知识卡片
调整集合不能堆入所有变量
纳入中介可能阻断本想估计的总效应路径;纳入碰撞点或其后代可能打开非因果路径。调整依据是目标与图条件,不是变量是否预测结果,也不是“能加尽加”;有效集合的任意超集也未必有效。
这张卡片看不懂?
这张卡片已通过人工审查,不再接受失败反馈。
已有反馈 · 0 条
还没有人提交反馈。
在整书笔记中的位置
原书出处 · 1 处
- · 查看原书依据
In many causal inference texts, particularly from econometrics, the default approach to causal inference is regression—specifically, regressing the outcome on the treatment and any confounders we wish to adjust for or “control for.” What we are doing in this case is using linear regression to estimate the backdoor estimand.
参考来源
- 位置:《Causal AI》第11章《构建因果推断工作流》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Causal AI/_epub-src/OEBPS/Text/chapter-11.xhtml)
- 结论依据:摘录中关于调整集合不能堆入所有变量的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。
- 原始内容:In many causal inference texts, particularly from econometrics, the default approach to causal inference is regression—specifically, regressing the outcome on the treatment and any confounders we wish to adjust for or “control for.” What we are doing in this case is using linear regression to estimate the backdoor estimand.