控制图的发展

[e-mail]1Levey-JenningsLeveyJennings1950/n=2—Xbar—R11.11924100ISO/GB1.21940—501924SPC1940—1950ManualonthePresentationofData(ASTM)Z1.1Z1.2Z1.31.312n=2-2.11n=2Xbar-R1946Wernimont,G.,UseofControlChartsintheAnalyticalLaboratory,Ind.Eng.Chem.Anal.Ed.,18:587-592(1946).ISSN-0019-78662GoogleWernimont,G.,ControlChartsWernimont3Wernimont,G.,n=2Xbar—RsX3±WernimontWernimontn=2XbarRn5XbarR2.2/1n=2Xbar-R1950Levey,S.,andJennings,E.R.,TheUseofControlChartsintheClinicalLaboratory.Amer.J.Clin.Path.,20:1059-1066(1950)LeveyJenningsWernimontn=[e-mail]22LeveyJenningsLeveyJennings1950/Levey,S.,andJennings,E.R./WernimontLevey,S.,andJennings,E.RLeveyJennings1[PubMed]LeveyJennings2LeveyJenningsLeveyandJennings[XbarR][+/-3s][XbarR][+/-3s]!1982ISO825819821991ISO82581991n=2Xbar-RISOLevey/Jennings3ISOn=2XbarRInternationalOrganizationforStandardizationISO1982ISO8258ShewhartcontrolchartsISO/TC691983GB4091.1GB4091.91983Shewhartcontrolcharts1991ISO8258ISO1991ISO82581991Shewhartcontrolcharts2001GB/T4091—2001ShewhartcontrolchartsISO82581991194650n=2Xbar—RISOWernimont,G.LeveyJennings3XbarRXbar3s1Levey-Jennings[Xbar3s][e-mail]3n=2Xbar—RHenry-SegaloveIHenry-SegaloveL-JHenry-SegaloveLevey-JenningsXbar3s21952Levey-Jennings/1990ISO/GBLevey-Jennings[Xbar3s]Henry-SegaloveI[Xbar3s]Grubbsoutliers1990ISO/GBStatisticsProcesscontrolSPC/ISO9000ISO/GB6SigmaSPC-6Sigma/ISO/GB1n=2Xbar—R[Xbar3s]/21-16Sigma36Sigma/!239899SPSSX-BarRsIndividualMovingRange-pnpcu2000MINITABLevey-Jennings1950ISO8258:1991ShewhartcontrolchartsGB4091—2001-X-BarR[Xbar+/-3s][e-mail]4HenrySegalove19529541[Xbar3s]ISO/GB[Xbar3s]Levey-JenningsSPCSASSPSSMINITABHenrySegalove1952[Xbar3s][Xbar3s]outliersHenry-SegaloveIoutliersMINITABBox-Cox2outliersoutlierscontrolchartsOutliersGrubbsDescriptivestatisticsExploreoutliersoutliers!outliersGrubbsoutliersnt-0.050.01GGrubbsoutlierst-outliersGtGrubbst-t-Grubbsoutlierst-3[OutliersControlcharts][e-mail]526Sigma3ShewhartcontrolchartsstatisticalqualitycontrolchartsqualitycontrolchartscontrolchartsSPCstatisticalprocesscontrol4://[Xbar+/-3s]Levey-JenningsHenry-SegaloveI+/-34123Grubbst-4Google55[]1.................................................................................................................................................526SIGMASPC..................................................83GRUBBST-................................................................................................................................104GOOGLE............................................................................................................................................125........................................................................................................[e-mail][e-mail][e-mail][e-mail][e-mail]103Grubbst-GrubbstestforOutliersPurpose:DetectionofOutliersGrubbs'test(Grubbs1969andStefansky1972)isusedtodetectoutliersinaunivariatedataset.Itisbasedontheassumptionofnormality.Thatis,youshouldfirstverifythatyourdatacanbereasonablyapproximatedbyanormaldistributionbeforeapplyingtheGrubbs'test.DefinitionGrubbs'testisdefinedforthehypothesis:H0:TherearenooutliersinthedatasetHa:sYYGi−=maxwhereYandsarethesamplemeanandstandarddeviation.TheGrubbsteststatisticisthelargestabsolutedeviationfromthesamplemeaninunitsofthesamplestandarddeviation.SignificanceLevel:αCriticalRegion:ThehypothesisofnooutliersisrejectedifG2)2,/(2)2,/(2)1(−−+−−NNNNtNtNNααwhere2)2,/(−NNtαisthecriticalvalueofthet-distributionwith(N-2)degreesoffreedomandasignificancelevelofN/ααtα−[e-mail]bbs'testfindsnooutliersinthedataset:InterpretationofSampleOutputTheoutputisdividedintothreesections.1.ThefirstsectionprintsthesamplestatisticsusedinthecomputationoftheGrubbs'testandthevalueoftheGrubbs'teststatistic.2.ThesecondsectionprintstheuppercriticalvaluefortheGrubbs'teststatisticdistributioncorrespondingtovarioussignificancelevels.Thevalueinthefirstcolumn,theconfidencelevelofthetest,isequivalentto100(α).WerejectthenullhypothesisatthatsignificancelevelifthevalueoftheGrubbs'teststatisticprintedinsectiononeisgreaterthanthecriticalvalueprintedinthelastcolumn.3.Thethirdsectionprintstheconclusionfora95%test.Foradifferentsignificancelevel,theappropriateconclusioncanbedrawnfromthetableprintedinsectiontwo.Forexample,forα=0.10,welookattherowfor90%confidenceandcomparethecriticalvalue3.24totheGrubbs'teststatistic2.92.Sincetheteststatisticislessthanthecriticalvalue,weacceptthenullhypothesisattheα=0.10level.Outputfromotherstatisticalsoftwaremaylooksomewhatdifferentfromtheaboveoutput.QuestionsGrubbs'testcanbeusedtoanswerth