简介:Thispaperstudiesthemodel-robustdesignproblemforgeneralmodelswithanunknownbiasorcontaminationandthecorrelatederrors.ThetrueresponsefunctionisassumedtobefromareproducingkernelHilbertspaceandtheerrorsarefittedbytheqthordermovingaverageprocessMA(q),especiallytheMA(1)errorsandtheMA(2)errors.Inbothsituations,designcriteriaarederivedintermsoftheaverageexpectedquadraticlossfortheleastsquaresestimationbyusingaminimaxmethod.Acaseisstudiedandtheorthogonalityofthecriteriaisprovedforthisspecialresponse.Therobustnessofthedesigncriteriaisdiscussedthroughseveralnumericalexamples.
简介:Countdatawithexcesszerosareoftenencounteredinmanymedical,biomedicalandpublichealthapplications.Inthispaper,anextensionofzero-inflatedPoissonmixedregressionmodelsispresentedfordealingwithmultileveldataset,referredashierarchicalmixturezero-inflatedPoissonmixedregressionmodels.AstochasticEMalgorithmisdevelopedforobtainingtheMLestimatesofinterestedparametersandamodelcomparisonisalsoconsideredforcomparingmodelswithdifferentlatentclassesthroughBICcriterion.AnapplicationtotheanalysisofcountdatafromaShanghaiAdolescenceFitnessSurveyandasimulationstudyillustratetheusefulnessandeffectivenessofourmethodologies.