AdvertisingResearch:Instructor’sManual

Copyright?2012PearsonEducation,Inc.publishingasPrenticeHall

PAGE

PAGE

AdvertisingResearch:Instructor’sManual

Copyright?2012PearsonEducation,Inc.publishingasPrenticeHall

15.QuantitativeDataAnalysis:DescriptiveStatistics

ChapterGoals

Thischapterisdividedintofivesections.Eachsectionisdesignedtohelpstudentsbetterunderstandtheprocessbywhichresearchersmovefromrawdatatomeaning.

? Thefirstsectionexplainsthefiveprimarymeasuresneededtoanalyzequantitativedata.

? Thesecondsectiondescribesthestartofthedataanalysisprocess,

focusingonhowtoensurethatdataisreliableenoughtoserveasthebasis

forsubsequentdecisions.

? ThethirdsectionshowshoweachofthespecifictypesofquestionsintroducedinChapter13areexaminedandanalyzed.

? Thefourthsectionshowshowtotakethelaststeptowarddatameaning,movingawayfromananalysisofthetotalsampleintoanalysesofsmallergroupsofrespondents.

? Thelastsectionbringsallofthepriorinformationtogetherthroughastep-by-stepdemonstrationoftheentiredataanalysisprocess.

NotestotheInstructor

TheChapterLectureprovidesaguidetokeytopicsandcontent.ThePowerPointslidesarenamed:davis_adresearch_ch15.ppt.

ChapterLecture

I.BasicMathandKeyMeasures

Slide15-

2

Slide15-

3

Fivebasicandeasytocomputemeasuresprovideallthemathneededtoderivemeaninginmostdataanalysissituations:percentage,averageormean,median,mode,andstandarddeviation.

A.Percentage

Thinkofascoreonanexam.Tocalculatepercentagedividenumberofcorrect

answersbytotalnumberofquestions,andthenmultiplyby100.Sixteencorrectanswersoutoftwentywouldgiveyouascoreof80%.

Slide15-

4

Slide15-

5

Slide15-

6

Slide15-

7

Slide15-

8

Slide15-

9

Useofpercentagesforanalysisofsurveydataworksexactlythesameway:dividenumberofpeoplegivingaparticularresponsebytotalnumberofpeopleansweringquestion.TableshowninSlide15-4showshowthisisdone.

? Firstcolumnshowsthepossibleanswerstothesurveyquestion,

? Secondcolumnshowsnumberofpeopleselectingeachoption,and

? Lastcolumnshowspercentageofpeopleselectingeachoption.Totalofthepercentagecolumnshouldalwaysaddupto100%.

Ifthesurveyusesoneoftheonlinesurveysites,thesitewillautomaticallyprovidepercentages.Slide15-5isexampleofasurveywith10respondents.Zoomerang’spresentationofresultsshowsthepercentageofrespondentsselectingeachoptionforeachquestion.

B.Average

Calculateanaverageforasurveyquestioninmuchsamewayasyoucalculateaclassaverage,exceptinthecaseofsurveyquestionswedividethetotalnumberofpointsforthequestionbythetotalnumberofpeopleansweringthequestion.ConsiderquestionshowninSlide15-6.

Eachresponseoptionhasbeenassignedavalue.Tenconsumersratethecommercialasfollows:1,2,3,2,3,4,42,1,2.Theaveragerelevanceratingwouldbe2.4,calculatedbydividingthesumofalloftheindividualscores,24,bythenumberofpeopleansweringquestion,10.

C.MedianandMode

Medianisanumberthatappearsinthemiddleofanorderedsetofdatawhenthescoresareorderedfromlowesttohighest.

Fivepeopleprovideage:21,23,25,26,99.Theaverageagewouldbe38.8.Butthisaverageseemsnottofitthedata,asnorespondenthasanagearound

39andinfactfouroutoffiveareaged26oryounger.Averageishighbecauseofindividualaged99.

