Ah, movie night! My wife and I agree thatmovie night, which we have in our houseevery Sunday, is our favorite night of theweek. Now, its not necessarily the moviethat we enjoy so much. If we happen towatch a dud of a movie, its still a goodnight. I guess its just the time spent togeth-er that we enjoy and of course, the pop-corn! We cannot have a movie night with-out those tender white morsels drizzled inbutter and salted to perfection. A moviejust wouldnt be the same without it.

To enhance our favorite night of theweek, we decided to experiment withour popcorn popping technique to try tomaximize the delight of our taste budsand minimize the wasted unpopped ker-nels at the bottom of our bowl. Theresnothing worse than getting shorted onthe pure enjoyment of eating popcorn.

Being experimenters, my wife and I havetried many popping methods, includingthe easiestmicrowave popcornbutweve found that nothing beats the tasteand enjoyment we get from cooking iton the stove with the Whirley-Poppopcorn popper. Its pictured above.

The Setup: Factor and Design SelectionThere are numerous variables to playwith when using this stovetop popper.One question we wanted to answer was,Do we really need to stand at the stoveand stir the whole time its popping? Asecond question arose from my parentsrecommendation that we pre-heat thepot by placing one kernel in the pan

Stat-Teaser News from Stat-Ease, Inc.

Workshop ScheduleA B O U T S TAT - E A S E S O F T W A R E , T R A I N I N G , A N D C O N S U LT I N G F O R D O EPhone 612.378.9449 Fax 612.746.2069 E-mail info@statease.com Web Site www.statease.com

April 2010 1

until it pops, and then adding in the restof the kernels. We had never done itthis way, so we wondered if it wouldmake a difference. Lastly, we wanted toaddress whether spending the extramoney for a premium popping corn wasworth the price. So, we had three fac-tors to investigate: Stirring, Pre-heattime, and Popcorn type.

We chose to do a full 23 factorial, testingall possible combinations. We alsoadded three center points. The full fac-torial with center points is a good choicebecause it allows us to test for curvature.If curvature is detected, we can augmentto a central composite design (CCD) tomodel the quadratic terms by adding afew extra runs to the original data. Inall, we tested three factors at 2 levels:

A. Stirring, in fraction of time stirredfrom 0.0 to 1.0

B. Pre-Heat Time, in seconds from 0to 360

C. Popcorn Type (categoric), from Cheap to Costly.

Because Factor C is categoric, the three

Whirley Pop DOEWhirley Pop DOE

Continued on page 2

Experiment Design Made EasyApril 27-28, 2010: Minneapolis, MNJune 15-16, 2010: Minneapolis, MNStudy the practical aspects of design ofexperiments (DOE) and two-level factorialdesigns. $1295 ($1095 each, 3 or more)

Response Surface Methods for Process OptimizationApril 29-30, 2010: Minneapolis, MNMaximize profitability by discovering optimal process settings via RSM.$1295 ($1095 each, 3 or more)

Mixture Design for Optimal FormulationsMay 18-19, 2010: Minneapolis, MN Learn the basics of mixture design.$1295 ($1095 each, 3 or more)

Advanced FormulationsMay 20-21, 2010: Minneapolis, MN See page 3 for information. $1495 ($1195 each, 3 or more)

Designed Experiments for Life SciencesMay 10-11, 2010: San Diego, CALearn how to apply DOE to Life Scienceproblems. $1495 ($1195 each, 3 or more)

European DOE User ConferenceJune 1-2, 2010: Lucerne, SwitzerlandSee page 3 for information. 420

Free Webinar: DOE Made Easy& More Powerful via DX8April 28, 2010 at 10:00 am CTSee http://www.statease.com/webinar.html.

PreDOE: Basic Statistics forExperimenters (Web-Based)For anyone who needs to go back to the basics of statistics. Seewww.statease.com/clas_pre.html formore information. FREE (a $95 value)Workshops limited to 16. Contact Elicia at612.746.2038 or workshops@statease.com.

2010 Stat-Ease, Inc. All rights reserved.

Brooks making Whirley Pop Popcorn

2 April 2010

center points we specified became six,three for each level of this categoric factor.

To facilitate using center points, wewanted to make as many factors numer-ic as possible. Factor A, stirring, was alittle tricky. We didnt want to just teststirring vs. not stirring, because thatwouldnt allow us to make a center pointand test an intermediate amount of stir-ring. To make it numeric, we defined itas the fraction of time stirred. The lowand high levels are self-evident, 0.0 is notstirring and 1.0 is constant stirring. Atthe center points, which required 0.5fraction of time stirred, we stirred for 15seconds, stopped stirring for 15 seconds,stirred 15 secondsand so on.

