縱向數(shù)據混合效應模型的研究.pdf

1、揚州大學碩士學位論文縱向數(shù)據混合效應模型的研究姓名：戴小飛申請學位級別：碩士專業(yè)：應用數(shù)學指導教師：宗序平20090501揚州大學碩：論文MixedEffectsModelforLongitudinalDataABSTRACT2一Longitudinaldataisreferredtodatainwhichindividualsaremeasuredrepeatedlythroughtime，SOitcombineselementsof

2、crosssectionaldataandtime—seriesdataManyofstatisticianshavebeeninterestedinlongitudinaldataatpresentItcombinesthecharacteristicsofcross—sectionaldataandtime—seriesdataItisobservedandgotintheorderoftimebythesamesample，SOl

3、ongitudinaldatacananalyzeeffectivelythechangeofindividualsandvariationamongindividualsItplaysallimportantpartinapplicationscontrasttocrosssectionaldataortimeseriesdataInthispaperwestudytwomodels：linearmixed—effectsmodela

4、ndsemiparametricmixedeffectsmodelThemaincontentofthispaperisasfollows：1Chapter1introduceswhatislongitudinaldataanditscharacterThenwecompareitwithtimeseriesdata，crosssectiondataandmultivariatestatisticaldata，SOwefindthead

5、vantageoflongitudinaldataAtlast，weintroducethestudysituationatpresentandthemainworkofthispaper；2Chapter2coversthelinearmixedeffectsmodeloflongitudinaldataanddiscusstheestimationofregressionparametersandcovarianceparamete

6、rsWeusetheNewton—Raphonformulatogettheestimationofunknowncovarianceparameters0；3Chapter3describesthehypothesistestsofsomeparameters；(1)Ftest，solvetheleastsquaresestimationofunknownparameters，thenwecangettheFstatisticalte

7、stIfthenullhypothesisistrue，互～F(q，N—npq)；Ifwegivethesignificantlevel口，itsrejectionregionisasfollows：曩巧一口(g，Ⅳ一印一q)；(2)LikelihoodratiotestWeconsiderthemaximumlikelihoodestimationofunknownparameters，thenproposetherelationbe

8、tween8and8LOnthebasisofthespectralrepresentationsoftheLRTweprovideasimplealgorithmtosimulatethenullfinitesampledistributionof三尺瓦Animportantfeatureofthealgorithmisthatitsspeeddependsonthenumberofrandomeffectsq(3)Restricte