简介:当处理回归分析时,heteroscedasticity是作者不得不面对的一个问题。特别如果很少信息都不能预先被得到,heteroscedasticity的察觉以及统计模型的评价能是甚至更困难的。到这个目的,这份报纸建议能有效地估计heteroscedastic功能的一个quantile差别方法(QDM)。这个方法,从吝啬的回归的评价是完全自由的工作,简单、柔韧、容易实现。而且,没有任何限制,QDM方法在错误术语启用heteroscedasticity的察觉,因而是广泛地适用。值得提及的,基于建议途径评估者是那两个都,吝啬的回归功能和heteroscedastic功能能被获得。最后,作者进行一些模拟检验建议方法的表演并且使用一个真实数据做一幅插图。
简介:Recentexperiencehasshownthatinterior-pointmethodsusingalogbarrierapproacharefarsuperiortoclassicalsimplexmethodsforcomputingsolutionstolargeparametricquantileregressionproblems.Inmanylargeempiricalapplications,thedesignmatrixhasaverysparsestructure.Atypicalexampleistheclassicalfixed-effectmodelforpaneldatawheretheparametricdimensionofthemodelcanbequitelarge,butthenumberofnon-zeroelementsisquitesmall.AdoptingrecentdevelopmentsinsparselinearalgebraweintroduceamodifiedversionoftheFrisch-NewtonalgorithmforquantileregressiondescribedinPortnoyandKoenker[28].Thenewalgorithmsubstantiallyreducesthestorage(memory)requirementsandincreasescomputationalspeed.Themodifiedalgorithmalsofacilitatesthedevelopmentofnonparametricquantileregressionmethods.Thepseudodesignmatricesemployedinnonparametricquantileregressionsmoothingareinherentlysparseinboththefidelityandroughnesspenaltycomponents.ExploitingthesparsestructureoftheseproblemsopensupawholerangeofnewpossibilitiesformultivariatesmoothingonlargedatasetsviaANOVA-typedecompositionandpartiallinearmodels.
简介:Recently,variableselectionbasedonpenalizedregressionmethodshasreceivedagreatdealofattention,mostlythroughfrequentist'smodels.ThispaperinvestigatesregularizationregressionfromBayesianperspective.OurnewmethodextendstheBayesianLassoregression(ParkandCasella,2008)throughreplacingtheleastsquarelossandLassopenaltybycompositequantilelossfunctionandadaptiveLassopenalty,whichallowsdifferentpenalizationparametersfordifferentregressioncoefficients.BasedontheBayesianhierarchicalmodelframework,anefficientGibbssamplerisderivedtosimulatetheparametersfromposteriordistributions.Furthermore,westudytheBayesiancompositequantileregressionwithadaptivegroupLassopenalty.Thedistinguishingcharacteristicofthenewlyproposedmethodiscompletelydataadaptivewithoutrequiringpriorknowledgeoftheerrordistribution.Extensivesimulationsandtworealdataexamplesareusedtoexaminethegoodperformanceoftheproposedmethod.Allresultsconfirmthatournovelmethodhasbothrobustnessandhighefficiencyandoftenoutperformsotherapproaches.
简介:Inthisarticle,weconsideraclassofkernelquantileestimatorswhichisthelinearcombinationoforderstatistics.Thisclassofkernelquantileestimatorscanberegardedasanextensionofsomeexistingestimators.Theexactmeansquareerrorexpressionforthisclassofestimatorswillbeprovidedwhendataareuniformlydistributed.Theimplementationoftheseestimatorsdependsmostlyonthebandwidthselection.Wethendevelopanadaptivemethodforbandwidthselectionbasedontheintersectionconfidenceintervals(ICI)principle.MonteCarlostudiesdemonstratethatourproposedapproachiscomparativelyremarkable.Weillustrateourmethodwitharealdataset.
简介:偏导长度的数据在许多重要领域里产生包括流行病学的队研究,癌症屏蔽试用和劳动经济。如此的数据的分析在文学吸引了许多注意。在这份报纸,我们为分析审查权利、偏导长度的数据建议一条quantile回归途径。我们导出反的可能性相应于quantile回归到的加权的估计方程由于长度偏爱采样并且增进知识的审查改正偏爱。这个方法能容易处理增进知识的审查由偏导长度的采样导致了。自从面对长度偏爱并且增进知识的审查获得风险因素的不偏的估计通常是困难的,这是我们的建议方法的一个呼吁的特征。我们用实验过程技术建立一致性和建议评估者的asymptotic分发。一个采样方法被采用估计评估者的变化。我们进行模拟研究评估它一个真实数据设置了说明建议方法的应用程序的有限样品性能和使用。
简介:1.IntroductionTheestimationofpopulationquaillesisofgrestillterestwhenone.isnotpreparedtoassumeaparametricformfortheunderlyingdistribution.Inaddition,quaillesoftenariseasthensturalthingtoestimatewhentheunderlyingdistributionisskewed.LetXIIXZ,’’’)Xubei...
简介:这份报纸与错过反应数据的nonignorable考虑分发函数和quantiles的评价问题。三条途径被开发估计分发功能和quantiles,即,Horvtiz-Thompson-type方法,回归归罪方法和扩充反的概率加权的途径。倾向分数被semiparametric指定指数的倾斜模型。为了在倾向估计倾斜的参数,得分,作者建议一个调整实验可能性的方法处理过去识别的系统。在一些常规条件下面,作者调查asymptotic性质为分发功能和quantiles建议了三个评估者,并且发现这些评估者有一样的asymptotic变化。大折刀方法被采用一致地估计asymptotic变化。模拟研究被进行调查建议方法论的有限样品表演。
简介:Quantileregressionisgraduallyemergingasapowerfultoolforestimatingmodelsofconditionalquantilefunctions,andthereforeresearchinthisareahasvastlyincreasedinthepasttwodecades.Thispaper,withthequantileregressiontechnique,isthefirstcomprehensivelongitudinalstudyonmathematicsparticipationdatacollectedinAlberta,Canada.Themajoradvantageoflongitudinalstudyisitscapabilitytoseparatetheso-calledcohortandageeffectsinthecontextofpopulationstudies.Oneaimofthispaperistostudywhetherthefamilybackgroundfactorsalterperformanceonthemathematicalachievementofthestrongeststudentsinthesamewayasthatofweakerstudentsbasedonthelargelongitudinalsampleof2000,2001and2002mathematicsparticipationlongitudinaldataset.Theinterestingfindingssuggestthattheremaybedifferentialfamilybackgroundfactoreffectsatdifferentpointsinthemathematicalachievementconditionaldistribution.