Gauss因果模型中因果效应识别方法的比较

Comparing Identifiability Criteria for Causal Effects in Gaussian Causal Models

一个Gauss因果模型中常常存在不只一种识别因果效应的方法,不同的方法对应的估计可能不同.该文对Pearl等人提出的前门准则,后门准则,工具变量准则等识别方法的估计精度进行了分析比较,并给出了相应的模拟结果,为实践中选择更优的识别准则提供了依据.

In a Gaussian causal model, there may exist several criteria for the identification of the same causal effect, but the corresponding estimators may be different from each other. This paper consideres three popular identifiability criteria, front-door, back-door and instrumental variables criteria for causal effects, compares their estimate efficieneies and gives some simulations. These results can offer guidances for choosing better identifiability criterion in practice.