部分线性EV模型中参数的经验似然比置信区域

《部分线性EV模型中参数的经验似然比置信区域》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。

1、0了{飞拭每嚼冷;、}f'5“布、‘可“Y,未‘万V夕t},*}}手卯产。‘义。““辛，“‘玲“‘·妞一省r}。‘雳v"4tul}‘令才乏丫，山·“乞雇r汉‘粉‘饰乏劣乡‘宕舜今争·k/司}j，哆扩}}}},p/i3}怜7鱼y,一Fti/}}7}_-cit‘云了枷‘呜塑比ACKNOWLEDGEMENTIamgratefultomytutorProfessorCuiHengjianandProfes-sorLiYongfortheirhelpduringthethreeyearsofgraduatestudy.EspeciallyPro

2、fessorCui'svaluableadviceandcarefulguidanceonmyresearchconductsmadegreatcontributiontomyimprovement.Hishelp,includingkindlysupplyingmewithsomeusefulreferences,isimportantformetofinishthisarticle.Hisseriousattitudetowardscientificresearchandhissophisticatedknowledgebenef

3、itmeverymuch.SoIwouldliketoexpressmygratitudetoProfessorCuiandotherteachersfortheirhardwork.Iwouldalsoliketothankmyclassmates,includingZhaoChun,AnJinbingandJinJiaofortheirhelpbothwithmystudyandwithmylife.ABSTRACTSThispaperproposesaconstrainedempiricallikelihoodconfi-den

4、ceregionforaparameter局Semi-linealErrors-in-Variables(EV)model:妖=x丁凡=x;+u;,(1<‘<。)whichisconstructedbycombiningthescorefunctioncor-respondingtothesquareorthogonaldistancewithacon-strainedregionof凡.Simulationsshowthattheproposedconfidenceregionhasquitesatisfactorycoverage

5、probabilityandshortinternallengthnotonlyforlargesamples,butalsoformediasamples.KEYM70RDSSemi-linear;Errors-in-Variables;Confidenceregions;Empiricallikelihood;Kernelfunction;Windowwidth;Coverageprobability.4互1IntroductionFirst.letushavealookatlinearErrors-in-Variables(EV

6、)regressionmodelY=x'丙+〔1二a+vwhere:i0isapx1vectorofunknownparameters.-landuarethepx1unobservablecov-ariatesandmeasurementerrorvectorsrespectively.Yisascalarresponseand‘isthemodelerror.Itisassumedthatxand(〔.u-)rareindependent.Thelasttwodecadeshaveseenanincreasingnumberofa

7、pplicationsofthelinearEVmodelsduetoitssimpleformandwideapplicability.Theempiricallikelihoodasanalternativetothebootstrapforconstructingconfidenceregionsnouparametrically~introduced勿Owen(1988).Butinsteadofresamplingwithequalprobabilityweightslikethebootstrap.itworks切一pro

8、filingamultinomiallike-lihoodsupportedonthesample.Themethoddefinesanempiricallikelihoodratiofunction.andusesit