应用spss软件实现二分类重复测量的GEE及GLMM分析

2012;32(12)JSouthMedUnivSPSSGEEGLMMs5105152012-09-232010LC07020-61360867E-mail:[e-mail]1generalizedestimatingequations,GEEgeneralizedlinearmixedmodels,GLMMsGEEGLMMsSAS2-5SPSS19.0GEEGLMMs1GEESPSSGEE1986Liang6GEE7GEE8-91A442468521.1SPSS3220id44week5PASI20=1=1.2Analyze-GeneralizedLinearModelsGeneralizedEstimatingEquationsRepeated▶SubjectvariablesidSPSSGEEGLMMsSPSS19.0GEEGLMMsSPSS19.0GEEGLMMsSPSS19.0GEEGLMMsSPSSR195.1A1673-4254201212-1777-04doi:10.3969/j.issn.1673-4254.2012.12.020Abstract:ObjectiveToanalyzebinaryclassificationrepeatedmeasurementdatawithgeneralizedestimatingequations(GEE)andgeneralizedlinearmixedmodels(GLMMs)usingSPSS19.0.MethodsGEEandGLMMsmodelsweretestedusingbinaryclassificationrepeatedmeasurementdatasampleusingSPSS19.0.ResultsandConclusionComparedwithSAS,SPSS19.0allowedconvenientanalysisofcategoricalrepeatedmeasurementdatausingGEEandGLMMs.Keywords:binaryclassification;repeatedmeasurement;generalizedestimatingequations;generalizedlinearmixedmodel;SPSSAnalysisofbinaryclassificationrepeatedmeasurementdatawithGEEandGLMMsusingSPSSSoftwareANShengli,ZHANGYanhong,CHENZhengDepartmentofBio-Statistics,SchoolofPublicHealthandTropicalMedicine,SouthernMedicalUniversity,Guangzhou510515,China··1777JSouthMedUniv32▶Withinsubjectvariableweek▶WorkingcorrelationmatrixUnstructuredunstructuredTypeofModel▶BinaryResponseorEvents/TrialsData:BinarylogisticLogisticResponse▶DependentvariablePASI▶BinaryReferenceCategoryFirstlowestvalue0=Predictions▶FactorsweekModel▶ModelweekweekStatistics▶PrintGoodnessoffitstatistics/Modelsummarystatistics/Parameterestimates/Workingcorrelationmatrix101.31P=0.00520=82461.1900.9600.7500.554P0.00983Unstructured0.581~0.939QICQICQIC4QICGEE2GLMMsSPSSGLMMs11-122wistar-SD402020123456789101428d12132.1SPSS4480id40groupSourceInterceptWeekWaldχ²4.26914.833df14P0.0390.0051Tab1Testsofthemodeleffects2Tab.2ParameterestimatesModeltermInterceptWeek=0Week=2Week=4Week=6Week=8Coefficients-0.0911.1900.9600.7500.5540SE0.30180.31040.27720.24560.2127-95%CILower-0.6830.5810.4170.2690.137-Upper0.5011.7981.5031.2320.971-Waldχ²0.09114.68511.9919.3326.773-df11111-P0.7630.0000.0010.0020.009-3Tab.3WorkingcorrelationmatrixMeasrurmentWeek=0Week=2Week=4Week=6Week=8MeasrurmentWeek=01.0000.9390.8460.7670.581Week=20.9391.0000.9490.8600.652Week=40.8460.9491.0000.9550.724Week=60.7670.8600.9551.0000.799Week=80.5810.6520.7240.7991.0004QICTab.4QICofeachworkingcorrelationmatrixWorkingcorrelationmatrixIndependentAR1Exchangeable1-DependentUnstructuredQIC288.97256894205645288.9725689420565288.97256894205645288.97256894205566288.97256894205634··1778,.SPSSGEEGLMMs1212day12RI012.2Analyze-MixedModelsGeneralizedLinearMixedModelsFields&Effects▶TargetRIRIMoreCustomizereferencecatego-ryReferencevalue0▶TargetDistributionandRelationshipLinkwiththeLinearModelBinarylogisticregressionLogistic▶FixedEffectsgroupdaygroup*day▶RandomEffectsAddBlockidBuildOptions▶TestsofmodeleffectsandcoefficientsAssumemodelassumptionarecorrectUserobustestimationtohandleviolationsofmodelassumptionsrobustcovariancesModelOption▶EstimateMeansTermsgroupdayEstimateMeansDisplayestimatedmeansintermsofLinkfunctiontransformation2.3group*dayF=0.000P=1.0005F=35.332P0.001F=6.402P0.00160=groupt=5.944P0.001day28d3dt=-0.025P=0.9804P0.0023GLMMsSPSS7GLMMSF=0.000P=1.00055.0%12.1%F=35.332P=0.000F=6.402P=0.0004P≤0.00228d82.5χ2GLMMs4GEEGLMMs5Tab.5FixedeffectEffectGroupDaydf1111df2467467F35.3326.402P0.0000.000Modelterminterceptgroup=1group=2day=1day=2day=3day=4day=5day=6day=7day=8day=9day=10day=14day=28Coefficient1.4928.7380a-30.026-30.026-30.026-13.425-11.178-11.178-6.677-6.677-4.118-4.118-3.3430aSE0.9501.470-1206.5251206.5251206.5251.6751.5351.5351.2991.2991.0271.0271.064-t1.5705.944--0.025-0.025-0.025-8.015-7.282-7.282-5.142-5.142-3.412-3.412-3.143-P0.1170.000-0.9800.9800.9800.0000.0000.0000.0000.0000.0010.0010.002-95%CILower-0.3755.849--2400.915-2400.915-2400.915-16.716-14.194-14.194-9.229-9.229-6.489-6.489-5.433-Upper3.35911.627-2340.8642340.8642340.864-10.134-8.162-8.162-4.125-4.125-1.746-1.746-1.253-6Tab.6Fixedcoefficients··1779JSouthMedUniv32LiangGEEGLMMs14-15GEEGLMMsGEEGLMMs11GEEGLMMsSPSS19.0SASStata1.J.,2003,5(1):67-70.2,,.J..,2010,27(2):122-8.3,,,.J.,2011,18(4):294-6.4,,,.CTJ.,2010,10(21):4173-6.5,,,.logisticJ.,2010,17(4):308-11.6LiangKY,ZegerST.LongitudinaldataanalysisusinggeneralizedlinearmodelsJ.Biometrics,1986,73(1):13.7.D.:,2007:1-35.8.D.:,2007:1-47.9.GEEJ.,2008,25(4):369-72.10,,.SPSSJ.,2011,28(2):199-201.11,.J.,2007,24(5):486-7.12Verbeke,G,Molenberghs.Linearmixedmodelsforlongitudinaldata//SpringerSeriesinStatisticsM.NewYork:Springer,2000:30-50.13.D.:,2011.14,,,.J.,2010,27(5):464-9.15.J.,2012,19(1):14-7.7*Tab.7Effectcomparisonof2corneatransplantationmethodsDayday=1day=2day=3day=4day=5day=6day=7day=8day=9day=10day=14day=28TotalControln202020202020202020202020240RI%00.000.000.0420.0840.0840.01680.01680.020100.020100.020100.020100.013255.0Experimentn202020202020202020202020240RI%00.000.000.000.000.000.0315.0315.0315.0315.0420.01365.02912.1Sumn404040404040404040404040