WSEAS Transactions on Mathematics
Print ISSN: 1109-2769, E-ISSN: 2224-2880
Volume 12, 2013
Calibration Estimation via a Smoothing Newton Method
Authors: , ,
Abstract: Calibration estimation is currently the most popular method of estimation using auxiliary information. Its major idea is to use auxiliary information to structure calibration weights, attaching them to survey data, in or- der to improve the accuracy of the gross or mean estimation. Calibration estimation problem with box constraints is equivalently to solve a nonlinear equations system. Mnnich at el proposed a semismooth Newton method for solving this system in [Calibration of estimator-weights via semismooth Newton Method. J. Glob. Optim. 52 (2012):471-485]. In this paper, we give more specific analysis about the semismooth Newton method and some numerical experiments have been reported. On the basis of that, a smoothing Newton method has been proposed and proved to be globally convergent without any assumptions and locally superlinearly convergent under certain assumptions. Numerical results show that both semismooth Newton method and smoothing Newton method are effect for solving the calibration estimation problem.
Search Articles
Keywords: Calibration estimation, Auxiliary information, Sample weights, Semismooth Newton method, Smoothing Newton method, Global convergence