资源描述：
《New Methods of Time Series Analysis时间系列分析的心方法》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、NewMethodsofTimeSeriesAnalysisofNon-StationaryEEGData:EigenstructureDecompositionsofTimeVaryingAutoregressionsAndrewD.Krystal,RaquelPradoandMikeWestMay31,1999AndrewKrystalisAssistantProfessorintheDepartmentofPsychiatryandBehavioralSciencesandDirectoroftheQuantitativeEEGLaboratoryatDukeUnivers
2、ityMedicalCenter,DurhamNC27710,USA.RaquelPradoisAssistantProfessor,DepartamentodeComputoCient coyEstadstica,UniversidadSimonBolvar,Caracas,Venezuela.MikeWestisArts&SciencesProfessorofStatisticsandDecisionSciences,andDirectoroftheInstituteofStatisticsandDecisionSciences,DukeUniversity,Du
3、rham,NC27708-0251,USA.TheworkreportedherewascarriedoutintheDepartmentofPsychiatryandtheInstituteofStatisticsandDecisionSciencesatDukeUniversity.Theauthorsacknowledgepartial nancialsupportundergrantsNIMHK20MH01151,R29MH57532andNSF/DMS-9704432.Addressforcorrespondence:Dr.A.D.Krystal,Box3309,DukeU
4、niversityMedicalCenter,DurhamNC27710.tel:(919)681-8742,fax:(919)681-8744.0AbstractOBJECTIVE:ThosewhoanalyzeEEGdatarequirequantitativetech-niquesthatcanbevalidlyappliedtotimeseriesexhibitingrangesofnon-stationarybehavior.Ourobjectiveistointroduceanewanalysistechniquebasedonformalnon-stationaryti
5、meseriesmodels.Thisnovelmethodprovidesadecompositionofthetimeseriesintoasetoflatent"componentswithtime-varyingfrequencycontent.Theidenti- cationofthesecomponentscanleadtopracticalinsightsandquanti-tativecomparisonsofchangesinfrequencystructureovertimeinEEGtimeseries.DESIGNandMETHODS:Thetechniq
6、uebeginswiththedevelopmentoftime-varyingautoregressivemodelsoftheEEGtimeseries.SuchmodelshavebeenpreviouslyusedinEEGanalysisbutweextendtheirutilitybytheintroductionofeigenstructuredecom-positionmethods.Wereviewthebasisandimplementationofthismethodandreportontheanalysisof2channelEEGdatarecordedd
7、uring3generalizedtonic-clonicseizuresinducedinanindividualaspartofacourseofelectroconvulsivetherapyformajordepression.RESULTS:Thistechniqueidenti edEEGpatternsconsistentwithpriorreports.Inaddition,itquanti edadecreaseindominantfre
