Selectionbiasincase-controlstudies-Smu选择偏倚的病例对照研究-SMU

《Selectionbiasincase-controlstudies-Smu选择偏倚的病例对照研究-SMU》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、Bias,Confounding,andEffectModificationSTAT6395Spring2008FilardoandNgAnysystematicerrorinthedesignorconductofastudythatresultsinamistakenestimateoftheassociationbetweenanexposureandadiseaseBiasisoftenamajorprobleminobservationalepidemiologicstudiesBiasExample:anassociationbetweenanexposureanadisease

2、inwhichthetruerelativeriskis2.0Systematicerror(bias)isdifferentthanrandomerrorIfthedesignandconductofastudyareunbiased,andthereisnoconfounding,andwerepeatthestudyaninfinitenumberoftimes,themeanrelativeriskwillbe2.0,withtheindividualrelativerisksfromthedifferentstudiesfluctuatingaround2.0Systematice

3、rror(bias)isdifferentthanrandomerrorIfthedesignorconductofthestudyisbiased,andwerepeatthestudyaninfinitenumberoftimes,themeanrelativeriskwilldifferfrom2.0(forexample,itmaybe1.2),withtheindividualrelativerisksfromthedifferentstudiesfluctuatingaround1.2Systematicerror(bias)isdifferentthanrandomerrorD

4、uetorandomvariation,anassociationthatisfarfromthetruthcanbeobservedinanunbiasedstudy,butitusuallywon’tbe.Systematicerror(bias)isdifferentthanrandomerrorDuetorandomvariation,thetrueassociationcanbeobservedinabiasedstudy,butitusuallywon’tbeSystematicerror(bias)isdifferentthanrandomerrorStatisticalsig

5、nificancedoesnotprotectagainstbiasSystematicerror(bias)isdifferentthanrandomerrorSelectionbiasInformationbiasTwomajorcategoriesofbiasErrorthatresultsfromcriteriaorproceduresusedtoselectstudysubjectsorfromfactorsthatinfluencestudyparticipation.Withselectionbias,therelationbetweenexposureanddiseaseis

6、differentforthosewhoareselectedforandparticipateinthestudyandthosewhoshouldbetheoreticallyeligibletoparticipate.SelectionbiasSelectionbiascanoccurasaresultof:IncorrectselectioncriteriaforstudysubjectsDifferencesincharacteristicsbetweeneligiblesubjectswhoagreetoparticipateandeligiblesubjectswhodonot

7、participateSelectionbiasErrorduetocollectionofincorrectinformationaboutstudysubjects.Duetothisincorrectinformation,subjectsareclassifiedintoincorrectexposureordiseasecategories.InformationbiasSelectionbiasi