"Sequential experimental design for precise parameter estimation. 1. Use of reparameterization". Comments
"Sequential experimental design for precise parameter estimation. 1. Use of reparameterization". Comments
复制标题
“精确参数估计的顺序实验设计。1.使用重新参数化”。
DOI：
10.1021/i200035a034
复制
发表时间：
1986
期刊：
影响因子：
--
通讯作者：
D. Rippin
中科院分区：
文献类型：
--
作者：
T. Rimensberger;D. Rippin
Sir: Agarwal和Brisk（1985）描述了如何使用重参数化来提高参数估计的精度和顺序实验设计过程的有效性。序贯实验设计程序为下一个实验选择实验变量设置的值，在该值下，下一次实验后参数置信区域的期望体积最小。然而，众所周知，参数置信区域的体积对参数变换是不变的（Federov, 1972）。因此，预期置信区域体积是实验变量设置的函数，并且该函数与模型中使用的变量转换无关。这与Agarwal和Brisk在他们的案例研究中报告的经验相反。他们的表II显示，经过几个初始实验不同的实验变量设置，他们的模型1和2要求最小化预期的置信区域体积。
Sir: Agarwal and Brisk (1985) describehow a reparam-eterization can be used to improve the precision of parameter estimates and the effectiveness of a sequential experimental design procedure. The sequential experimental design procedure selects for the next experiment those values of the experimental variable settings at which the expected volume of the confidence region of the parameters after the next ex-periment is minimized. However, it is well-known that the volume of the confidence region of theparameters is invariant against parameter transformations (Federov, 1972). Thus, it is expected that the confidence region volume is a function of the experimental variable settings and that this function is independent of what variable transformations have been used in the model. This is contrary to the experience of Agarwal and Brisk as reported in their case study. Their Table II shows that after a few initial experiments different experimentalvariable settings are called for bytheir models 1 and 2 to minimize the expected confidence region volume.