简介:ConsiderthepartitionedlinearregressionmodelA=(y,X1β1+X2β2,σ^2V)anditsfourreducedlinearmodels,whereyisann×1observablerandomvectorwithE(y)=XβanddispersionmatrixVar(y)=σ^2V,whereσ^2isanunknownpositivescalar,Visann×nknownsymmetricnonnegativedefinitematrix,X=(X1:X2)isann×(p+q)knowndesignmatrixwithrank(X)=r≤(p+q),andβ=(β'1:β'2)'withβ1andβ2beingp×1andq×1vectorsofunknownparameters,respectively.InthisarticletheformulaeforthedifferencesbetweenthebestlinearunbiasedestimatorsofM2X1β1underthemodelAanditsbestlinearunbiasedestimatorsunderthereducedlinearmodelsofAaregiven,whereM2=I-X2X2^+.Furthermore,thenecessaryandsufficientconditionsfortheequalitiesbetweenthebestlinearunbiasedestimatorsofM2X1β1underthemodelAandthoseunderitsreducedlinearmodelsareestablished.Lastly,wealsostudytheconnectionsbetweenthemodelAanditslineartransformationmodel.
简介:作为数学工作者和产业界专业人士共同探讨实际问题的数学建模及其解决方案的一种国际合作形式,诞生40多年的StudyGroupwithIndustry(工业与应用数学国际研讨会)活动遍及世界各大洲,为数学工作者参与高技术领域的创新、产业部门发现和培养高素质人才提供了有效的平台。在中国科学院院士李大潜、英国皇家学会会员(Fellow)J.Ockendon的大力推动下,这项活动在中国的影响力正在逐步扩大[1]。