The ResultsWe measured four responses for thisexperiment. Taste was rated on a 1 to 5scale with 5 being the best. Texture wasalso rated on a 1 to 5 scale with 1 beinghard and 5 being soft. We prefer tenderpopcorn, so a value of 5 was desirablehere. We counted the number ofunpopped kernels (UPKs) after poppingand this was our third response. Thefinal response was an estimate of the vol-ume of popcorn in the pot after popping.Unfortunately, we didnt have a graduat-ed cylinder for measuring volume.Thats probably the only piece of super-fluous kitchenware that we didnt receiveas a wedding gift. So, we just estimatedthe volume in the pot after popping.Given the wide range of our responses, itseems we had enough discrimination todraw some conclusions here.

Stirring and Pre-Heat both had a signif-icant effect on taste and texture. In eachcase, it was best to go with constant stir-ring and no pre-heating; so much forour dreams of not having to stir at all. Iguess Whirley-Pop knew what theywere doing when they added that stir-ring mechanism to the pot. On theother hand, not having to pre-heat is a

Continued from page 1

win for us. There is one caveat, though.Take a look at the stirring/pre-heatinteraction graph for taste (Figure 1).At a stirring rate of 1.0 (right side ofgraph), it appears that no pre-heat (B-)will give the best taste. However, theerror bars overlap slightly, so it cant bedeemed statistically significant. We dohave enough information to go withoutpre-heating in the future, though. Evenif we had just equivalent taste with andwithout the pre-heating, we can savetime by cutting out this step. Whenthings are statistically equivalent, gowith the easier or cheaper option.

Another thing to note is the significantlack of fit detected by our center points.Look at the green dots in Figure 1.These are the results at our center pointtreatments using the expensive popcorn.The true data points lie above the linespredicted by our model. It appears wemay be able to get better tasting popcornif we only stir half of the time or someintermediate proportion. I like the soundof that: less work and better taste! TheANOVA summary table also showedthat there is significant curvature with ap-value of 0.0058 for the adjusted model.

The significant curvature suggests that wecan better fit the data by adding a quadrat-ic term, either A2 or B2, but because weonly did a factorial design, we dont haveenough data to determine which term isneeded. We will have to augment the cur-rent design with more points to get a CCD

for response surface methods (RSM),which will tell us which term is correct.

For UPKs, the only factor found to besignificant in the design space was factorA, stirring. More stirring reduced thenumber of unpopped kernels signifi-cantly (from a mean of 73.9 UPKs toonly 2.8 UPKs). The volume of popcornin the pot after popping varied mostlydue to the stirring and pre-heat factors.Just like with taste, more stirring and nopre-heating was the best combination,giving us the highest volume of popcorn.

What did we learn?Constant stirring (a level of 1.0) with nopre-heating gave the most desirableresponses for taste, texture, and volumeof popcorn. The brand of popcorn did-nt make a statistically significant dif-ference over the range tested, but it didappear to have some effect. We shouldinclude it in further studies. The analy-sis also detected significant curvaturefor the taste response that would sug-gest that we can get a better taste ratingif we stir for an intermediate proportionof time. However, constant stirringreduced the number of UPKs. Todecide whether we should use constantstirring to reduce UPKs or an interme-diate level to possibly increase taste andtexture, well have to upgrade to aresponse surface design so that we canmodel that intermediate proportion ofstirring. If we want to keep the samefactors and levels, we can simply aug-ment the current design with a fewmore runs to get a CCD. However, wemay want to throw the other knownfactors (time and temperature) that wedidnt test into the final characteriza-tion and optimization study. Stay tunedfor Part II.

Brooks Henderson, DOE Consultantbrooks@statease.com

Brooks is the newest trainer/consultant atStat-Ease, Inc. Take a look at his biographyat http://www.statease.com/brooksh.html.

B: Pre-heat

Interaction

A: Stirring

Tast

e

B+

B-

Fig. 1: Effect of stirring and pre-heat on the taste response

April 2010 3 Stat-Teaser News from Stat-Ease, Inc.

New Advanced Mixture Design WorkshopCombine one mixture with another(frosting on the cake)Map mixtures to process space viaratios, thus enabling the application offactorials and response surface methods(RSM)Become an expert on innovative mix-ture models such as the partial quadraticUnderstand when it is advantageousto invert the mixture space via re-cod-ing of c