SPSS混合线性模型.ppt

《SPSS混合线性模型.ppt》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、MixedAnalysisofVarianceModelswithSPSSRobertA.Yaffee,Ph.D.Statistics,SocialScience,andMappingGroupInformationTechnologyServices/AcademicComputingServicesOfficelocation:75ThirdAvenue,LevelC-3Phone:212-998-34021OutlineClassificationofEffectsRandomEffectsTwo-WayRandomLayoutSolutionsandestimates

2、GenerallinearmodelFixedEffectsModelsTheone-waylayoutMixedModeltheoryPropererrortermsTwo-waylayoutFull-factorialmodelContrastswithinteractiontermsGraphingInteractions2Outline-Cont’dRepeatedMeasuresANOVAAdvantagesofMixedModelsoverGLM.3DefinitionofMixedModelsbytheircomponenteffectsMixedModelsc

3、ontainbothfixedandrandomeffectsFixedEffects:factorsforwhichtheonlylevelsunderconsiderationarecontainedinthecodingofthoseeffectsRandomEffects:Factorsforwhichthelevelscontainedinthecodingofthosefactorsarearandomsampleofthetotalnumberoflevelsinthepopulationforthatfactor.4ExamplesofFixedandRand

4、omEffectsFixedeffect:Sexwherebothmaleandfemalegendersareincludedinthefactor,sex.Agegroup:MinorandAdultarebothincludedinthefactorofagegroupRandomeffect:Subject:thesampleisarandomsampleofthetargetpopulation5ClassificationofeffectsTherearemaineffects:LinearExplanatoryFactorsThereareinteraction

5、effects:Jointeffectsoverandabovethecomponentmaineffects.67ClassificationofEffects-cont’dHierarchicaldesignshavenestedeffects.Nestedeffectsarethosewithsubjectswithingroups.AnexamplewouldbepatientsnestedwithindoctorsanddoctorsnestedwithinhospitalsThiscouldbeexpressedbypatients(doctors)doctors

6、(hospitals)89BetweenandWithin-SubjecteffectsSucheffectsmaysometimesbefixedorrandom.TheirclassificationdependsontheexperimentaldesignBetween-subjectseffectsarethosewhoareinonegrouporanotherbutnotinboth.Experimentalgroupisafixedeffectbecausethemanagerisconsideringonlythosegroupsinhisexperimen

7、t.Onegroupistheexperimentalgroupandtheotheristhecontrolgroup.Therefore,thisgroupingfactorisabetween-subjecteffect.Within-subjecteffectsareexperiencedbysubjectsrepeatedlyovertime.Trialisarandomeffectwhenthereareseveraltrialsintherepeatedmeasuresdesign;a