資源描述：
《On Weight Smoothing in the Current Employment Statistics Survey October 2014 在加權(quán)平滑在當(dāng)前就業(yè)統(tǒng)計(jì)調(diào)查2014年10月》由會(huì)員上傳分享，免費(fèi)在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫(kù)。
1、JSM2014-SurveyResearchMethodsSectionOnWeightSmoothingintheCurrentEmploymentStatisticsSurveyOctober2014JulieGershunskaya,MichaelSverchkovBureauofLaborStatistics,2MassachusettsAve.NE,Washington,DC,20212AbstractThesamplingweightintheCurrentEmploymentStatisticsSurveyisdete
2、rminedatthetimeofsampleselection.Itdependsonaunit’sState,industry,andsizeclass.However,thepopulationofbusinessesishighlydynamic.Establishmentsconstantlygroworcontract;sometimestheyalsochangetheirindustrialclassificationorgeographicallocation.Eventhenumberofpopulationun
3、itsisnotfixedbutcontinuouslychangesovertime.Aunitmaychangeitssizeclassatthetimeofestimationorthecontentoftheoriginalstratummaychange.Undersuchcircumstances,applicationoftheoriginalsurveyweightsmayincreasevolatilityofsurveyestimates.Inthispaperweinvestigateifthesurveyes
4、timatescanbeimprovedbyadjustingtheoriginalweights.KeyWords:samplingweights,extremeobservation,businesssurvey,stratumjumper1.IntroductionUndertheclassicaldesign-basedapproachtoinferencesfromsurveysampling,thesamplingweightsaredefinedatthedesignstageofasurveyandviewedasn
5、on-randomquantitiesattheestimationstage.Incontrast,PfeffermannandSverchkov(1999)andBeaumont(2008)viewthesamplingweightsasrealizationsofarandomvector.Thisallowsmodelingtheweightsandapplyinganew“smoothed”setofweightsestimatedfromthemodel.Theapproachhasthepotentialforthei
6、mprovedefficiencycomparedtotheestimatorbasedontheoriginalweights.Consideraversionoftheexpansionestimatorforthepopulationtotal,whereasetofsmoothedweightsisusedinplaceoftheoriginalsamplingweights.Themethodhasthepotentialtogivegoodresultswhenthenewweightsarebetterrelatedt
7、otheresponsevariable.Theusuallayperson’sinterpretationofthesamplingweightsintheexpansionestimatorgoesasfollows:thinkofsampleunit’sweightasthenumberofcorrespondingunitsinthepopulationhavingthesamevalueofthesampleunit’sresponsevariable.Ifthiscorrespondencewouldholdexactl
8、yforallsampleunits,thenthesampleweightedestimatorprovidesaperfectestimatorofthepopulationtotal.It“estimates”thetotalw