Medianagewouldbe25;obtainedbyfirstputting/makingcertainscoresareinorder,thenfindingthescorethatisinthemiddle.Thisnumber(25)ismuchmorerepresentativeoftheagesinthedatasetversustheaverage.

Whenyouhaveanaveragethatisdistortedbyafeweitherveryhighorverylowscores,thenmedianoftenbecomesbetterwaytosummarizesetofscores.

Slide15-

10

Slide15-

11

Slide15-

12

Slide15-

13

Slide15-

14

Slide15-

15

? Whenyouhaveanoddnumberofscores,medianisthescoreexactlyinthemiddleoftheorderedsetofdata.

? Whenyouhaveanevennumberofscores,themedianistheaverageofthetwomiddlescores.

Modeisthenumberinasetofresponsesthatappearsmostoften.Modeisusefulwhenyouneedtoknowmostfrequentresponse.

ModefordatashowninSlide15-12wouldbe“1”representing“Agree.”

D.RelationshipofMean,MedianandMode

Mean,median,andmodeeachprovideadifferentinsightintoasetofscores’distribution.Imagine100studentstakeanexam.Distributionofexamscoresissymmetricalwhenmean,median,andmodearenearlyidentical.Inthesecases:(1)distributiontorightofmean,median,andmodeisanearmirrorimageofdistributiontotheleftofthesemeasuresand(2)majorityofscoresfallintothecenterofdistribution.Thisdistributionistypical“bellcurve.”Whendistributionis(orisnear)symmetricalthenmeanisanaccurateandpreferreddescriptorofdistribution.

Distributionsdonothavetobesymmetrical.

? Distributionsinwhichthevalueofthemodefallsbelowtheme-dianthatinturnfallsbelowmeanskewsleft.Distributionhasalargenumberofvaluesatlowerendofdistributionandfewvaluesatthehighendofdistribution.

? Distributioninwhichthevalueofthemodefallsabovemedianthatsubsequentlyfallsabovethemeanskewsright.Distributionhasalargenumberofvaluesathighendofdistributionandfewvaluesatlowendofdistribution.

Themoreskewedthedistribution,thelesswellthemeanprovidesagoodsummaryofthatdistribution.Asaresult,forheavilyskeweddistributions,eithermedianormodeisoftenpreferreddescriptorversusaverage.

E.StandardDeviation

Standarddeviationisameasureofdispersion.Dispersionreferstohowspreadoutorclusteredasetofscoresisaroundthemean.

Slide15-

16

Slide15-

17

Slide15-

ConsiderdatashowninSlide15-16.Datarepresentsfiftyconsumers’purchaseintentafterseeingoneofthreecommercials.Dataineachcolumnrepresentspercentageofselectingeachresponse.Dataonthebottomofthetablereportsmeanpurchaseintentforeachcommercial.

Meanpurchaseintentforeachcommercialisidenticalinspiteofthefactthatunderlyingdistributionsarequitedifferent.ResponsestoAd1areevenlyspreadoutamongresponseoptions.ResponsestoAd2fallexclusivelyattheendsofthepurchaseintentscale.ResponsestoAd3resembleabellcurve.

Standarddeviationisawaytounderstandandcomparedistributionsofscoreswithouthavingtoactuallyplotandexamineeachandeveryscore.Foranyparticularscale,thegreaterthestandarddeviation,thegreaterthedispersionofscoresandasaresult,thelessaccuratethemeanforsummarizingasetofscores.ThetableshowninSlide15-17addsstandarddeviation.

StandarddeviationissmallestforAd3,largerforAd1andmuchlargerforAd

2.Aresearcherwouldconcludethatunderlyingpatternsofresponsetothethreecommercialsareverydifferent,andasaresult,needstolookdeeperintotheresponsepattern.OnlyinthecaseofAd3isthemeanagoodsummary

measureofthedistribution.

Itisimportanttorememberthatincomparingstandarddeviations,anyparticularstandarddeviationisareflectionoftheunderlyingscale.Standarddeviationscannotbecomparedwhenscaleshavedifferentnumbersofoptions.

