简介：为概括线性模型(GLM)，以防regressors是随机的并且有不同分布，参数的最大的可能性的估计(MLE)的asymptotic性质被学习。在合理条件下面，我们证明弱、强壮的一致性和asymptotic规度。关键词概括了线性模型-一致性-中国自然科学基础支持的Asymptotic规度先生(2000)题目分类62J12工程

简介：Theactivitiesofgeomagneticstormsaregenerallycontrolledbysolaractivities.Thecurrentsolarcycle(SC)24isfoundtobemild;comparedtoSCs19–23,thestormoccurrenceandsizederivedbyaveragingtheoccurrencenumberandDstaroundthesolarmaximumarereducedbyabout50–82%and36–61%,respectively.Weestimateseparately,forSC19to24,therepeatintervalsbetweengeomagneticstormsofspecificDst,basedonfitsofpower-lawandlog-normaldistributionstothestormdataforeachSC.RepeatintervalsbetweensupergeomagneticstormswithDst≤–250nTarefoundtobe0.36–2.95year(s)forSCs19–23,butabout20yearsbasedonthedataforSC24.Wealsoestimatetherepeatintervalsbetweencoronalmassejections(CMEs)ofspecificspeed(VCME)sinceCMEsareknowntobethemaindriversofintensestormsandtherelatedstatisticsmayprovideinformationaboutthepotentialoccurrenceofsupergeomagneticstormsfromthelocationoftheSun.OuranalysisfindsthataCMEwithVCME≥1860km/smayoccuronceper3and5monthsinSC23and24,respectively.BasedonaVCME-Dstrelationship,suchafastCMEmaycauseastormwithDst=–250nTifarrivingattheEarth.BycomparingtheobservedgeomagneticstormstostormsexpectedtobecausedbyCMEs,wederivetheprobabilityofCMEcausedstorms,whichisdependentonVCME.ForaCMEfasterthan1860km/s,theprobabilityofaCMEcausedstormwithDst≤–250nTisabout1/5forSC23or1/25forSC24.AlloftheaboveresultssuggestthatthelikelihoodoftheoccurrenceofsupergeomagneticstormsissignificantlyreducedinamildSC.

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：Inthispaper,weconsidertheempiricallikelihood-basedinferencesforvaryingcoefficientmodelsY=X~τα(U)+εwhenXaresubjecttomissingatrandom.Basedontheinverseprobability-weightedidea,aclassofempiricallog-likelihoodratios,aswellastwomaximumempiricallikelihoodestimators,aredevelopedforα(u).Theresultingstatisticsareshowntohavestandardchi-squaredornormaldistributionsasymptotically.Simulationstudiesarealsoconstructedtoillustratethefinitesamplepropertiesoftheproposedstatistics.

简介：ConsidertwolinearmodelsXi=U′iβ+εi,Yj=Vj′γ+ηjwithresponsevariablesmissingatrandom.Inthispaper,weassumethatX,Yaremissingatrandom(MAR)andusetheinverseprobabilityweightedimputationtoproduce‘complete’datasetsforXandY.Basedonthesedatasets,weconstructanempiricallikelihood(EL)statisticforthedifferenceofXandY(denotedas?),andshowthattheELstatistichasthelimitingdistributionofχ21,whichisusedtoconstructaconfidenceintervalfor?.ResultsofasimulationstudyonthefinitesampleperformanceofEL-basedconfidenceintervalson?arereported.

简介：Missingdataandtime-dependentcovariatesoftenarisesimultaneouslyinlongitudinalstudies,anddirectlyapplyingclassicalapproachesmayresultinalossofefficiencyandbiasedestimates.Todealwiththisproblem,weproposeweightedcorrectedestimatingequationsunderthemissingatrandommechanism,followedbydevelopingashrinkageempiricallikelihoodestimationapproachfortheparametersofinterestwhentime-dependentcovariatesarepresent.Suchprocedureimprovesefficiencyovergeneralizedestimationequationsapproachwithworkingindependentassumption,viacombiningtheindependentestimatingequationsandtheextractedadditionalinformationfromtheestimatingequationsthatareexcludedbytheindependenceassumption.Thecontributionfromtheremainingestimatingequationsisweightedaccordingtothelikelihoodofeachequationbeingaconsistentestimatingequationandtheinformationitcarries.Weshowthattheestimatorsareasymptoticallynormallydistributedandtheempiricallikelihoodratiostatisticanditsprofilecounterpartfollowcentralchi-squaredistributionsasymptoticallywhenevaluatedatthetrueparameter.Thepracticalperformanceofourapproachisdemonstratedthroughnumericalsimulationsanddataanalysis.

简介：

简介：Duetotheopennessofthecognitiveradionetwork,spectrumsensingdatafalsification(SSDF)canattackthespectrumsensingeasily,whilethereisnoeffectivealgorithmproposedincurrentresearchwork,sothispaperintroducesthemalicioususersremovingtotheweightsequentialprobabilityradiotest(WSPRT).Theterminals'weightisweightedbytheaccuracyoftheirspectrumsensinginformation,whichcanalsobeusedtodetectthemalicioususer.Ifoneterminalownsalowweight,itcanbetreatedasmalicioususer,andshouldberemovedfromtheaggregationcenter.SimulationresultsshowthattheimprovedWSPRTcanachievehigherperformancecomparedwiththeothertwoconventionalsequentialdetectionmethodsunderdifferentnumberofmalicioususers.

