scientific article; zbMATH DE number 4026531

From MaRDI portal
Publication:3768149





The main results in this paper are stated as follows. 1. Let Y be an n-vector such that \(EY=\theta\), \(DY=\sigma^ 2V\), \(V\geq 0\). Then under the quadratic loss function \((LY-S\theta)'(LY-S\theta)\), LY\(\sim S\theta\) iff (a) \(LVS'\) is symmetric, (b) \(LVL'\leq LVS'\), and (c) \(rk(L-S)V=rk(L-S)\), where \(LY\sim S\theta\) means that LY is an admissible estimate of \(S\theta\) in the class of all linear estimators. 2. Let \((Y,X\beta,\sigma^ 2V,V\geq 0)\) be a Gauss-Markov model and \(S\beta\) be estimable. Then under the matrix loss function \((LY- S\beta)(LY-S\beta)'/\sigma^ 2\), \(LY\sim S\beta\) iff (a) \(\mu\) (VL')\(\leq \mu (X)\), (b) \(LX=S\) or when \(LX\neq S\), \[ LXTX'L'-STS'+a(LX-S)T(LX- S)'\geq 0 \] does not hold for all \(a\in (0,1)\), where \[ T=X^+(V- VH(V^{1/2}H)^+V^{1/2})X'{}^+,\quad H=I-XX^+. \]




Cited in
(25)








This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3768149)