Optimal designs for cubic regression without an intercept

碩士 === 國立中山大學 === 應用數學研究所 === 84 === For cubic regression without an intercept on the symmetric interval $[-a,a]$, $phi_p$-optimal designs $(-inftyleq p leq 1)$ are computed for six parameter subsystems. For the present setting, it sufficies to consider a two-dimensional class of symmetric designs d...

Full description

Bibliographic Details
Main Authors: JU, JEHNG HER, 朱政和
Other Authors: CHANG, FU CHUEN
Format: Others
Language:en_US
Published: 1996
Online Access:http://ndltd.ncl.edu.tw/handle/63598631116679380690
Description
Summary:碩士 === 國立中山大學 === 應用數學研究所 === 84 === For cubic regression without an intercept on the symmetric interval $[-a,a]$, $phi_p$-optimal designs $(-inftyleq p leq 1)$ are computed for six parameter subsystems. For the present setting, it sufficies to consider a two-dimensional class of symmetric designs depending on two parameters. The $D$-, $E$- and $T$-optimal designs for all parameter subsystems are found and numerically optimal designs for some selected values of $a$ are computed. Some special properties of these optimal discussed. In particular, the $D$-optimal designs are invariant under scaling the interval. Moreover, the asymptotic properties of the optimal designs are also derived as $a o infty$ and $a o 0$. Finally, the efficiency of a $phi_p$- optimal design, when it compares with a $phi_q$-optimal design, is also investigated.