简介：Toestimatethespreadingsequenceofthedirectsequencespreadspectrum(DSSS)signal,afastalgorithmbasedonmaximumlikelihoodfunctionisproposed,andthetheoreticalderivationofthealgorithmisprovided.Bysimplifyingtheobjectivefunctionofmaximumlikelihoodestimation,thealgorithmcanrealizesequencesynchronizationandsequenceestimationviaadaptiveiterationandslidingwindow.Sinceitavoidsthecorrelationmatrixcomputation,thealgorithmsignificantlyreducesthestoragerequirementandthecomputationcomplexity.Simulationsshowthatitisafastconvergentalgorithm,andcanperformwellinlowsignaltonoiseratio(SNR).

简介：Inthispaper,wenotonlyconstructtheconfidenceregionforparametersinamixedinteger-valuedautoregressiveprocessusingtheempiricallikelihoodmethod,butalsoestablishtheempiricallog-likelihoodratiostatisticandobtainitslimitingdistribution.Andthen,viasimulationstudieswegivecoverageprobabilitiesfortheparametersofinterest.Theresultsshowthattheempiricallikelihoodmethodperformsverywell.

简介：目标用无线传感器网络追踪在传感器之中要求有效合作到在精力消费和追踪的精确性之间的折衷。这份报纸论述用最大的可能性的评价和Kalman过滤器的联合在无线传感器网络追踪途径的一个合作目标。簇领导人把收到的非线性的距离大小变换成线性观察模型并且接近用最大的可能性的评价的变换测量噪音的协变性，然后使用Kalman递归地用变换大小更新目标状态估计的过滤器。最后，一项措施被领导人基于最大的可能性的评价的菲希尔信息矩阵使用为进一步的追踪作为新追踪簇选择最增进知识的传感器。建议合作追踪的途径的优点经由模拟结果被表明。

简介：

简介：Theauthorsstudytheempiricallikelihoodmethodforpartiallylinearerrors-in-variablesmodelwithcovariatedatamissingatrandom.Empiricallikelihoodratiosfortheregressioncoefficientsandthebaselinefunctionareinvestigated,andthecorrespondingempiricallog-likelihoodratiosareprovedtobeasymptoticallystandardchi-squared,whichcanbeusedtoconstructconfidenceregions.Thefinitesamplebehavioroftheproposedmethodsisevaluatedbyasimulationstudywhichindicatesthattheproposedmethodsarecomparableintermsofcoverageprobabilitiesandaveragelengthofconfidenceintervals.Finally,theEarthquakeMagnitudedatasetisusedtoillustrateourproposedmethod.

简介：Inthisarticle,weconsideralifetimedistribution,theWeibull-Logarithmicdistributionintroducedby[6].Weinvestigatesomenewstatisticalcharacterizationsandproperties.WedevelopthemaximumlikelihoodinferenceusingEMalgorithm.AsymptoticpropertiesoftheMLEsareobtainedandextensivesimulationsareconductedtoassesstheperformanceofparameterestimation.Anumericalexampleisusedtoillustratetheapplication.

简介：Blindidentification-blindequalizationforFiniteImpulseResponse(FIR)MultipleInput-MultipleOutput(MIMO)channelscanbereformulatedastheproblemofblindsourcesseparation.Ithasbeenshownthatblindidentificationviadecorrelatingsub-channelsmethodcouldrecovertheinputsources.TheBlindIdentificationviaDecorrelatingSub-channels(BIDS)algorithmfirstconstructsasetofdecorrelators,whichdecorrelatetheoutputsignalsofsubchannels,andthenestimatesthechannelmatrixusingthetransferfunctionsofthedecorrelatorsandfinallyrecoverstheinputsignalusingtheestimatedchannelmatrix.Inthispaper,anewapproximationoftheinputsourceforFIR-MIMOchannelsbasedonthemaximumlikelihoodsourceseparationmethodisproposed.TheproposedmethodoutperformsBIDSinthepresenceofadditivewhiteGaussiannoise.

简介：Epidemiologicstudiesuseoutcome-dependentsampling(ODS)schemeswhere,inadditiontoasimplerandomsample,therearealsoanumberofsupplementsamplesthatarecollectedbasedonoutcomevariable.ODSschemeisacost-effectivewaytoimprovestudyefficiency.Wedevelopamaximumsemiparametricempiricallikelihoodestimation(MSELE)fordatafromatwo-stageODSschemeundertheassumptionthatgivencovariate,theoutcomefollowsagenerallinearmodel.Theinformationofbothvalidationsamplesandnonvalidationsamplesareused.Whatismore,weprovetheasymptoticpropertiesoftheproposedMSELE.