lecture5缺失值处理策略

lecture5缺失值处理策略

ID:36202159

大小:2.09 MB

页数:55页

时间:2019-05-07

lecture5缺失值处理策略_第1页
lecture5缺失值处理策略_第2页
lecture5缺失值处理策略_第3页
lecture5缺失值处理策略_第4页
lecture5缺失值处理策略_第5页
资源描述:

《lecture5缺失值处理策略》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库

1、1Lecture5 IncompletedataZiadTaibBiostatistics,AZMay3,2011OutlineoftheproblemMissingvaluesinlongitudinaltrialsisabigissueFirstaimshouldbetoreduceproportionEthicsdictatethatitcan’tbeavoidedThereisnomagicmethodtofixitMagnitudeofproblemvariesacrossareas8-weekdepressiontrial:25%−50%maydropoutbyfinalv

2、isit12-weekasthmatrial:maybeonly5%−10%2DateName,department3OutlineofthelecturePartI:MissingdataPartII:MultipleimputationExample:Theanalgesictrial45DateName,department6PartI:MissingdataInrealdatasets,like,e.g.,surveysandclinicaltrials,itisquitecommontohaveobservationswithmissingvaluesforoneormore

3、inputfeatures.Thefirstissueindealingwiththeproblemisdeterminingwhetherthemissingdatamechanismhasdistortedtheobserveddata.LittleandRubin(1987)andRubin(1987)distinguishbetweenbasicallythreemissingdatamechanisms.Dataaresaidtobemissingatrandom(MAR)ifthemechanismresultinginitsomissionisindependentofi

4、ts(unobserved)value.Ifitsomissionisalsoindependentoftheobservedvalues,thenthemissingnessprocessissaidtobemissingcompletelyatrandom(MCAR).Inanyothercasetheprocessismissingnotatrandom(MNAR),i.e.,themissingnessprocessdependsontheunobservedvalues.http://www.emea.europa.eu/pdfs/human/ewp/177699EN.pdf

5、1.Introductiontomissingdata??????VariablesCases??=missing7Whatismissingdata?Themissingnesshidesarealvaluethatisusefulforanalysispurposes.Surveyquestions:WhatisyourtotalannualincomeforFY2008?Whoareyouvotingforinthe2009electionfortheEuropeanparlament?8Whatismissingdata?Clinicaltrials:StartFinishce

6、nsoredatthispointintimetime9MissingnessItmatterswhydataaremissing.Supposeyouaremodellingweight(Y)asafunctionofsex(X).Somerespondentswouldn'tdisclosetheirweight,soyouaremissingsomevaluesforY.Therearethreepossiblemechanismsforthenondisclosure:Theremaybenoparticularreasonwhysomerespondentstoldyouth

7、eirweightsandothersdidn't.Thatis,theprobabilitythatYismissingmayhasnorelationshiptoXorY.InthiscaseourdataismissingcompletelyatrandomOnesexmaybelesslikelytodiscloseitsweight.Thatis,theprobabilitythatYismissingdependsonlyonthe

当前文档最多预览五页,下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页,下载文档查看全文
温馨提示:
1. 部分包含数学公式或PPT动画的文件,查看预览时可能会显示错乱或异常,文件下载后无此问题,请放心下载。
2. 本文档由用户上传,版权归属用户,天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容,确认文档内容符合您的需求后进行下载,若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误,付费完成后未能成功下载的用户请联系客服处理。