版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领
文档简介
Presenter:ZHANGJinLIYue2010.10.25DataAnalysisandEthicsinBusinessResearchOutline1.PrinciplesandProceduresofExploratoryDataAnalysis2.MissingData3.LongitudinalandCross-sectionalTests4.DataAnalyticTrendsandDoctoralTraining5.EthicalIssuesinBusinessResearch6.FromtheEditors:anEthicalQuiz7.AcademyofManagementCodeofEthicalConduct1.1PrinciplesofExploratoryDataAnalysisWhatisEDA?AdetectiveworktodiscoverpatternsindatawithdifferentmethodsAnattitudetoexploredataconsistentlyandthoroughlyAplausiblestorytotellratherthandrawingconclusionsDefinitionExploratorydataanalysis(EDA)isawell-establishedstatisticaltraditionthatprovidesconceptualandcomputationaltoolsfordiscoveringpatternstofosterhypothesisdevelopmentandrefinement.EDA&CDAFormulationvs.Test(Generatingthedirectionvs.Testingthemyth)EDAhelpstointerpretresultsofCDAandmayrevealunexpectedormisunderstandingofpatternsinthedata.1.2ProceduresofEDABelief:GetarichdescriptionofdataUseofgraphics;Processofiterativemodelfit;ResidualAnalysisUnderstandtheContextInteractionofpriorknowledgeandpresentdataanalysisQuantitativeknowingdependsonqualitativeknowingFundamentalobservationswithstatisticalabilityUseGraphicRepresentationofData
“GraphicanalysisiscentraltoEDA”PortraynumerousdatavaluessimultaneouslySimpleplot;stem-and-leafplot;dotplot;boxplot;densitysmoothers;interactivecomputergraphics.1.2ProceduresofEDA——GraphicsGraphicRepresentationofData(1):SimplePlotLinearregressionStraightandsimpletounderstand1.2ProceduresofEDA——GraphicsGraphicRepresentationofData(2):BoxPlotMarkfirstandthirdquartilesOfferinformationaboutthelocationofkeyelementsinthedistributionandomitmoresubtledetailsUsefulwhenanumberofdistributionneedtobecompared1.2ProceduresofEDA——GraphicsGraphicRepresentationofData(3):DensitySmoothersDisclosesomehiddeninformationOverlayingdensityfunctionallowsdirectcomparisonofshape1.2ProceduresofEDADevelopModelsinanIterativeProcessandTentativeModelSpecificationandResidualAssessmentData=Fit+ResidualBuildaTwo-WayFitApplythefit-plus-residualframeworkiterativelyinbothdimensionsExample:LauverandJones’researchoncareer-self-efficacyCollectoccupationalpreferencedatafromethnicallydiversegroupsPercentageofstudentsconsideringthatcareeranoption1.2ProceduresofEDA——ModelFitBuildaTwo-WayFit:Example1.2ProceduresofEDA——ResidualAnalysisBuildaTwo-WayFit:Example1.2ProceduresofEDA——IterativeProcessUseRobustandResistantMethodsResistance;Smoothness;BreadthApproachestoassessresistance(breakdownpoint;trimean,etc.)PayAttentiontoOutliersCorrectmentalandcomputationalmodelsConsiderimportantdataandphenomenaoriginallyunanticipatedReexpresstheOriginalScalesTransformation(avoidradicalchangeofunderlyinginformation)Usualmethod(logarithmicscale;standardscore)PuttingitAllTogether
–
AnIterativeProcesstoFollowGraphics—Initialmodel,fit-plus-residual—ResidualAnalysis—Transformation,Outliers—Modificationofmodel(Iteratively)1.3ConclusionsEDAisFindpatternsinthedatatobuildrichmentalmodelsEspeciallyusefulwhenlittletheoreticalbackgroundavailablePromotetheorydevelopmentandtestingmendationsAWillingnesstoExplore;APhilosophyofYourOwn2.1FundamentalsWhatisMissingData?In
statistics,
missingvalues
occurwhenno
data
value
isstoredforthe
variable
inthecurrent
observation.(Wikipedia)TypeandPatternsofNonresponseUnitnonresponse←reweightingItemnonresponse←singleimputationWavenonresponse←MI(multipleimputation)&ML(maximumlikelihood)Univariatepattern(figure1)MonotonepatternArbitrarypattern2.1Fundamentals2.2OlderMethodsCaseDeletionDiscardunitwhoseinformationispleteSimplicity;GenerallyvalidonlyunderMCAR;inefficiencySingleImputationImputingunconditionalmeansImputingfromunconditionaldistributionsImputingconditionalmeansImputingfromaconditionaldistribution