II.MakingCertainYouHaveGoodData

Successfuladvertisingandrelateddecisionsarebasedongood(i.e.,correctandappropriate)responses.Tomakecertaindatais“good”youmust“clean”yourdatapriortoconductinganyanalyses.

A.DataReview,DecisionsandEditing

Firststepinpreparingdataistocheckresponsestoeachquestion,lookingforquestionsthathavealargenumberofmissingresponsesorquestionsthathaveodd,inappropriate,orunexpectedresponses.Youcanperformthisbyexaminingeachquestion’spercentagedistributions.Youmustdecidewhichquestionsareproblematicandwhattodowitheach:acceptasis,eliminateoredit.

ConsiderinformationprovidedbyZoomerangforfourquestions.

? Question1focusesonbeeradvertiser’scompetition.Almostall

18

Slide15-

19

Slide15-

20

Slide15-

21

whoansweredquestion(90%)choseHeinekin.Whilethishighpercentagemightnotbeexpected,dataappearstobe“good”astheredoesnotappeartobeanyproblemwithquestionwordingorresponseoptions.Appropriatedecisionatthispointwouldbetoacceptandusedataasis.

? Question2showsthesamepatternofresponseasQuestion1.

Allrespondentschosejustoneresponse.Aclosereviewofthis

question’sresponseoptionsshowsthatthereweretypographicalerrors-thefinaltwooptionsshouldhaveread“unimportant”insteadof“important.”Asaresult,scaleisincorrectanddatafromthisquestionisnotuseable.

? Question3isaproblem,butforadifferentreason.Whiletenindividualscomprisetotalsampleonlytwoindividualsansweredthisquestion.Anyquestionthathasahighnumberof“noresponses”shouldbeeliminatedfromanalysis.

? Question4illustratesanadditionalproblem.Tworespondentsprovidedanswersthatseemunreasonable.Youcanexaminethetotalpatternofresponseforthesetwoindividuals:ifpatternseemsreasonableyoucanacceptdataasis;ifthepatternseemsreasonablebutthisonemeasureseemsoddorinconsistentyoucanrecodetheresponseas“missing”;orifthepatternofresponseisproblematicacrossanumberofquestions,youcaneliminatetherespondentfromthesample.

Examinationofdataonquestionlevelshouldbefollowedbyareviewofeachrespondent’sindividualpatternofresponse.Here,lookingforanyrespondentswhohavealargenumberofmissingoroddresponsesacrossallsurveyquestions.

ImagineacasewheretenrespondentsareaskedtorateYouTube.Analysisofeachquestionshowsarangeofresponseand80%oftotalsampleansweredeachquestions.Analysisonthislevelappearspositive.

Thingsaredifferent,however,onindividualrespondentlevel.

Slide15-21showsresponsesofeachrespondentonthefoursurveyquestions.Dataindicatesthatthefirstsixrespondentsansweredallofthequestionswhilerespondentsnineandtenonlyansweredhalfofthequestions.

? Researcherstypicallydeletefromsamplerespondentswhohavefailedtoansweralargenumberofquestions.

Additionally,responsesofrespondentsfiveandsixalsoappeartobeaproblem,aseachindividualgaveexactsameanswertoallquestions.

? Wheneveryansweristhesame,itisquitelikelythattherespondentdidnotgivethesurveyfullattentionandresponsesareprobablynotvalid.Theserespondentsshouldalsobeeliminatedpriortodataanalysis.

III.DataAnalysisforSpecificQuestionTypes

UsequestionsfromBeboresearchstudy.[MakecertainstudentshaveFigure

15.7availableforreference.]

A.Classification,ChecklistandOtherNominalLevelQuestions

Slide15-

22

Slide15-

23

Slide15-

24

Slide15-

25

Slide15-

26

Questions1through3arenominallevelquestions.Becausenumbersusedtocoderesponsestonominallevelquestionshavenointrinsicmeaning,averagesareinappropriate.Instead,mostcommonlyusepercentages(showninSlide15-

22).

