中文摘要：

在线性回归模型中，如果自变量存在样本归并，普通的LS估计不再一致，Rigobonand Stoker（2004，2007）建议使用部分样本回归或完整形式分析的方法来获得参数的一致估计，但是，它们都不是有效估计。本文使用基于EM算法的ML估计来获得参数的有效估计，在正态混合模型设定下详细推导了观测样本的似然函数以及相应的EM迭代方程。数值模拟的结果表明，基于EM算法的ML估计比部分样本回归和完整形式分析的方法具有更好的小样本表现。

英文摘要：

Since LS estimators are biased for linear models with censored regressors, Rigobon and Stoker （2004, 2007 ） suggest applying partly sample regression or complete case analysis to obtain consistent estimators of parameters. Unfortunately, neither their methods is efficient. In this paper, we apply ML estimation via the EM algorithm to obtain the efficient estimators, and calculate the sample likelihood function and the EM iteration equations in normal mixed-censoring models. Results of simulation show that the ML estimation has better small sample performance than partly sample regression or complete case analysis.