知识卡片

普通读书笔记卡 · 6588 · 生成模型:GPT-6

模型与观测列共同决定识别

理解等级:中级

因果图刻画DGP假设,数据列决定哪些变量可观察。没有观测某个上游共同原因,不一定破坏识别:若其到处理和结果的后门路径已被可观测集合阻断,仍可调整;应按路径检验而非统计“缺少几个变量”。

这张卡片看不懂?

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

      Our level 1 observational distribution includes all the variables in the DAG except Prior Experience. Thus, Prior Experience is a latent variable (figure 11.4). Figure 11.4 Prior Experience is not observed in the data; it is a latent (unobserved) variable with respect to our DAG. We specified the base distribution for the estimand using y0’s domain-specific language for probabilistic expressions: Identification.from_expression( graph=dag, query=query, estimand=observational_distribution )

    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:摘录中关于模型与观测列共同决定识别的定义、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:Our level 1 observational distribution includes all the variables in the DAG except Prior Experience. Thus, Prior Experience is a latent variable (figure 11.4). Figure 11.4 Prior Experience is not observed in the data; it is a latent (unobserved) variable with respect to our DAG. We specified the base distribution for the estimand using y0’s domain-specific language for probabilistic expressions: Identification.from_expression( graph=dag, query=query, estimand=observational_distribution )