Initialreportingofpercentagesshouldalwayspresentthedataintheformcollectedbysurveyquestion.Next,summarizedatainwaysthatdonotdirectlycorrespondtotheoriginalresponsecategories.Here,originalresponsecategoriesarecombinedinlogicalfashionandtotalpercentagesfornewcategoriesarepresented.

? Responsestotheeducationalstatusquestioncanberecomputedtoallowforaneasiertoseecomparisonofthosewhoareinanytypeofschoolversusnoschool,asshowninSlide15-23or(asshowninSlide

15-24)tocomparesizeofgroupswhoarecurrentlynotenrolled,enrolledinhighschoolandenrolledinanytypeofcollege.

B.ChecklistQuestions

Question4,achecklist,measuresindividuals’reactionstoadvertisingonBebo.Thefirststepinanalysiscalculatesthepercentageofrespondentswhochecked

eachitem.TableshowninSlide15-25providesthisdata,presentingdatainthesameorderasitemsappearedonquestionnaire.

Thetableaccuratelypresentsdata.But,becausethereisnologicalorganizationtoordering,itisdifficulttoseeoverallpatternofresponse.

? ThetableinSlide15-26fixestheproblembyorderingitemsbyabsolutelevelofselection,makingiteasiertoseewhichitemshaverelativelymoreorlessagreement.

Slide15-

27

Slide15-

28

Slide15-

29

Slide15-

30

? ThetableshowninSlide15-27alsofixesproblembygroupingpositiveandnegativeitemstogether,makingiteasiertoseewhichpositiveandnegativeitemshaverelativelymoreagreement.

Presentationofpercentagesforchecklistitemsisoftensufficientfordiscoveringoverallpatternsofresponse.Anadditionalcomputation,whichcanprovidedeeperinsightintorespondents’thoughtsandattitudes,requiresthat

theresearcher(1)manipulatetherawdatafile,andthen(2)examineandthenclassifyingeachrespondent’spatternofresponseaccordingtoapredeterminedsetofcriteria.

Givenbothpositiveandnegativeitemsonachecklist,foranyindividualoneofthreepatternsofresponseispossible:

? arespondentchecksonlypositiveoptions

? arespondentchecksonlynegativeoptions

? arespondentchecksbothpositiveandnegativeoptions

Distinguishingbetweenthesethreetypesofindividualsisimportant,becauseitallowstheresearchertoidentifytherelativepercentofrespondentswhoareentirelypositive,entirelynegativeorwhohavemixedfeelings.

SurveyQuestion5onthesurveyasksrespondentstochecktheirfeelingstowardadvertisingthatappearsonMyAOL’shomepage.Atfirstglance,responsestoBebo(frompriorquestion)andMyAOLappeartobesimilar,asshownintableinSlide15-29.

Withoutfurtheranalysis,theresearchermightconcludethatreactionstoadvertisingontwosites’homepagesareidentical.However,thisconclusionprovesfalsewhenyouexaminethepercentageofpeoplewhoareentirelypositive,entirelynegativeorhavemixedreactions.ThisanalysisshowninSlide15-30.

Analysisshowsthatreactionstoadvertisingonthetwositesareverydifferent.Thesameproportionofsamplehadentirelypositiveattitudestowardtheadvertising.But,mostoftheremainingreactionstoBeboweremixed,withafewindividualshavingentirelynegativeattitudes.ReactionstoMyAOLwereverydifferent,withfewindividualshavingmixedfeelingsandmostindividualshavingentirelynegativereactions.DataindicatesthatMyAOL’shomepageadvertisingisperceivedasbeingmuchmoreofaproblemthanBebo’shomepageadvertising.Theseinsightsintoconsumers’attitudesweremasked

byinitialanalysisofindividualitempercentages.

C.RankingandOtherOrdinalLevelQuestions

Question6isarankorderquestion.Onlypercentages(andmediansandmodes)

areappropriate.

Thefirststepinanalysisofrankingdataisthecreationofapercentagedistri-butionthatreportsthenumberofeachrankingassignedtoeachrankeditem.Inthiscasewewanttocounthowmany“1,”“2,”and“3”ranksweregiventoeachofsocialnetworkingsites.

