COMPARINGRANDOMIZATIONANDARANDOMRUN…：比较的随机和随机运行…

《COMPARINGRANDOMIZATIONANDARANDOMRUN…：比较的随机和随机运行…》由会员上传分享，免费在线阅读，更多相关内容在教育资源-天天文库。

1、NewResultsaboutRandomizationandSplit-PlottingbyJamesM.Lucas2003Quality&ProductivityResearchConferenceYorktownHeights,NewYorkMay21-23,2003ContactInformationJamesM.LucasJ.M.LucasandAssociates5120NewKentRoadWilmington,DE19808(302)368-1214JamesM.Lucas@worldnet.att.net2J.M.LucasandAssociatesResearch

2、TeamHueyJuJeetuGanjuFrankAnbariPeterGoosMalcolmHazelDerekWebbJohnBorkowski3J.M.LucasandAssociatesPRELIMINARIESHowdoyourunExperiments?QUESTIONSHowmanyofyouareinvolvedwithrunningexperiments?Howmanyofyou“randomize”toguardagainsttrendsorotherunexpectedevents?Ifthesamelevelofafactorsuchastemperature

3、isrequiredonsuccessiveruns,howmanyofyousetthatfactortoaneutrallevelandresetit?5J.M.LucasandAssociatesADDITIONALQUESTIONSHowmanyofyouhaveconductedexperimentsonthesameprocessonwhichyouhaveimplementedaQualityControlProcedure?Whatdidyoufind?6J.M.LucasandAssociatesCOMPARINGRESIDUALSTANDARDDEVIATIONF

4、ROMANEXPERIMENTWITHRESIDUALSTANDARDDEVIATIONFROMANIN-CONTROLPROCESMYOBSERVATIONSEXPERIMENTALSTANDARDDEVIATIONISLARGER.1.5XTO3XISCOMMON.7J.M.LucasandAssociatesHOWSHOULDEXPERIMENTSBECONDUCTED?“COMPLETERANDOMIZATION”(andthecompletelyrandomizeddesign)RANDOMIZEDNOTRESET(AlsoCalledRandomRunOrder(RRO

5、)Experiments)(OftenAchievedWhenCompleteRandomizationisAssumed)SPLITPLOTBLOCKING(EspeciallyWhenThereareHard-to-ChangeFactors)8J.M.LucasandAssociatesRandomizedNotReset(RNR)ExperimentsAlargefraction(perhapsalargemajority)ofindustrialexperimentsareRandomizednotReset(RNR)experimentsPropertiesofRNRex

6、perimentsandadiscussionofhowexperimentsshouldbeconducted:“LkFactorialExperimentswithHard-to-ChangeandEasy-to-ChangeFactors”JuandLucas,2002,JQT34,411-421[studiesoneH-T-CfactorandusesRandomRunOrder(RRO)ratherthanRNR]“FactorialExperimentswhenFactorLevelsAreNotNecessarilyReset”Webb,LucasandBorkowsk

7、i,2003,JQT,toappear[studies>1HTCFactor]9J.M.LucasandAssociatesRNREXPERIMENTS(RandomRunOrderWithoutResettingFactors)OFTENUSEDBYEXPERIMENTERSNEVEREXPLICITLYRECOMMENDEDADVANTAGESOftenachievessuccessfulresultsCanbecost-effectiveDISADV