HLM-多层线性模型的原理及应用-雷雳

()JournalofCapitalNormalUniversity20022(SocialSciencesEdition)(145)*雷雳1张雷2(1.,100089;2.):本文对多层线性模型(HierarchicalLinearModels,HLM)的理论缘起应用范围以及其应用原理进行了阐述,在指出经典统计技术处理多层数据结构上的局限的同时,表明了多层线性模型在这方面的优越性本文最后对多层线性模型的效果及局限性进行了简要分析:多层数据;回归;线性模型;多层模型:G44:A:1004-9142(2002)02-0110-05:2001-12-12:(1968-),,,,,;,,,,,*:100089,83,[e-mail](HierarchicalLinearModels,HLM),,;,,,,,,,,ANOVA,,,,,,,,,,,110,,,,,,,??,,,,?,,,(independenceofobserva-tions),,,,,,80%90%,,,,,,,,,,,,,(disturbances)(variancecomponentmodels);,,,,,,(random-coefficientregressionmodels),,,,,,,,,111,,,,:,,(),,,,(),,,,,,,,,,,,,,,,,,:(1)();(2);(3),,,,,,,(meta-analysis),,,,,,,,,,,,,,,,,HLM,112,,,,2641,51,$$(VIC),$$(SHY)(MALE)(BADC)(BULT)(HARSHC),,;SHY,MALEBADCBULTHAR-SHC:(1)MALEBADC?MALEMALE(BADC),(MALE)(2)BULTHARSHC?BULTBULTSHY(HARSHC),SHY(3)MALEBADCBULTHARSHC?(4)MALEBADCBULTHARSHC,?:(1)(2),SHY(3)(4)0,1:SHY,:VIC=B0+B1(SHY)+r,=+SHY+()2,:B0=C00+u0,=+();B1=C10+u1,SHY=SHY+SHY;Var(u0)=S0,;Var(u1)=S1,;Cov(u0,u1)=S,;:B0=C00+C01(BADC)+C02(MALE)+u0,BADCBADC??B1=C10+C11(BULT)+C12(HARSHC)+u1,SHY(HARSHC)?SHY(BULT)?:Var(u0)=S0Var(u1)=S1,,?,,MALEBADCBULTHARSHC,HLM(1):1131FixedEffectCoefficientSET-ratiod.f.P-valueForINTRCPT1,B0INTRCPT2,G00-0.5290.173-31062480.004BADC,G010.4480.0469.750480.000MALE,G020.0100.0032.854480.007ForSHYslope,B1INTRCPT2,G100.3710.0419.153480.000BULT,G11-0.0860.042-2.077480.043HARSHC,G12-0.1820.082-2.223480.0312RandomEffectSDVarianceComponentdfCh-isquareP-valueINTRCPT1,U00.0240.0014825.68101500SHYslope,U10.2630.0694826314090.000leve-l1,R0.7520.562,VIC(BADC)BADC,,1,BULT,HARSHC,2,2,,,32u1(01263,p=.000),2SHY,,4HLMHLM:,,,,,,,,,,,;,,,,,(责任编辑:肖子)PrincipleandApplicationoftheHierarchicalLinearModelLeiLiZhangLei(Thedepartmentofpsychology,CapitalNormaluniversity,Beijing100089)Abstract:ThisarticleanalysesbrieflytheorytoHierarchicalLinearmodelsgenesis,rangeofapplication,andtheirapplicationprinciple.Keyword:Multileveldata;Regress;Linearmodel;Hierarchicalmodel114