Slide15-

31

Slide15-

32

Slide15-

33

Thetableshowsthat380peoplegaveBeboarankingof‘1',80peoplegaveitarankingof'2',and40peoplegaveitarankingof'3'.Sinceall500peopleinthesampleprovidedrankings,totalrankingsforeachsiteaddsuptoatotalof500.

Thenextstepinanalysisandpresentationofrankingdataistranslationofeachfrequencydistributionintoapercentagedistributionforeachrankeditem.Thisisdonebydividingeachnumber(ofresponses)inacolumnbythetotalresponsesforthecolumn.ShowninSlide15-32.Datainthistableisreadbothdown(bycolumn)andacross(byrow).

Readingfromtoptobottomofeachcolumnshowspatternofresponseforeachsiteindependentoftheothersites.

Readingacrosswithinarowcomparesacrosssites.

Oncethesummarychartiscreated,additionalsummarymeasurescanbecalculatedtosimplifydataandassistincommunicatingkeyfindings.AsshowninSlide15-33,youcanaddasummarylinetothetablewhichpresentstheresultsofacomputationthatsubtractsthepercentageofworstrankings(inthiscase‘3’rankings)fromthepercentageofbest(‘1’)rankingsinordertoobtainanoverallassessmentofpositivepreference.Newmeasureiscalled“NetBestRankings”The66%reportedforBebo,forexample,iscomputedby

subtractingthepercentageof“3”rankingsfromthepercentageof“1”rankings,inthiscase76-8.

SummarymeasureclearlyreinforcespriorconclusionthatBebohasmostpreferredhomepage:onlyBebohadmoreindividualsgivingita“best”versusa“worst”ranking.

D.RatingScalesandOtherIntervalLevelQuestions

Questions7to9areratingscales.Themostcommonwaytoreportintervalleveldataisthroughpercentsandaverages.

Slide15-

34

Slide15-

35

Slide15-

36

Slide15-

37

Slide15-

38

Thefirststepinanalysisiscalculationofpercentofrespondentsselectingeachresponseoption,andcalculationofanoverallaverageandstandarddeviation(asshowninSlide15-34).ThemajorityofrespondentswerepositivetowardBebo’shomepage.Theaverageisnearthevalueofthe“SlightlyAgree”responseoption.

Afterdetaileddatahasbeenreportedforeachresponseoption,datacanthenbesimplifiedtomakeunderlyingtrendseasiertosee.Oneapproachistocombinechoicesthataredirectionallyrelated,inthiscasepositive“agree”choicesandnegative“disagree”choices(asshowninSlide15-35).

Also,similartotheapproachtakenforrankingdata,youcansubtractpercentageofnegativesfrompositivesinordertoobtainameasureofoverallpositiveresponses.

Eachofpriorapproachesprovideaslightlydifferentlookattheresponsestoeachindividualquestion.Whenthereisasetofmultiple,relatedquestions,itisoftenusefultoprepareasummarytablethatpresentsthekeyfindings.The

tableshowninSlide15-37presentssummarymeasuresforBebowhenreportingtheresponsestoQuestions7to9.Thismakesitveryeasytoseerelativestrengthsandweaknesses.

E.ConstantSumandOtherRatioLevelQuestions

Question10isaconstantsumquestion.Percentagesandaverages,aswellasmediansandmodes,areappropriateforsummarizingresponsestothistypeofquestion.

Thefirststepiscreationofadistributionthatreportstheaveragenumberofpointsgiventoeachlisteditem.Distributionfor500respondentswhorankedimportanceofeachsiteattributeisshowninSlide15-38.Notethatinpreparationofthistable:(a)thesumaddstothenumberofpoints(100)notthenumberofrespondentsand(b)theitemsevaluatedhavebeenplacedindescendingorder(tomaketheunderlyingtrendeasiertosee)ratherthanorderinwhichitemsappearedonthequestionnaire.

