搜索
mynotesformeta-analysis•1.BasicinformationAresearchsynthesiscanbedefinedastheconjunctionofaparticularsetofliteraturereviewcharacteristics;andResearchsynthesesattempttointegrateempiricalresearchforthepurposeofcreatinggeneralizationsResearchsynthesiscanbebothqualitativeandquantitative.Generally,therearethreemethodsforquantitativeresearchsynthesis:Methodsforvote-counting,methodsforcombiningprobabilities,andmodernmeta-analyticmethods.Modernmeta-analytictechniquesarecomputationallydifferentfromcombinedprobabilitymethodsinthattheycombinethemeasuresofeffectsfromindividualstudiesintoanestimateoftheoverallstrengthoftheeffect,andthenusethistodeterminesignificance.Aconceptcrucialtomodernmeta-analysisisthatoftheeffectsize.Unlikemoretraditionalstatisticalapproaches(e.g.,ANOVA),effectsizesfromindividualstudiesinameta-analysisshouldbecombinedusingweightedstatisticalmodels.Severalmeta-analyticalmodelsexist,allofwhichareweightedlinearmodelswhoseweightsarebasedonthestudies’samplingvariances.Thestepsofameta-analysis•1)Definequestion(s)tobeanswered2)Conductaliteraturesearchandcollectdata3)Chooseanappropriateeffectsizeandanalyticmodel(s)4)Calculateeffectsizes5)Explorestatisticalproperties/characteristicsofeffectsizes6)Conductsummaryanalyses7)ExamineandinterpretresultsMetaWinisacommonstatisticalsoftwareformeta-analysisanditcanbeusedtoperformsteps4through6above.3.Howtodoameta-analysisusingMetaWin•Taketheeffectsizeofresponseratioforexample.1)Thedatayouneedtocollectinclude:XcXe,SecSee,NcNe,andXcisthemeanofthecontrolXeisthemeanofthetreatmentSecisthestandarddeviationofthecontrolSeeisthestandarddeviationofthetreatmentNcisthesamplesizeofthecontrolNeisthesamplesizeofthetreatmentLikethisUsethistochangecolumnformattonumerictypeAsabove,thevalueofeffectsizeis1.2324,andthevalueofBootstrapCIisfrom1.1534to1.3145.IfthevalueofBootstrapCIoverlaps1,theresponseofthevariabletothetreatmentisnotsignificant;Ifboththeendsarehigherthan1,theresponseispositive;ifboththeendsarelowerthan1,theresponseisnegative.Ifthedataaredividedintodifferentgroups,heterogeneitybetweengroups(b)isneededtobecalculatedasfollowsIfthep(Chi-Square)valueofbishigherthan0.05,theheterogeneitybetweengroupsisnotsignificant;Ifthep(Chi-Square)valueofbislowerthan0.05,theheterogeneitybetweengroupsissignificant.YourgroupOthercalculationprocess,youshouldreadtwoliteratures,andsomepapersWhichisaboutmetaanalysis.