2.3MLEstimationML(maximumlikelihood)MLestimatesarenotsubstantiallybiasedunderMCARorMARbutarequitebiasedunderMNARAssumingthesampleislargeenoughDependingontheparticularapplicationAssumingunderMARconditionSoftwareforMLEstimationBMDP;SPSS;EMCOV;NORM;SAS;Mplus;S-PLUS;LISREL;Amelia
2.4MultipleImputation2.4MultipleImputationFeaturesofMIRelyingonlarge-sampleapproximationsRequiringassumptionsaboutthedistributionofmissingnessMissingvaluesforeachparticipantarepredictedfromhisorherownobservedvaluesThejointrelationshipsamongthevariablesmustbeestimatedfromallavailabledataingroupMISoftwareNORMSASprocedure:PROCMIS-PLUSAmelia
3.1GravitationtoJobsCommensuratewithAbilityGravitationalHypothesisIndividuals,overthecourseoftheirlabormarketexperiences,willsortthemselvesintojobscompatiblewiththeirinterests,valuesandabilities.Goodperson-jobfitConceptsofFitIndividual’sbeliefandorganization’scultureIndividual’sabilityandabilityrequirementsforjobTwoTestsDirectionswithDifferentDatabaseLongitudinal–individual–directtestCross-sectional–job–indirecttest3.2Study1——LongitudinalTestHypothesisOvertime,lowerabilitypeoplewillgravitateintolowercomplexityjobsandhigherabilitypeoplewillgravitateintohighercomplexityjobsParticipantsDatafromNLSYdatabase;asampleof3887participantsValidscoresforASVABsubtest;occupationcodesVariablesAge(controlvariable)Cognitiveability(“g”fromASVABsubtestscores)Jobcomplexityin1982&1987OAPMap:sortjobsin13categoriesanddifferentiatebetweenjobsonthebasisofcognitiveabilityrequiredtoperformthejobto10levels3.2Study1——LongitudinalTestsOAPMap3.2Study1——LongitudinalTestsResultsSupportGravitationalHypothesis(Table2)IndividualsmovinglowerinthehierarchyovertimeshouldhavelowerabilityscoresthanthosewhoremainatthesamelevelorthosewhoproceedupwardThosemovinghigherhavethehighestmeangscoresCognitiveability(gscore)isasignificantpredictorofOAPmap(Table3)3.3Study2——Cross-sectionalTestsHypothesisAmoreexperiencedgroupofemployeesinaparticularjobwillexhibitlessvariabilityincognitiveabilitythanalessexperiencedgroupParticipantsDatafromUSES;asampleof60job-firmcombinationfor6051participantsValiddataforbothfirmandjobexperience;GATBscoresVariablesCognitiveability(GATBabilityscores)Firmandjobexperience(USES,self-reported)Jobcomplexity(5-categorysystemdevelopedbyJohnHuster)ResultBothfirmandjobexperiencearesignificantlyrelatedtovarianceofcognitiveabilityLess-experiencedgrouptendtohavelargervariability3.4ConclusionGravitationtoJobsCommensurateAbilityIndividualwithhighercognitiveabilitymoveintojobsrequiringmorecognitiveability.Groupshigherinbothfirmandjobexperiencehavesmallervariance.
Whatcouldwelearnfromthispaper?Twodifferentapproachestotesthypothesis:longitudinalandcross-sectionalDirectandindirecttestsDifferentdatabasetouse4.1DataAnalyticTrendsandTraininginStrategicManagementHittetal.’sassertion“Strategicmanagementresearchismovingbeyondcross-sectional,multipleregressionapproachestomethodsmoreattunedtothespecificproblemsandissueslikelytoinfluencestrategyresearch,suchasnetworkanalysis,eventstudies,andPoisson/negativebinomialregression”Doctoralstudentsshouldbetrainedincertainspecializedmethodratherthantraditionalmethod
ATwo-studyDesignTracktrendsintheuseofdataanalytictechniquesUnderstandthelevelofmasteryrecentdoctoralgraduatespossesswithbothtraditionalandspecializedmethods4.2Study1——DataAnalyticTrendsSampleandDataArticlespublishedinStrategicManagementJournal(SMJ)from1980to2001Asampleof297presentedoriginalempiricalstudiesCodeandgroupanalyticmethodsindifferentcategorizations4.2Study1——DataAnalyticTrends4.2Study1——DataAnalyticTrendsSampleandDataArticlespublishedinStrategicManagementJournal(SMJ)from1980to2001Asampleof297presentedoriginalempiricalstudiesCodeandgroupanalyticmethodsindifferentcategorizationsResultsBasictechniquesfalloutoffavor(e.g.testofmeans)GLMremainthedominantgroupoftechniques(e.g.multipleregressionandhierarchicalregression)Specializedtechniquesgrowinuse4.3Study2——PhDTrainingSampleandData77strategicmanagementPhDresearchers“Whenyouleftgraduateschool,howcompetentwereyouwitheachmethod?”