摘要
有偏估计是线性回归模型参数估计的一类重要方法,它能有效地克服设计阵X的多重共线性(Multicollinearity)问题。但是,有偏估计却不具稳健性。当数据中还含有少量的异常值(Outliers)时,估计的效果将受到严重干扰。为此,本文在有偏估计的特征根方法基础上,借助近年新兴的PP(Projection-Pursuit)的思想与方法,提出一种新的既能克服多重共线性影响又能抵抗异常值干扰的稳健的有偏估计方法。Monte Carlo模拟与实例应用结果均表明该法具有良好的特性。
Biased estimator is a kind of importent method widely used to estimate the regression coefficients in linear regression model. It can effectively overcome the difficulty caused by multicollinearity. However, the biased estimator is nonrobust, and can easily be influenced by outliers. In this paper, a new robust biased estimation method was developed based on the Webster's 'Latent Root Regression' and by means of the Projection Pursuit technique. The most attractive advantage of this method is that it can overcome the multicollinearity and resist to outliers simultanously. In the last part, a medical example is given.
出处
《中国卫生统计》
CSCD
北大核心
1991年第2期16-21,共6页
Chinese Journal of Health Statistics
关键词
回归分析
稳健估计
有偏估计
Regression analysis Projection pursuit Biased estimator Robust estimator Multicollinearity Outliers
