内容简介
AthoroughreviewofthemostcurrentregressionmethodsintimeseriesanalysisRegressionmethodshavebeenanintegralpartoftimeseriesanalysisforoveracentury.Recently,newdevelopmentshavemademajorstridesinsuchareasasnon-continuousdatawherealinearmodelisnotappropriate.Thisbookintroducesthereadertonewerdevelopmentsandmorediverseregressionmodelsandmethodsfortimeseriesanalysis.Accessibletoanyonewhoisfamiliarwiththebasicmodernconceptsofstatisticalinference,RegressionModelsforTimeSeriesAnalysisprovidesamuch-neededexaminationofrecentstatisticaldevelopments.Primaryamongthemistheimportantclassofmodelsknownasgeneralizedlinearmodels(GLM)whichprovides,undersomeconditions,aunifiedregressiontheorysuitableforcontinuous,categorical,andcountdata.TheauthorsextendGLMmethodologysystematicallytotimeserieswheretheprimaryandcovariatedataarebothrandomandstochasticallydependent.Theyintroducereaderstovariousregressionmodelsdevelopedduringthelastthirtyyearsorsoandsummarizeclassicalandmorerecentresultsconcerningstatespacemodels.Toconclude,theypresentaBayesianapproachtopredictionandinterpolationinspatialdataadaptedtotimeseriesthatmaybeshortand/orobservedirregularly.Realdataapplicationsandfurtherresultsarepresentedthroughoutbymeansofchapterproblemsandcomplements.Notably,thebookcovers:*ImportantrecentdevelopmentsinKalmanfiltering,dynamicGLMs,andstate-spacemodeling*AssociatedcomputationalissuessuchasMarkovchain,MonteCarlo,andtheEM-algorithm*Predictionandinterpolation*Stationaryprocesses