

ChemometricsandIntelligentLaboratorySystems52 2000.87–104
doc.guandang.net
Multipleoutlierdetectionformultivariatecalibrationusingrobust
statisticaltechniques
RandyJ.Pell)
AnalyticalSciencesLaboratory,TheDowChemicalCompany,1897SBuilding,Midland,MI48667USA
Received14February2000;receivedinrevisedform24May2000;accepted25May2000
Abstract
Outliersthatareincorporatedintoamultivariatecalibrationmodelcansignificantlyreducetheperformanceofthemodel.Inthecaseofmultipleoutliers,thestandardmethodsforoutlierdetectioncanfailtodetecttrueoutliersandevenmistakenlyidentifygoodsamplesasoutliers.Robuststatisticalmethodsarelesssensitivetooutliersandcanprovideapowerfultoolforthereliabledetectionofmultipleoutliers.Thispaperexaminestheuseofrobustprincipalcomponentregression PCR.anditerativelyreweightedpartialleastsquares PLS.formultipleoutlierdetectioninaninfraredspectroscopicapplication.q2000ElsevierScienceB.V.Allrightsreserved.
Keywords:Robustregression;IterativelyreweightedPLS;Leasttrimmedsquares;PLS;PCA;Resamplingbyhalfmean;Outliers
1.Introduction
Buildinghigh-qualitymultivariatecalibrationmodelsdependsontheexecutionofseveralstepsw1x.Oneofthemostimportantofthesestepsisoutlierdetection.Soimportantisoutlierdetectioninmulti-variatecalibrationthatMartensandNæsw2xdevoteanentirechaptertothissubjectintheirbook.Theydis-cussthestandarddiagnostictoolsforbothcalibrationandpredictionoutlierdetection.Whilethesetoolsarequitepowerfulfordetectionofpredictionoutliersandsinglecalibrationoutliers,theyareinefficientatbestandcanbemisleadingatworstforthedetectionofmultipleoutliersinthecalibrationphase.Detectionof
)
Tel.:q1-517-638-4967;fax:q1-517-636-4882.E-mailaddress:pellrj@dow.com R.J.Pell..
multipleoutliersincalibrationusingthestandarddi-agnosticscanbeseverelylimitedduetomaskingandswamping.Maskingisthecasewhereoneormoreoutliersareincorrectlyidentifiedasnormalsamplesbecauseotheroutliersaremaskingtheirpresence.Swampingisthecasewherenormalsamplesaremadetoappeartobeoutliers.
Itmayseemimprobablethatmultipleoutlierswouldbepresentindatacollectedforacalibrationexperiment,giventhehighdegreeofcontrolnor-mallyaccompanyingsuchexperiments.Thismaybethecaseforlaboratorycalibrationexperiments;butfordatacollectedfromexperimentsinwhichthean-alysthaslesscontrol,itisreasonabletoexpectthatoutlierswillbemoreprevalent.Theapplicationofmultivariateanalysistechniquestodatacollectedfromon-lineanalyticalinstrumentationordatafromcollectionsofunivariateprocesssensors e.g.tem-
0169-7439r00r$-seefrontmatterq2000ElsevierScienceB.V.Allrightsreserved.PII:S0169-7439 00.00082-4
4.1 Statistical Methods for Multivariate Outlier Detection Multivariate outlier ...tractability of multiple outlier detection methods is the use of biased ...
Multiple outlier detec... 15页 1下载券 OUTLIER DETECTION 11页 免费 ...Probability density function of a multivariate normal di t ib ti distribution...
Outlier Detection A brief introduction Fuzhou University 智能技术课程---D.Y....dataset must be known in prior.(2)not suitable for multivariate applications...
Multiple outlier detecti... 15页 1财富值 Online Outlier Detection... 12...cation of outliers in multivariate data. Outliers often provide useful ...
Density-based trajectory outlier detection algorithm_材料科学_工程科技_专业资料。Journal of Systems Engineering and Electronics Vo1.24,N...
Risk pre-warning of tender evaluation for civil projectsan outlier detection model_专业资料。The marking scheme method removes the low scores of the ...
Many algorithms have been proposed in recent years for outlier detection 7, 8, 10, 22, 23, 25, 26], but they are not methods which are speci ...
1财富值 Multiple outlier detecti... 15页 1财富值 Outlier Detection for Tr... 暂无评价 8页 免费 Shape Outlier Detection ... 暂无评价 8页 免费...
Therefore multivariate techniques must be employed, and PCA has been one of the most accepted methods for outlier detection (Daszykowski et al., 2007; ...
A data set may have multiple types of outlier One object may belong to ...Transform the multivariate outlier detection task into a univariate outlier ...
热门文档