为限制异常数据对回归模型预报值的干扰,提出预报值影响有界的回归估计量。基于稳健 统计的有关结果,证明了这些估计量的最优性和可容许性,进一步建立了Bp-稳健 性与Vp-稳健性之间的关系。本文结果能有效地控制异常数据对预报值的影响,在数据 受污染时的回归分析问题中有广泛的应用价值。%For various reasons, abnormal values exist in engineering data and ordinary regr ession methods often lead to serious errors. We propose to reduce such errors by a new method developed by us after several years of research on application of robust statistics. Our new method employs a class of regression estimators which possess bounded predicted-value influence. Using robust statistics, we prove the optimality and admissibility of these estimators and establish the relations hip between Bp-robustness and Vp-robustness.
展开▼