A note on \(E\)-optimal designs for weighted polynomial regression (Q688382): Difference between revisions

This note generalizes the results of \textit{F. Pukelsheim} and \textit{W. J. Studden} in ibid., No. 1, 402-415 (1993) (see the preceding review) and allows specific nonconstant variances of different observations, namely \(\sigma^ 2/\lambda(x)\) with \[ \lambda(x)=(1+x)^ u(1-x)^ v;\;u,v\in\{0,1\};\;x\in[-1,1]. \] An example is given that the support points of an \(E\)-optimal design cannot be transformed linearly when the design space is changed from \([-1,1]\) to \([-b,b]\).