A new biased estimator for multivariate regression models with highly collinear variables [Elektronische Ressource] / vorgelegt von Julia Wissel

Description

henLehrstuhlWisself?riusMaximiliansUnivStatistikErlangungInstitutyf?rorgelegtMathematiFknaturwissenscUnivderersit?thenW?rzburgW?rzburgAonnewM?mbiased2009estimatordesforhaftlicmDoktorgradesultivBaariateeriscregressionJulmoersit?tdelsvwithvhighlyJuliacollinearausvbrisariablesebruarDissertationzurhenLehrstuhlWisself?riusMaximiliansUnivStatistikErlangungInstitutyf?rorgelegtMathematiFknaturwissenscUnivderersit?thenW?rzburgW?rzburgAonnewM?mbiased2009estimatordesforhaftlicmDoktorgradesultivBaariateeriscregressionJulmoersit?tdelsvwithvhighlyJuliacollinearausvbrisariablesebruarDissertationzur117ConErrortenLineartsTheInEstimatortroLinearductionof1TheChapterA1.EconomicSp57ecicationedofEthePropMoDLSEdel7.5RidgeChapterof2.RidgeCriteria6.foroComparingofEstimators579742.1.theMultivMatrixariate6.5.MeannSquaredDataErrorEstimator10Economic2.2.tiMatrixnMean43SquaredFErrorRid105.5.Chapterof3.49StandardizationDisturbofetheMoRegressionDerivCoDisturbecienMotsThe15Squares3.1.MeanCenertiesteringLRegression6.4.MoSquareddelsDLSE15Squared3.2.ofScaling96Cenfortered6.7.RegressionRidgeMo6.8.delsof19112Chapterul4.izaMulticollinearitoyin23Regression4.1.5.4.SourcesGeneraloformMulticollinearittheyge23444.2.

Sujets

• Mathematics

Informations

• Publié par: julius-maximilians-universitat_wurzburg
• Publié le: 01 janvier 2009
• Nombre de lectures: 16
• Langue: English
• Poids de l'ouvrage: 2 Mo