Thetableshowsthattheabilitytopostpicturesandblogwereperceivedasthemostimportantbenefits,receivingnearlythesameamountofhighpoints.Abilitytopostvideosandinstantmessagewereseenastheleastimportant.Theresearchermightconcludethatthefocusoneitherpostingpicturesorbloggingcouldsafelybeselectedassiteattributethatshouldbepromoted.But,withoutadditionalanalyses,itwouldbepremature(andthuspotentiallydangerous)todoso.

Additionalanalysesmakecertainthataveragesdonotleadtoerroneous

conclusions.Theyrequireanexaminationofeachofthedistributions.Oneapproachgroupspointallocationsintolargergroups,notingthepercentageofthesamplethatfallsintoeachgroup.Whenperformingthisanalysis,focusshiftsfromtheaveragingofpointstoanexaminationofthesample’sallocationbehavior.Thiscancreatetwocategoriesofresponse:

? Extremelyimportant:Percentofsamplewhoallocatedmorethan75%

ofpointstoanattribute

? Extremelyunimportant:Percentofsamplewhoallocatedlessthan25%

ofpointstoanattribute

Slide15-

39

Slide15-

40

Slide15-

41

TheresultsofthisanalysisareshownintheSlide15-39,wherepercentagesreflectpercentageofrespondentsfallingintoeachcategory.

Thetableindicatesthatinspiteoftheroughlyequivalentaveragenumberofpointsgiventopostingpicturesandblogging,theunderlyingpatternofresponsepresentsaquitedifferentpicture.Respondentswereoverwhelmingly

infavorofpostingpictures,with74%ofthesamplegivingit75ormorepoints

andonly15%ofthesamplegivingit25orfewerpoints.Thepaatternforbloggingisquitedifferent,withrelativelyequalnumbersofpeoplefeelingthatitwasimportantandunimportant.Postingpicturesisclearlythemorepowerfulbenefit.

IV.TheImportanceofSubgroupAnalysis

Priorapproachesentailedexamining,summarizing,andreportingtheresultsforthetotalsample.Thisisoftenusefultoexamineandcompareresponsesofsmallergroupsinsample.

Analysisofsmallersubgroupsoftenprovidesdeeperinsightsintodatatrendsandprovidesabasisfordrawingimportantmeaningfromthedata.Slide15-40illustrateshowyoungerindividualsareextremelynegativewhileolderindividualsareextremelypositive.

Youcantakesubgroupanalysestoanylevelofdetail.Prioranalysiscomparedtwoagegroups.But,youcanzoominevenfurthertodetermineifpositivereactionsof30to39yearoldsvaryasafunctionofeducation.Inthismoredetailedanalysis,youcancompareopinionsofindividualsaged30to39withacollegeeducationandthoseaged30to39withoutacollegeeducation.TableshowninSlide15-41accomplishesthis,whereitappearsthat(withinthisagegroup)lesswell-educatedindividualsaremorepositivethanthosewithacollegeeducation.

Guidelinesforsubgroupanalysis:

Slide15-

42

Slide15-

43

Slide15-

44

? Thinkaboutthedatafromtheresearchend-user’sperspective.

Whatquestionsaretheylikelytoask?Whatfindingswillthey

findmostintriguing?Whatfindingsaretheyleastlikelyto

accept?Whatdepthofdetailisrequiredformanagementtomakefullyinformeddecisions?Oncequestionsareanswered,thenconductappropriatesubgroupanalysestoprovidedetails.

? Lookfortheunexpected.Becreativeinquestionsasked;don’tjustfocusontheobviousoranticipatedfindings.Trytofindvaluablemeaninginthedatabydiscoveringunexpectedtrendsandrelationshipsinsubgroupanalyses.

? Beopen-minded.Whenyouseesomethingunexpected,takeastepbackandaskwhyresultsmightbeastheyare.Asknewquestionsandreformulateoldquestions.Answerthesequestionswiththeappropriatesubgroupanalyses.

