欢迎来到天天文库
浏览记录
ID:45591841
大小:65.09 KB
页数:14页
时间:2019-11-15
《病例队列试验研究的经验似然方法》由会员上传分享,免费在线阅读,更多相关内容在工程资料-天天文库。
1、病例队列试验研究的经验似然方法郁文复旦大学管理学院统计学系,上海200433摘要:病例•队列抽样设ilM-种广泛适用于大型流行病学跟踪研究中的降低抽样成本的抽样方式。本文在Cox模型假设下,提出了一种分析病例-队列设计抽样数据的经验似然方法,所构造的回归系数的对数经验似然比统计量渐近服从卡方分布。本文提出的经验似然方法可对回归系数的任意线性组合进行统计推断,并且避免了复杂的估计量方差估计,对于现存的Wald型推断方法提供了一种有效的备选。最后使用模拟计算与一个实例来展示所提出方法的表现。关键词:病例■队列设计,Cox模型,经验似然,似然比检验,Wilks定理中图分类号:0212
2、Empiricallikelihoodmethodforcase-cohortstudiesYUWenDepartmentofStatistics,SchoolofManagement,FudanUniversity,Shanghai200433Abstract:Thecase-cohortdesigniswidelyusedinlargeepidemiologicalstudiesandpreventiontrialsforcostreduction.Inthispaper,weproposeanempiricallikelihoodbasedinferentialproce
3、dureforthecase-cohortdesignunderCoxmodel.Theproposedlog-empiricallikelihoodratioteststatisticsfortheregressionparametersareshowntopossesschi-squaredlimitingdistributions.Theprofileempiricallikelihoodcanbeappliedtomakeinferencesaboutlinearcombinationsoftheentireparametervector.Theproposedappr
4、oachwhichavoidsthecomplexvarianceestimationisanattractivealternativetotheexistingWald-typeinferentialprocedures・Simulationstudiesareconductedtoassessthefinitesampleperformancesoftheproposedinferentialprocedures・Arealexampleisalsoprovidedforillustration.Keywords:Case-cohort;Coxmodel;Empirical
5、likelihood;Likelihoodratiotest;Wilkstheorem0IntroductionInmanylargeepidemiologicalstudiesandpreventiontrials,thousandsofindividualswithrareeventsarefollowedformanyyears・Usuallyresourceconstraintsprohibitcollectingcovariatehistoriesonallcohortmembers.Thecase-cohortdesignproposedbyPrentice[1]
6、provides师类资助课题(项目徧号:20110071120023),国家自然科学基金资助课题(项n编号:11101091).作者简介:YuWen,male,associateprofessor,majorresearchdirection:semiparametricmodel,empiricallikelihood,survivalanalysis・aneffectivesolutionfortheproblem.UnderPrentice^design,covariatesarecollectedonlyforthecases,i.e.,thosewhoexperie
7、ncetheeventofinterestduringthefollow-upperiod,andforasmallsimplerandomsampleoraBernoullisamplefromtheentirecohort,whichisknowntobeasubcohort・Thedesignreducesthecostsignificantlyandmakesthelargecohortstudiesfeasibletoconductfortheresearchers・Thecase
此文档下载收益归作者所有