DefiningandCollectingDataChapter1ObjectivesInthischapteryoulearn:

Tounderstandissuesthatarisewhendefiningvariables.HowtodefinevariablesHowtocollectdataToidentifydifferentwaystocollectasampleUnderstandthetypesofsurveyerrorsClassifyingVariablesByTypeCategorical(qualitative)variablestakecategoriesastheirvaluessuchas“yes”,“no”,or“blue”,“brown”,“green”.Numerical(quantitative)variableshavevaluesthatrepresentacountedormeasuredquantity.DiscretevariablesarisefromacountingprocessContinuousvariablesarisefromameasuringprocessDCOVAExamplesofTypesofVariablesDCOVAQuestionResponsesVariableTypeDoyouhaveaFacebookprofile?YesorNoCategorical(Qualitative)Howmanytextmessageshaveyousentinthepastthreedays?---------------Numerical(discrete)Howlongdidthemobileappupdatetaketodownload?---------------Numerical(continuous)TypesofVariablesVariablesCategoricalNumerical

DiscreteContinuousExamples:MaritalStatusPoliticalPartyEyeColor

(Definedcategories)Examples:NumberofChildrenDefectsperhour

(Counteditems)Examples:WeightVoltage

(Measuredcharacteristics)DCOVACollectingDataCorrectlyIsACriticalTaskNeedtoavoiddataflawedbybiases,ambiguities,orothertypesoferrors.Resultsfromflaweddatawillbesuspectorinerror.Eventhemostsophisticatedstatisticalmethodsarenotveryusefulwhenthedataisflawed.DCOVADevelopingOperationalDefinitionsIsCrucialToAvoidConfusion/ErrorsAnoperationaldefinitionisaclearandprecisestatementthatprovidesacommonunderstandingofmeaningIntheabsenceofanoperationaldefinitionmiscommunicationsanderrorsarelikelytooccur.Arrivingatoperationaldefinition(s)isakeypartoftheDefinestepofDCOVADCOVAEstablishingABusinessObjectiveFocusesDataCollectionExamplesOfBusinessObjectives:Amarketingresearchanalystneedstoassesstheeffectivenessofanewtelevisionadvertisement.Apharmaceuticalmanufacturerneedstodeterminewhetheranewdrugismoreeffectivethanthosecurrentlyinuse.Anoperationsmanagerwantstomonitoramanufacturingprocesstofindoutwhetherthequalityoftheproductbeingmanufacturedisconformingtocompanystandards.Anauditorwantstoreviewthefinancialtransactionsofacompanyinordertodeterminewhetherthecompanyisincompliancewithgenerallyacceptedaccountingprinciples.DCOVASourcesofDataPrimarySources:ThedatacollectoristheoneusingthedataforanalysisDatafromapoliticalsurveyDatacollectedfromanexperimentObserveddataSecondarySources:ThepersonperformingdataanalysisisnotthedatacollectorAnalyzingcensusdataExaminingdatafromprintjournalsordatapublishedontheinternet.DCOVASourcesofdatafallintofivecategoriesDatadistributedbyanorganizationoranindividualTheoutcomesofadesignedexperimentTheresponsesfromasurveyTheresultsofconductinganobservationalstudyDatacollectedbyongoingbusinessactivitiesDCOVAExamplesOfDataDistributedByOrganizationsorIndividualsFinancialdataonacompanyprovidedbyinvestmentservices.Industryormarketdatafrommarketresearchfirmsandtradeassociations.Stockprices,weatherconditions,andsportsstatisticsindailynewspapers.DCOVAExamplesofDataFromADesignedExperimentConsumertestingofdifferentversionsofaproducttohelpdeterminewhichproductshouldbepursuedfurther.Materialtestingtodeterminewhichsupplier’smaterialshouldbeusedinaproduct.Markettestingonalternativeproductpromotionstodeterminewhichpromotiontousemorebroadly.DCOVAExamplesofSurveyDataAsurveyaskingpeoplewhichlaundrydetergenthasthebeststain-removingabilitiesPoliticalpollsofregisteredvotersduringpoliticalcampaigns.Peoplebeingsurveyedtodeterminetheirsatisfactionwitharecentproductorserviceexperience.DCOVAExamplesofDataCollectedFromObservationalStudiesMarketresearchersutilizingfocusgroupstoelicitunstructuredresponsestoopen-endedquestions.Measuringthetimeittakesforcustomerstobeservedinafastfoodestablishment.Measuringthevolumeoftrafficthroughanintersectiontodetermineifsomeformofadvertisingattheintersectionisjustified.DCOVAExamplesofDataCollectedFromOngoingBusinessActivitiesAbankstudiesyearsoffinancialtransactionstohelpthemidentifypatternsoffraud.EconomistsutilizedataonsearchesdoneviaGoogletohelpforecastfutureeconomicconditions.Marketingcompaniesusetrackingdatatoevaluatetheeffectivenessofawebsite.DCOVADataIsCollectedFromEitherAPopulationorASamplePOPULATIONApopulationconsistsofalltheitemsorindividualsaboutwhichyouwanttodrawaconclusion.Thepopulationisthe“l(fā)argegroup”SAMPLEAsampleistheportionofapopulationselectedforanalysis.Thesampleisthe“smallgroup”DCOVAPopulationvs.SamplePopulationSampleAlltheitemsorindividualsaboutwhichyouwanttodrawconclusion(s)AportionofthepopulationofitemsorindividualsDCOVACollectingDataViaSamplingIsUsedWhenSelectingASampleIsLesstimeconsumingthanselectingeveryiteminthepopulation.Lesscostlythanselectingeveryiteminthepopulation.Lesscumbersomeandmorepracticalthananalyzingtheentirepopulation.DCOVAThingsToConsider/DealWithInPotentialSourcesOfDataIsthesourceofdatastructuredorunstructured?Howiselectronicdataformatted?Howisdataencoded?DCOVAStructuredDataFollowsAnOrganizingPrinciple&UnstructuredDataDoesNotAStockTickerProvidesStructuredData:Thestocktickerrepeatedlyreportsacompanyname,thenumberofshareslasttraded,thebidprice,andthepercentchangeinthestockprice.Duetotheirinherentstructure,datafromtablesandformsarestructureddata.E-mailsfromfivepeopleconcerningstocktradesisanexampleofunstructureddata.Inthesee-mailsyoucannotcountontheinformationbeingsharedinaspecificorderorformat.ThisbookdealsexclusivelywithstructureddataDCOVAAllOfTheMethodsInThisBookDealWithStructuredDataTousethetechniquesinthisbookonunstructureddatayouneedtoconverttheunstructuredintostructureddata.Formanyofthequestionsyoumightwanttoanswer,thestartingpointcan/willbetabulardata.DCOVADataCanBeFormattedand/orEncodedInMoreThanOneWaySomeelectronicformatsaremorereadilyusablethanothers.Differentencodingscanimpacttheprecisionofnumericalvariablesandcanalsoimpactdatacompatibility.Asyouidentifyandchoosesourcesofdatayouneedtoconsider/dealwiththeseissuesDCOVADataCleaningIsOftenANecessaryActivityWhenCollectingDataOftenfind“irregularities”inthedataTypographicalordataentryerrorsValuesthatareimpossibleorundefinedMissingvaluesOutliersWhenfoundtheseirregularitiesshouldbereviewed/addressedBothExcel&MinitabcanbeusedtoaddressirregularitiesDCOVAAfterCollectionItIsOftenHelpfulToRecodeSomeVariablesRecodingavariablecaneithersupplementorreplacetheoriginalvariable.Recodingacategoricalvariableinvolvesredefiningcategories.Recodingaquantitativevariableinvolveschangingthisvariableintoacategoricalvariable.Whenrecodingbesurethatthenewcategoriesaremutuallyexclusive(categoriesdonotoverlap)andcollectivelyexhaustive(categoriescoverallpossiblevalues).DCOVAASamplingProcessBeginsWithASamplingFrameThesamplingframeisalistingofitemsthatmakeupthepopulationFramesaredatasourcessuchaspopulationlists,directories,ormapsInaccurateorbiasedresultscanresultifaframeexcludescertainportionsofthepopulationUsingdifferentframestogeneratedatacanleadtodissimilarconclusionsDCOVATypesofSamplesSamplesNon-ProbabilitySamplesJudgmentProbabilitySamplesSimpleRandomSystematicStratifiedClusterConvenienceDCOVATypesofSamples:

NonprobabilitySampleInanonprobabilitysample,itemsincludedarechosenwithoutregardtotheirprobabilityofoccurrence.Inconveniencesampling,itemsareselectedbasedonlyonthefactthattheyareeasy,inexpensive,orconvenienttosample.Inajudgmentsample,yougettheopinionsofpre-selectedexpertsinthesubjectmatter.

DCOVATypesofSamples:

ProbabilitySampleInaprobabilitysample,itemsinthesamplearechosenonthebasisofknownprobabilities.ProbabilitySamplesSimple

RandomSystematicStratifiedClusterDCOVAProbabilitySample:

SimpleRandomSampleEveryindividualoritemfromtheframehasanequalchanceofbeingselectedSelectionmaybewithreplacement(selectedindividualisreturnedtoframeforpossiblereselection)orwithoutreplacement(selectedindividualisn’treturnedtotheframe).Samplesobtainedfromtableofrandomnumbersorcomputerrandomnumbergenerators.DCOVASelectingaSimpleRandomSampleUsingARandomNumberTableSamplingFrameForPopulationWith850ItemsItemNameItem#BevR. 001UlanX. 002. .. .. .. .JoannP. 849PaulF. 850PortionOfARandomNumberTable492808892435779002838116307275111000234012860746979664489439098932399720048494208887208401TheFirst5ItemsinasimplerandomsampleItem#492Item#808Item#892--doesnotexistsoignoreItem#435Item#779Item#002DCOVADecideonsamplesize:nDivideframeofNindividualsintogroupsofkindividuals:k=N/nRandomlyselectoneindividualfromthe1stgroupSelecteverykthindividualthereafterProbabilitySample:

SystematicSampleN=40n=4k=10FirstGroupDCOVAProbabilitySample:

StratifiedSampleDividepopulationintotwoormoresubgroups(calledstrata)accordingtosomecommoncharacteristicAsimplerandomsampleisselectedfromeachsubgroup,withsamplesizesproportionaltostratasizesSamplesfromsubgroupsarecombinedintooneThisisacommontechniquewhensamplingpopulationofvoters,stratifyingacrossracialorsocio-economiclines.PopulationDividedinto4strataDCOVAProbabilitySample

ClusterSamplePopulationisdividedintoseveral“clusters,”eachrepresentativeofthepopulationAsimplerandomsampleofclustersisselectedAllitemsintheselectedclusterscanbeused,oritemscanbechosenfromaclusterusinganotherprobabilitysamplingtechniqueAcommonapplicationofclustersamplinginvolveselectionexitpolls,wherecertainelectiondistrictsareselectedandsampled.Populationdividedinto16clusters.RandomlyselectedclustersforsampleDCOVAProbabilitySample:

ComparingSamplingMethodsSimplerandomsampleandSystematicsampleSimpletouseMaynotbeagoodrepresentationofthepopulation’sunderlyingcharacteristicsStratifiedsampleEnsuresre