? Besensitivetosamplesize.Sizeofindividualsubgroupstypicallydeclinesasgroupsbecomemorenarrowlydefined.Bewaryofanalyzinggroupsthatcontainsmallnumberofindividuals.

V.DataAnalysisinAction

A.Situation

Researchpriortoproductiontoidentifycommercial’sstrengthsandweaknesses.Twentyrespondentsaremembersoftheproduct’stargetaudience.Thesurveyusesfivemeasurestotestreactionstothecommercial.Thedatacollectedbythesemeasures,includingdemographicinformation,are:

? Respondentage

? Respondentgender

? Ratingofpurchaseintentafterseeingthecommercial

? Ratingofcommercialmessagebelievability

? Ratingofcommercialmessageuniqueness

? Reactionstothecommercial(twochecklistitems)

Alldatawerecollectedafteranindividualviewedtestcommercial.

B.TheAnalysis

Datacollectedfrom20respondentsshowninSlide15-44where:

? Ageiscoded“1”or“2,”wherea“1”represents25to34yearsoldand

“2”represents35to49yearsold

? Genderiscoded“1”or“2,”where“1”representsawomananda“2”

representsaman.

? Ratingsofpurchaseintent,messagebelievabilityandmessageuniquenessareallonafive-pointscalewherehighernumbersindicatemorepositiveresponses.

? Checklistreactionsusean“x”toindicateanitemwaschecked.A

blankspaceindicatesthatanitemwasnotchecked.

Onewaytoapproachdataanalysisistoimaginetypesofquestionsendusersofresearchwillaskandthencarryoutnecessarycomputationstoanswerthesequestions.Sincethisisacommercialtest,mostfundamentalquestionlikelytobeaskedbythosewhodesignedthecommercialis:Howdidthecommercialdo?

Theanswertothisquestionrequiresanexaminationofallofthemeasuresforthetotalsample,usingdescriptivestatisticstosummarizeresults.Thisprovidesafirstlookathowallrespondentsreactedtocommercial.Appropriatedescriptivestatisticsare:

Slide15-

45

Slide15-

46

? Purchaseintent(interval):Mean,percentage,median,mode

? Believability(interval):Mean,percentage,median,mode

? Uniqueness(interval):Mean,percentage,median,mode

? Checklistitems(nominal):Percentage

Withthisinmind,thebottomtworowsofpriortable(nowshowninSlide15-

46)presentappropriatedescriptivestatisticsforalltwentyindividuals.

Averageshavebeencalculatedonlyforintervallevelmeasures,notforgenderandage,whichwillbeusedasclassificationvariables.

Averagesandpercentagesindicatethatthecommercialperformedfairlywell.Purchaseintentwasatmidpointofscale,messagebelievabilitywashigh,andmessageuniquenesswasatmidpointofscale.Mostrespondentsfoundthecommercialinterestingandlessthanone-thirdfounditconfusing.

Atthispointitwouldbeeasytothinkthatthejobisdone.But,thiswouldbewrongconclusion-notbecausecalculationswereincorrect,butbecauseanalysisfailedtoanticipateadditionalquestions.Areasonablenextquestiontoaskwouldbe:Didthecommercialworkequallywellamongallmembersofthetargetaudience?

Slide15-

47

Slide15-

48

Slide15-

ThetableinSlide15-47looksattheagesubgroupsdetermined.Itshowsoriginaldatasetdividedintotwoagegroups.Allinyoungergroup(thosecodedwitha“1”forage)havebeenplacedtogetherinonegroupwhileallinoldergroup(thosecodedwitha“2”forage)havealsobeengroupedtogether.Descriptivestatisticsarecalculatedforeachgroup.Thereappearstobenodifferenceincommercialperformancebyagegroup:

? Averagesforratingscalesareidentical(purchaseintentforbothagegroupsequals3.1;messagebelievabilityforbothagegroupsequals4.2;messageuniquenessforbothagegroupsequals3.1).

? Percentagesofrespondentsineachgroupwhosaythecommercialisinteresting(80%)and“confusing(30%)arethesame.

Itthereforeappearsthatthecommercia