The linear regression model: Lpnorm estimation and the choice of p
作者:
A. H. money,
J. F. Affleck-Graves,
M. L. Hart,
G. D. I. Barr,
期刊:
Communications in Statistics - Simulation and Computation
(Taylor Available online 1982)
卷期:
Volume 11,
issue 1
页码: 89-109
ISSN:0361-0918
年代: 1982
DOI:10.1080/03610918208812247
出版商: Marcel Dekker, Inc.
关键词: generalised variance;linear regression;Lp: norm estimates;monte carlo simulation
数据来源: Taylor
摘要:
The LPnorm provides alternatives to least squares for estimating the coefficients of a linear regression model. Monte Carlo studies were performed using six different values of p (including the least squares case of p =2) and were compared by means of the relative efficiencies based on the generalised variances. Values of p other than 2 showed improvement over least squares for all non-normal error distributions. A rule for determining a suitable p on the basis of the kurtosis of the error distribution is proposed.
点击下载:
PDF (569KB)
返 回