse occurs when thresholds, such as statistical significance, are used to determine the presence of an effect and is most severe when thresholds are stringent and studies are too small and thus have low power.”

—“Power failure: why small sample size undermines the reliability of neuroscience,” Nature Reviews Neuroscience, Katherine Button, John Ioannidis, et al.

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

Acknowledgements to contributors:
—Students of Stat-Ease training and users of Stat-Ease software
—Stat-Ease consultants Pat Whitcomb, Shari Kraber, Wayne Adams and Brooks Henderson
—Statistical advisor to Stat-Ease: Dr. Gary Oehlert
—Stat-Ease programmers led by Neal Vaughn
—Heidi Hansel Wolfe, Stat-Ease sales and marketing director, Karen Dulski, and all the remaining staff that provide such supreme support!