资源描述:
《Hybrid bootstrap aided unit root testing》由会员上传分享,免费在线阅读,更多相关内容在学术论文-天天文库。
1、ComputStatDOI10.1007/s00180-011-0290-0ORIGINALPAPERHybridbootstrapaidedunitroottestingC.Jentsch·J.-P.Kreiss·P.Mantalos·E.PaparoditisReceived:27November2009/Accepted:24October2011©Springer-Verlag2011AbstractInthispaper,weproposeahybridbootstrapprocedureforaugmentedDickey-Fuller(ADF)testsforthep
2、resenceofaunitroot.Thishybridproposalcom-binesatimedomainparametricautoregressivefittothedataandanonparametriccorrectionappliedinthefrequencydomaintocapturefeaturesthatarepossiblynotrepresentedbytheparametricmodel.Itisknownthatconsiderablesizeandpowerproblemscanoccurinsmallsamplesforunitroottes
3、tinginthepresenceofanMAparameterusingcriticalvaluesoftheasymptoticDickey-Fullerdistribution.Theben-efitofthesievebootstrapinthissituationhasbeeninvestigatedbyChangandPark(JTimeSerAnal24:379–400,2003).Theyshowedasymptoticvalidityaswellassubstantialimprovementsforsmallsamplesizes,buttheactualsize
4、softheirbootstraptestswerestillquitefarawayfromthenominalsize.ThefinitesampleperformancesofourprocedureareextensivelyinvestigatedthroughMonteCarlosimulationsandcomparedtothesievebootstrapapproach.Regardingthesizeofthetests,ourresultsshowthatthehybridbootstrapremarkablyoutperformsthesievebootstr
5、ap.KeywordsHybridbootstrap·Sievebootstrap·Unitroottesting·ADFtestsC.JentschDepartmentofEconomics,UniversityofMannheim,L7,3-5,68131Mannheim,GermanyJ.-P.KreissInstitutfürMathematischeStochastik,TechnischeUniversitätBraunschweig,Pockelsstraße14,38106Braunschweig,GermanyP.Mantalos(B)DepartmentofSt
6、atistics,SwedishBusinessSchool,UniversityofÖrebro,70182Örebro,Swedene-mail:panagiotis.mantalos@oru.seE.PaparoditisDepartmentofMathematicsandStatistics,UniversityofCyprus,P.O.Box20537,1678Nicosia,Cyprus123C.Jentschetal.1IntroductionWhenwestudythesmallsamplepropertiesofatestprocedurebycomparingd
7、ifferenttests,twoaspectsareofprimeimportance:(a)findingthetestthathasactualsizeclosesttothenominalsize,and,giventhat(a)holds;(b)findingthetestthathasthegreatestpower.Inmostcases,however,thedistributionoftheteststatisticweuseisknownonlyasy