/“Towhatextentareyoucompetentnowwiththesemethods?”Collectanswers(scales1-5)from1996to20014.3Study2——PhDTraining4.3Study2——PhDTrainingSampleandData77strategicmanagementPhDresearchersbetween1996to2001“Whenyouleftgraduateschool,howcompetentwereyouwitheachmethod?”/“Towhatextentareyoucompetentnowwiththesemethods?”Collectanswers(scales1-5)from1996to2001ResultsTraditionaltechniques:welltrainedSpecializedtechniques:notimprovedsincegraduationMorerecentgraduatesleftgraduateschoolpossessmoreconfidencesometechniquesthenearliergraduates4.4ConclusionDataAnalyticMethodTrendRiseofsomespecializedtechniquesRelianceofregressionmodelManyresearchersarenotfullyexploitingtheirdataDoctoralTrainingPhDgraduatesarecompetentwithacoresetoftechniquesInadequatetrainingforvitalmethodsofcurrentandfutureknowledgedevelopmentDoctoralprogramsshouldworktoclosethegapbetweenwhatstudentsknowandwhattheyneedtoknow5.1RightsandObligationsoftheRespondentRightsoftheRespondentPrivacyBeinginformedObligationsoftheRespondentBeingtruthful5.2RightsandObligationsoftheClientSponsorRightsoftheClientSponsorPrivacyBeinginformedObligationsoftheClientSponsorObservinggeneralbusinessethicswhendealingwithresearchsuppliersAvoidingmisusingtheresearchfindingstosupportitsaimsRespectingresearchsubjects’privacyBeingopenaboutitsintentionsandbusinessproblems5.3RightsandObligationsoftheResearcherRightsoftheResearcherBeinginformedObligationsoftheResearcherAdheringtothepurposeoftheresearchMaintainingobjectivityAvoidingmisrepresentingresearchfindingsProtectingsubjectsandclients’righttoconfidentialityAvoidingshadingresearchconclusions6FromtheEditors:anEthicalQuizScenario1:PlagiarismWhatdoyouthinkaboutthefollowingbehaviors?ReusingadescriptionofasampleyouwroteforanotherpaperReusingadescriptionofascaleyouusedinanotherpaperThe“codeofethical”ofAOMAOMmembersexplicitlyciteothers’workorideas,includingtheirown,eveniftheworkorideasarenotquotedverbatimorparaphrased.Scenario2:Data(Re)useThe“codeofethical”ofAOMWhenAOMmemberspublishdataorfindingsthatoverlapwithworktheyhavepreviouslypublishedelsewhere,theycitethesepublications,andtheymustsendthepriorpublicationworktotheAOMjournaleditors.6FromtheEditors:anEthicalQuizScenario3:InstitutionalReviewBoard(IRB)The“codeofethical”ofAOMWhenAOMmembersconductresearch,theyshouldobtaintheinformedconsentoftheindividualsScenario4:CoauthorsWhatdoyouthinkaboutthefollowingbehaviors?SubmittingapapertoajournalorconferencewithoutallofthecoauthorsbeingawareofitAddingacoauthorwithoutgettingthepermissionofthosealreadyonthepaperThe“cod
温馨提示
- 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
- 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
- 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
- 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
- 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
- 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
- 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
最新文档
- 2024年度消防工程单项工程劳务供应协议2篇
- 2025年度太阳能热水系统安装劳务分包合同范本3篇
- 2025年度百货商场儿童游乐区安全协议3篇
- 2024年度文化教育机构代运营合作协议3篇
- 2024年物联网设备研发与销售合同
- 2025年度老旧小区水暖设施改造合同模板2篇
- 2024年版权转让合同(音乐作品)
- 2024年精装修住宅工程合同5篇
- 2024年环保复印纸采购协议
- 2025版智能物流系统合伙投资经营协议书2篇
- 粉末涂料有限公司危废库安全风险分级管控清单
- 750更换齿轮箱作业指导书
- 安全生产信息管理制度全
- 住宅物业危险源辨识评价表
- 世界主要国家洲别、名称、首都、代码、区号、时差汇总表
- 2023学年广东省广州市越秀区铁一中学九年级(上)物理期末试题及答案解析
- 《报告文学研究》(07562)自考考试复习题库(含答案)
- 安全操作规程
- 电源日常点检记录表
- 人教版小学三年级语文上册期末测试卷.及答题卡2
- 钢轨接头位置及接头联结形式
评论
0/150
提交评论