多维协变量具有测量误差的结构回归模型

Structural Regression Model with Measurement Errors of Covariates

提出具有测量误差的结构回归模型,研究可交换条件下多维协变量的测量误差对 平均处理效应估计的影响.在没有其它的附加条件下,尽管大多数模型参数不可识别,平均处理 效应仍可识别.由于平均处理效应的极大似然估计求解困难,建议在实际中使用拟极大似然估计 作为替代.

A structural regression model with measurement errors is established to study the impact of measurement errors of covariates on estimating the average treatment effect under exchangeability conditions. Without other additional conditioins, the average treatment effect is still identified even though most of the population parameters could not be identified. The quasi-maximum likelihood estimator of the average treatment effect is proposed to use in practice instead of the maximum likelihood estimator because of the complicated computation.