超饱和设计数据新地分析方法.pdf

1 Introduction 1 Introduction industrial containa ManypreliminaryscreeningexperimentstypicaUy large few numberof relevant a arebelievedtobe potentiallyfactors．Amongthem，only The constructed large supersa七urated active(haviIlgefrects) designs(SSDs)，nrst Boothand receiveda dealofattention systelnaticallyby Cox(1962)，haSgreat sincethe of After areseveralnewmethodsof appearanceLin(1993)that，there the include constructingssDs(choosingdesignmatrix)．TheseLin(1995)，、Vu and (1993)，Nguyen(1996)，cheng(1997)，Liwu(1997)，TangandⅥ～(1997)， Liuand eta1． soon．Nowit出soextendsto zhang(2000)，Butler(2001)and includeYamadaand and mixed—levelssDs Lin(2002)，YamadaMatsui(2002)， etal etal et et Fhng (2003)，Fhng(2004)，Lia1．(2004)，1ramadaa1．(2006)and Liuet this focuson oftwo-level al_(2006)．Inpaper，we analyzing supersaturated designs isa becausechenumberofeX- Analyzingsupersaturateddesignschallenge islessthanthenumberoffactors．‘rb丘ndthe active periments sparse effbcts， variable becomesfundamentalinthe ofsuch ex— selection analysisstage screening isas吼lmedthatthereisnointeractionbetweenthefactorsandthe periments．1t modelfor efrectislinearinthelevel