可交换条件下具有测量误差的结构回归模型

Structural Regression Model with Measurement Errors under Exchangeability Conditions

摘要: 提出具有测量误差的结构回归模型，研究可交换条件下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．如果暴露组与对照组中协变量的测量误差同分布，那么测量误差只对平均处理效应估计的精度有影响．

Abstract: A structural regression model with measurement errors is proposed to study the impact of measurement errors on estimating the average treatment effect under exchangeability conditions. Without other additional conditions, the average treatment effect is still identified even though most of the population parameters could not be identified. If the distribution of the covariate measurement errors in the exposed group is the same as that under control, it will only have some influence on the preciseness of estimating the average treatment effect.