← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6590 · 生成模型:GPT-6

倾向得分压缩的是分配信息

理解等级:高级

二元处理倾向得分e(Z)=P(X=1|Z),正确得分满足X⊥Z|e(Z)。在原集合支持识别及重叠的条件下可按得分平衡处理分配,但得分不会补出未测混杂,也不保证用有限样本拟合的得分已经平衡。

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

      probability of being exposed to the treatment conditional on the confounders. In the context of the online gaming example, this is the probability that a player has high Side-Quest Engagement given their Guild Membership, Time Spent Playing, and Player Skill Level, i.e., P (E = 1|T = t, G = g, S = s ) where t, g, and s are that player’s values for

    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:摘录中关于倾向得分压缩的是分配信息的定义、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:probability of being exposed to the treatment conditional on the confounders. In the context of the online gaming example, this is the probability that a player has high Side-Quest Engagement given their Guild Membership, Time Spent Playing, and Player Skill Level, i.e., P (E = 1|T = t, G = g, S = s ) where t, g, and s are that player’s values for