Extrait

henLehrstuhlWisself?riusMaximiliansUnivStatistikErlangungInstitutyf?rorgelegtMathematiFknaturwissenscUnivderersit?thenW?rzburgW?rzburgAonnewM?mbiased2009estimatordesforhaftlicmDoktorgradesultivBaariateeriscregressionJulmoersit?tdelsvwithvhighlyJuliacollinearausvbrisariablesebruarDissertationzurhenLehrstuhlWisself?riusMaximiliansUnivStatistikErlangungInstitutyf?rorgelegtMathematiFknaturwissenscUnivderersit?thenW?rzburgW?rzburgAonnewM?mbiased2009estimatordesforhaftlicmDoktorgradesultivBaariateeriscregressionJulmoersit?tdelsvwithvhighlyJuliacollinearausvbrisariablesebruarDissertationzurk
~!
Ridgetrotheductiono1TheChapterRidge1.57Sp6.5.ecicationChapterofofthe6.Moofdel745MatrixChapter6.6.2.Example:Criteria117forAComparingEstimatorEstimatorsEconomic9Linear2.1.57MultivLinearariateedMeanErrorSquaredEErrorof10Prop2.2.UnstandardizedMatrixRegressionMeantheSquaredtiErrorn1043ChapterF3.RidStandardization5.5.ofofthe49RegressionDisturbCoeecienMotsDeriv15Disturb3.1.MoCenTheteringSquaresRegressionMeanMoertiesdelsL156.4.3.2.SquaredScalingDLSECenSquaredteredofRegressionDLSEMo105delsas19111ChapterDLSE4.DataMulticollinearitulyn23o4.1.inSourcesRegressionof5.4.MulticollinearitGeneralyorm23the4.2.geHarmful44EectsExample:ofRegressionMulticollinearittheyData24Chapter4.3.TheMulticolledinearitRygressiDniagdeln6.1.oationstheticsed28Regression4.4.delExample:6.2.MulticollDisturbinearitLeastyEstimatorof6.3.theSquaredEconomicPropDataof31DChapterS5.82RidgeTheRegressionMean35Error5.1.theThe93RidgeMeanEstimatorErrorofertieststenInHo96erlTheandforKDataennard6.7.35DLSE5.2.RidgeEstimatingEstimatorThe6.8.BiasingTheFofaEconomicctor112Con7.Sim40a5.3.oStandardStudyizaitiAlgebra7.1.wledgmenThe151AlgorithmA.118Bibliograph7.2.cThe155SimMatrixulation137Resultsy124AConclusionkno135tAppendixthInnormallytropredictedductionereInoftentheonescienhticnon-eldsonofalsoastronomrepresenymaximandbgeoedesyappliscienandtistsdhainvaethealwtheatermsyshlocaseokyedula-foresolutionsytoineasefreethefewnacationvigationtheonwthe1widewhomateceadel.nminimizesetandedtovalloofwdel.advleastenalentuorersntoariables.discosquaresvkerGaussMarknewsquaresanlidNevunknboremainingwnunlandscapfeweserrorduringAstheareAgespofaExploration.estimationItbwdaseathecommontecpracticedernatitthattotimeparameterstoregressionrelytonapproaclandsumsighdierencestingseentothedeterminealuestheendencurrenwhictthepsquaredositiontheofathetheship.estimator,Hoalsowtoevliker,estimatortoindepenabletoerrorvpertheseartratracedvtheelsofthvTheorem,etheptoositionbofunthetimashiinptherehadrathertoimplicationsbneclassdeterminedestimators.bwyariablesmeansmeasuremenofortheccelestialconsequence,bmoospdies.usThereforeciitresultwoutcomeasleastanused.issueoftook,describUllaherightheirtobbehabanevior.squaresAopular,rsthniquemamojorstatistics,steperetoisdescribusedeesti-thethemoofvlinearemenmotCurrenandandhencetraditionalpredicthesthethetraofjectorysquaredofbawstarthewandasobservpverformedofbdepytCarlariable,FhriedrictshsumGauss.theHeerroruofsedmoobservAsationssolutionandgetstherebordinaryysquaresdevwhicelopisedequivthetfundamenthetalsumforeliholeastdsquaresinanalysisofine1795.denHoandwdistributedevver,ThetheopularithonorofofleastpresenestimatotingcantheerstbacpublicationtoofformthetionleastthesquaresomethowhicdstatesbleastelongsestimatortobLegendre.theWithinesthisnearwbiasedork,spublishedtor.inertheless,1805,manhecasespresenwilltsethesevrulestatisticalasofaoconlyvthee-ofnienbiasedtInmethoeddorkonlyv,arewhereasofatrstand/proareofstoofhastic.theamethoonlydstatisticalwdelsascorrectlypublishedecied,bthytheseRobeertAerrorsdinrabiasediwhenneinsquares1808.isThInuprefacestheirtheotraditionalVinoleastandsquaresfounstatistictheistprobablyordsonedescribofthetheooldest,andbofutleaststillestimator:mostpew."Sincewthyeesexpaoec-sibiasedtiondel.ofthetherenklerGaussMarkeciedomatrixvinTheoremar[...]leastpracticallyploallmathematicaleldsTherefore,inythsquaresepredictivnat-vuraleandinformationsoterscialalmostsciencestohayvneofwidelyyandtistsome-yieldtimesnotblindlycusedbtheacceptOrdinarysmallerLeastwroteSquaresoptimalmethotod".ofItewise,ismorethereforesamplenottionastonishisincengethatBecausethemresultwofyCharlestheSteinforinngthedierenlatev1950's,alterna-statingothatathformemethoretextexistsoaappbestimatoetterresult,alternativteedevtoisthemaleastmoresquarestageousestimatorslighunestimator-ariance.der58certainonconditions,hrstlyatremainedcanunnoticed.oSomequal-ycorrectlyearsdel.later,toJamesoneandationSteinfrompropimproosedbadantheexplicitnotbiasedanestimatoryforwnwhicyhoftentheproblemimpronearitveementheretwtothantheEspleastestimatorsquareseestimatorulticollininTherefore,termspastomanfkindsthetorsmeanbsquaredtederroreswestimatorsasthequiteparameterssubstanregressiontialofinusethesusualotherslinearthemothedel.oinPvaralleldierentoacthis,sameHowitherlAsandyKennardguaransuggestedthatanothertotalbiasedarianestimator,ethsmall.eitso-calledyridgeeregressionadvestimator,nstilltooantlebiasedofwiththevmostTp(1981,[opular])biased"Insistingestimatorleasttowhicdameansyt.anProbablypricetheleadbpesortemotivitiesationaforspfurthermoresearcLikhtryinginremotheeeldorofobservbiasedvestima-torstionthewtoavsthethecondi-theofproblemregressorofismadvisibleurelevlticollinearittyma.bInthrocaseaofam".ultionecoencounlthelinearitofyulticolli,yi.applied.ork,ifistherenoiasoutausingstrongestimators.depeciallyendencyridgebseemsetbwapplicableemenethedata.coludurimthen30searsofythteofdesigntimamhaatrix,etheeeleastpresensquaresasestimatestivtendtotosquaresbfervestimationerytheunstableofandlinearunreliable.moAlthSomeoughthesethulationseBaGauss-MarkeoianvTheoremds,assuresem-thatytheconleastofsquaresfrequenestimatorphastminimalftotalivOftenariancetinrothehesclasstheoforthrsesimilarunforms.biasedaestimators,manit2is! ; j = 1;:::;pj
!
X2
np T
R X ; j = 1;:::;p j x jj j
X j X X ; j = 1;:::;pj j
btemanotherebiasedfewestimator,awithoneimprou-vforedformeanmoresquaredsquarederrorapropbertiestothanestimationtheoutleastsquarsquaresinestimator,thisandwilltoancomparesquaresitstandardizedwithsturbtheridgefamous,ridpropgatoreenestimator.aTheandthesiscolumnispapdivideditsinbtoissevyentries.cchaptersvandwishorganizedeasthanfollosquaresws:eIndataChapterestimator1thethewillusedaregressiontomoerformancedelleastisandspClosingeciedourandliographthebiasednecessaryrenkleracolumnssthump-oftionsdenotedareoexplained.leAfterwestimatorardpropsdataanpresenin6.troonductionquantonriskgoalfun500ctionsonandregressor.espeeciallythattoantheysmean,squaredtheerrorofforedcomparingisestimatorscorrespisthegivBesidesendel,inalsoChapterfor2.theTheleastmethobdeddedofofstandardiWzthisationmeansofulationregressionhcoaecienthets,foftendisturbusedestimatorinleastregressionestitheoryChapterandsectionappliedordswGiorkmatrix,hisisedescribeyedvin1981,detailAlreadyinestimators.Chapterthe3.wTherebTheyabthetodis-ycussionhaabandoutdtheastadvesanandtagestheoreticalandertiesdistandardizedsawilldevtedanChaptertagesItofbasedstandardizationaddinginsmallregressiontittheoryuscriptismanotofignored.TheAfterenathanreviewtainedofeathehproblItebmproofenmwulticollinearitcyal-inaChapternd4,conansucinthattromeanductionerrortothridgedisturbestimationleastwillestimatorbsmallerethegivondingenofinleastChapterestimator.5.theBecausemoofwthewillexhaustindvsolutioneunstandardizedinandvdies-edtigationssquaresdonewillinethisbeld,initclassistheinestimators.tendedetoconcludegivthesiseyonlyofasimroughstudyowhicvtrieserview.evThereblyatethepemphasisoliestheonosedtheedridgesquaresestimatorcomparedoftheHosquareserlridgeandmKennardinand7.itsthisstatisticalapropwerties,regardingwhicnotation:hvwillabyebibusefulandforwthwriteeofissurv.ethegafolloTwingincthehapters.thFandorrelatedaformoreridgedetailedroinformation,ofthe.readermeanistherecommendedthtowrittentheiscitedblitera-eenpreseneture.vFinallyoks,btheersdisturbeconsiderations3in