资源描述：
《Identification Environment and RobustForecasting for Nonlinear Time Series 识别环境与鲁棒性 非线性时间序列的预测》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、ComputationalEconomics7:37-53,199491994KluwerAcademicPublishers.PrintedintheNetherlands.IdentificationEnvironmentandRobustForecastingforNonlinearTimeSeriesBERLINWUDepartmentofMathematicalSciences,NationalChengchiUniversity,Taiwan(Received:June1992)Abslract.Inthispaper,themet
2、hodsoftimeseriesfornonlinearityarebrieflysurveyed,withparticularattentionpaidtoanewtestdesignbasedonaneuralnetworkspecification.Theproposedintegratedexpertsystemcontainstwomaincomponents:anidentificationenvironmentandarobustforecastingdesign.Theidentificationenvironmentcanbev
3、iewedasaintegrateddynamicdesigninwhichcognitivecapabilitiesariseasadirectconsequenceoftheirself-organizationalproperties.Theintegratedframeworkusedfordiscussingthesimilaritiesanddifferencesinthenonlineartimeseriesbehaviorispresented.Moreover,itsperformanceinpredictionprovesto
4、besuperiorthantheformerwork.Fortheinvestigationofrobustforecasting,weperformasimulationstudytodemonstratetheapplicabilityandtheforecastingperformance.Keywords,nonlineartimeseries,bilinear,Lagrangemultipliertest,neuralnetwork,forecasting,robust.1.IntroductionTheanalysisintimes
5、eriesmodelshasbeenconcernedwithprocesseswhicharestationary.Testsforunitrootsintimeseriesdatahavebeenthesubjectofattentionineconometricsaswellaswithstatisticiansinthelasttwodecades.Muchoftheresearchhasconcentratedonthedistributiontheorythatisnecessarytodevelopthesetestsandthea
6、nalysisofthepowerofvarioustestsunderdifferentalternativehypotheses.However,inamajorityofeconomicsapplica-tionstheneedofanonlineartes~tisaprioriratherthanunitrootstestfornonstationarity.AsthepapersbyTsay(1991),Granger(199t),andGooijerandKumar(1992)indicate,theinterestinapplyin
7、gnonlineartimeseriesmodelshasconsiderablyincreasedrecently.Therealsoseemstobestrongbeliefamongeconomiststhatrelationshipsbetweeneconomicvariablesarenonlinear;pro-ductionmodelbeinganexample.Specifically,givenparametricrelationship,suchastheCobb-DouglasorCESproductionmodels,sta
8、ndardeconometricstech-niquesprovidewaysofestimationoftheparametersas
