Effects of the starting value and stopping rule on robust estimates obtained by iterated weighted least squares
作者:
Jeffrey B. Birch,
期刊:
Communications in Statistics - Simulation and Computation
(Taylor Available online 1980)
卷期:
Volume 9,
issue 2
页码: 141-154
ISSN:0361-0918
年代: 1980
DOI:10.1080/03610918008812144
出版商: Taylor & Francis Group
关键词: m-estimator;rates of convergence;relative change stopping rule;average number of iterations to convergence;first-order auto-regressive model
数据来源: Taylor
摘要:
To obtain certain estimators in statistical models, for example, M-estimators, it is often necessary to use an iterative computational technique. One such technique is the iterated weighted least squares method, while another is the well-known Newton1s method. Both methods require a starting value as well as a stopping rule incorporated into their iterative scheme. This paper attempts to indicate the possible effects of using various combinations of starting values and stopping rules on M-estimators, especially when applied to results of Monte Carlo studies. In this paper, H-estimators are computed for the first-order autoregressive model
点击下载:
PDF (323KB)
返 回