搜索
(510224)李丽霞郜艳晖周舒冬邹宗峰张瑛,,localscoring,,,,,,1985Stone(ad-ditivemodels),,,1990,HastieTibshirani,(generalizedadditivemodels)YX1,X2,Xp:E(Y|X1,X2,,Xp)=0+1X1+2X2++pXp0,1,p:E(Y|X1,X2,,Xp)=s0+s1(x1)+s2(x2)++sp(xp)si(.),i=1,2,,p,Esj(Xj)=0,si(.)backfitting1,2,g()=s0+s1(x1)+s2(x2)++sp(xp)=s0+pi=1si(Xi),=E(Y|X1,X2,,Xp),,si(.),,,YX,Y(randomcom-ponent)(additivecomponent)g(.)(linkfunction),Y,PoissonGamma,:fY(y;;)=expy-b()()+c(y,),,,,,,(sem-iparametricgeneralizedadditivemodels),:g()=s0+X+si(xi),si(.)localscoring[1,2],:(1):si=g(E(y)),s01=s02==s0p=0,m=0(2):zi,wi,zi,wi2,m-1i=s0+pj=1sm-1j(xij)m-1i=g-1(m-1i)wi=(i/i)2m-1V-1izi=m-1i+(yi-m-1i)(i/i)m-1zbackfittingsmj(.),Deviance,localscoringSAS90STATprocgam,997,1age(0:20;1:20~25;2:25~30;3:30~40;4:40),xz(0:;1:),rg(0:;1:),,2500glogistic,,,,logistic,xzrg,age243200762431xzrg01100010024111140010SAS:dataaa;inputagexzrgycount@@;procgamdata=aa;/*GAM*/modely=param(xzrg)spline(age)/method=GCVdist=binomial;/*xzrg,logistic;method=GCV:(generalizedcrossvalidation)*/freqcount;outputout=bbp;run;,2~4,,xz(t=8.49,P0.0001),rg(t=2.26,P=0.0237),(age)(t=-1.96,P=00504),GCV,,,age,2=181147,P=000081,:2tPIntercept-3.599480.32006-11.250.0001xz1.813890.213588.490.0001rg0.479330.211822.260.0237Linear(age)-0.282900.14455-1.960.05043(GCV)Spline(age)0.1339963.5865501.499278542PSpline(age)3.586556.15487518.11470.0008,,,,,3-51,,(con-curvity)[6],,,,,,,,,,,GeneralizedAdditiveModelsanditsApplicationLiLixia,GaoYanhui,ZhouShudong,etal.DepartmentofMedicalStatistics,GuangdongCollegeofPharmacy(510224),GuangzhouAbstractObjectiveToexploretheapplicationofGeneralizedAdditiveModelsinmedicalreserch.MethodsIthasbeenillustratedhowtouseGeneralizedAdditiveModelsandsasprogrambyanexamplethatisabouttheriskfactorofthepregnancyinducedhypertensionsyndrome.Re-sultsThismodelcanuncoverthenon-linearrelationshipbetweendiseaseandage.ConclusionItdemandsnoassumptionoflinearityandisapow-erfultoolthatcanflexiblyexploresthenon-linearrelationshipbetweeninde-pendentvariableanddependentvariable.KeywordsGeneralizedadditivemodels,Additivemodels,Localscoringalgorith1HastieT,TibshiraniR.GeneralizedAdditiveModels.ChapmanandHall,1990.2SASInstituteInc.SAS/STATUserGuide.Version9.Cary,NC:SASInstituteInc,1992.3DanielBerg.Bankruptcypredictionbygeneralizedadditivemodels.2005(statisticalresearchreport).4,,,...2005,22(4):252-255.5,,,..,2004,25(11):962-966.244ChineseJournalofHealthStatistics,Jun2007,Vol.24,No.3