A balanced approach to region estimation with tables for the normal model
作者:
T. L. Bratcher,
A. Hobbs,
J. Paul,
期刊:
Communications in Statistics - Simulation and Computation
(Taylor Available online 1984)
卷期:
Volume 13,
issue 6
页码: 801-821
ISSN:0361-0918
年代: 1984
DOI:10.1080/03610918408812416
出版商: Marcel Dekker, Inc.
关键词: confidence intervals;decision theory;minimum risk;expected size;joint estimation
数据来源: Taylor
摘要:
Practitioners of statistics are too often guilty of routinely selecting a 95% confidence level in interval estimation and ignoring the sample size and the expected size of the interval. One way to balance coverage and size is to use a loss function in a decision problem. Then either the Bayes risk or usual risk (if a pivotal quantity exists) may be minimized. It is found that some non-Bayes solutions are equivalent to Bayes results based on non-informative priors. The decision theory approach is applied to the mean and standard deviation of the univariate normal model and the mean of the multivariate normal. Tables are presented for critical values, expected size, confidence and sample size.
点击下载:
PDF (797KB)
返 回