条件广义岭估计及其诊断统计量

Conditional Generalized Ridge-Type Estimation and Its Diagnostic Statistic

研究了带约束线性模型中回归参数的有偏估计及其数据删除模型中度量影响大小的诊断统计量.提出条件广义岭估计,给出它在不同模型中的关系式,利用Welsch-Kuh统计量和特征值的性质,推导出删除k组数据后度量拟合值影响的上确界.通过实例分析检验了诊断统计量的有效性.

This paper studies the biased estimation in linear model with equality restrictions and its diagnostic statistic which can assess the influence of deleting date in the date-deleted model.The conditional generalized ridge-type estimation is proposed,and its relation fomula in different models is given.The upper bound of influence in date-deleted model is derived by using Welsch-Kuh statistic and some properties of eigenvalue,and the effect of diagnostic statistic is examined by means of a numerical example.