实证研究——实验总结

总结（1）实验法Experimentaldesignisaprocedurefordevisinganexperimentalsettingsuchthatachangeinadependentvariablemaybeattributedsolelytothechangeinanindependentvariable.定义：是一种通过系统操作一个或多个自变量以确定此变化是否对其他一个或者多个因素造成影响的方法。Ch52自变量Independentvariablesarethosevariableswhichtheresearcherhascontroloverandwishestomanipulate.Categoryvs.Continuousvariables(problemswithmediansplit)Forexample:levelofadexpenditure;typeofadappeal;price;productfeatures,etc.Ch53因变量Dependentvariablesarethosevariablesthatwehavelittleornodirectcontrolover,yetwehaveastronginterestin.Exampleswouldbecustomersatisfaction,attitude,purchaseintention,actualbehavior,choice,etc.Ch54额外变量Extraneousvariablesarethosevariablesthatmayhavesomeeffectonadependentvariableyetarenotindependentvariables.Extraneousvariablesmustbecontrolledthroughproperexperimentaldesign.Ch55额外变量常有的额外变量实验者效应：由研究者引起的对被试行为产生的无意的作用双盲程序：与被试接触的主试及被试本身都不知道研究假设噪音温度被试的个体差异额外变量控制额外变量的技术直接消除，如控制噪音保持额外变量恒定，比如性别对行为的影响，可以全采用男被试或者女被试测量被试属性，然后按属性匹配的方法随机化——最常见的控制额外变量的方法把它们作为自变量或者协变量实验法Doestheindependentvariablecausechangesinthedependentvariable?Ch58思考困难Choicedeferralorcompromisecause实验法必须对变量下一个操作定义实验法Characteristicsmanipulationofavariablesolvesthedirectionofcausalityproblemrandomlyassigntoconditionssolvesthe3rdvariableproblemCh510实验法操纵和选择自变量自变量的每个值被称为水平（level)要操纵变量，被试必须被分配到该变量的两个水平上（最简单实验：实验组vs控制组）与每个水平相联系的条件叫做该实验的一个处理（treatment)被试可以接受一个处理（被试者间设计）或所有处理（被试者内设计）如果实验做得不好,thengarbagein,garageout)实验法也会出现一些问题(1)Themanipulationsmaynotbestrongenoughtoallowforameaningfultestofthetheory;(2)Amanipulationdesignedtovaryoneparticularindependentvariablemayalsoaffectotherindependentvariablesinthestudy,(3)Otherlatentvariables(i.e.,confoundingvariables)maybeinadvertentlyaffectedbythemanipulations,Ch514实验法也会出现一些问题(4)Theexperimentalproceduresmightcreatedemandartifacts,(5)Thedesignofthemanipulationsmaysuggestthatthegeneralizabilityofthefindingstoreal-worldmarketingsituations,thatis,thestudy'sexternalvalidityishighlyquestionable.Ch515实验法的优缺点strengths(内在效度)allowsforcausalconclusionstobemade--besttestoftheoryWeaknesses(外在效度)notallquestionsareamenabletoexperimentsdemandcharacteristicsconcernsaboutgeneralizabilityCh516内部效度问题：如何严格地测试理论?ManipulationandconfoundingchecksDirectmeasureofprocessesMultiplestudiesusingdifferentmanipulationmethods,productcategories,etc.Ch5171.MANIPULATIONANDCONFOUNDINGCHECKSIndependentvariableDependentvariable(SourceCredibility)(BrandAttitude)Whatifthemanipulationsthemselvesareconfounded?Concrete,observableindependentvariables(e.g.,priceand/oradvertisingexpendituresindollars,color,numberofsalescalls).Unobservable,latentindependentvariables(e.g.,perceptionsofasalesperson'sexpertise,feararousal,groupcohesiveness).Ch518TIMINGOFMANIPULATIONANDCONFOUNDINGCHECKSPretestvs.MainExperimentMostlyatthepreteststageCostissmall.Includethesechecksinthemainexperimentcanpresentproblems.Inthepretest,manipulationcheckquestionsshouldbeusedimmediately.Youcanalsousequalitativetoolssuchasinterviewsandconcurrentverbalprotocoltofindsomepotentialconfoundingfactors.Ch519CHECKSWITHINTHEMAINEXPERIMENTPotentialproblems:YouneedtomeasureyourmainDVsfirsttoavoiddemandcharacteristics,fatigue,etc.However,themanipulationmayonlycauseatemporarychange.Themeasuredelaymaycausethesubstantialdissipationofchecks.MainDVsmaybiasestheirresponsestothosechecks.Ch520SCALESFORMANIPULATIONANDCONFOUNDINGCHECKSThemeasuresforthemanipulationandconfoundingchecksshouldbeconstructedwiththesamecareasthoseforthedependentvariables.Itisimportanttoestablishthatthemanipulationproducedalargeenoughvarianceintheintendedindependentvariabletoprovideforameaningfultestofthehypothesesofinterest.Developreliablemeasuresofthosechecks.Ch521ANALYSISOFMANIPULATIONANDCONFOUNDINGCHECKSWhentherearemultiplefactors,directionalTtestorsimpleone-wayANOVAmaynotbesufficient.Youneedtousethefull-factorialANOVA.Example:2(involvement:highvs.low)X2(emotion:fearvs.anger)onriskperception.OnesimpletestoftheconfoundingvariablesinvolvesenteringtheconfoundingchecksascovariatesintheANOVAmodelusedtoanalyzethedependentmeasures.Ch5222.DIRECTMEASUREOFPROCESSProcessmeasuresareimportantbecauseitdirectlytapsintothetheoreticalprocesses.Thesemeasuresincluderesponsetime,self-reportedrecallorthoughtlisting,andphysiologicalresponsessuchasheartrateandpupildilation.Ch5233.MULTIPLESTUDIESDifferentwaysofvaryingthelevelsofindependentvariables(e.g.,measuredvs.manipulation,traitvs.prime).Differentmethodsofmanipulation.Ch524方向的经验Alotofpretests.Stimuliarethekey.Putmoretimeonplanning.Noperfectdesignorstudy(tradeoff).Ch525外部效度问题：实验室实验有用吗？实验法的长处与局限长处：与相关研究相比，实验研究的魅力在于有能力确定因果关系局限做实验的人有不愉快的trade-off,控制越精细，结果使用范围越小？？外部效度问题：实验室实验有用吗？Question:Willtheresponsebethesameoutsidethelaboratory?Canwegeneralizeourresultstothepopulationofinterest?Labexperimentstendtobelowonexternalvaliditybecauseconditionsnotlikethe“realworld”MostinternalvaliditythreatsalsoimpactexternalvalidityCh527外部效度问题：实验室实验有用吗？Question:Willtheresponsebethesameoutsidethelaboratory?Selectivitybiasmaybeworsto