Calibration using a piecewise simple linear regression model
Calibration using a piecewise simple linear regression model
复制标题
DOI：
10.1081/sta-100002027
复制
发表时间：
2001-01-01
影响因子：
0.8
通讯作者：
Preater, J
中科院分区：
文献类型：
--
作者：
Ndlovu, P;Preater, J
我们证明经典校准方法（[1]）是平行拟合方法（[2]）的一个特例，经典校准设计是平行拟合设计的一个组成部分。这种标定方法之间的关系促使将经典标定的优化设计扩展到平行拟合的优化设计。该扩展涉及将平行拟合设计中的经典校准设计组件限制为经典校准设计的最优设计，然后优化平行拟合设计中剩余组件的选择。通过对各自最优设计标定方法的比较，表明平行拟合方法比经典标定方法具有更强的鲁棒性和更高的估计效率。仿真研究表明，渐近校正设计对于小样本校正问题的估计也是有效的。
We show that the classical calibration method([1]) is a special case of the parallel fitting method ([2]) and that the design for classical calibration is a component of the design for parallel fitting. This relationship between the calibration methods motivates extending the optimal designs for classical calibration to designs for parallel fitting. The extension involves restricting the classical calibration design component in the design for parallel fitting to an optimal design for classical calibration and then optimizing the choice of the remaining components of the design for parallel fitting. A comparison of the calibration methods at their respective optimal designs shows that the parallel fitting method is more robust and more efficient for estimation than the classical calibration. A simulation study shows that the asymptotic designs for calibration are also effective for estimation in small sample calibration problems.