目录 Adjusting for Selection Bias Conditioning on a Mediator Adjusting for Selection Bias 不幸的是,纠正选择偏倚绝非易事。在我们一直在讨论的例子中,即使有随机对照试验,ATE也无法识别,仅仅是因为你无法在对那些回应了调查的人进行条件化后,关闭新功能与客户满意度之间的非因果关联流。为了取得一些
凡是包含because, since, for, if, thus, hence, so that, so...that..., so...as to..., therefore, consequently, as a result, accordingly, in that, result in, result from, cause, give rise to, derive from,
目录 What is Non-Parametric About? What is Non-Parametric About? 在我们继续之前,我只想强调一个常见的误解。当我们考虑使用非参数 Double-ML 模型来估计 CATE 时,我们似乎会得到一个非线性治疗效果。例如,让我们假设一个非常简单的数据生成过程(DGP),其中 discont 对销售额的影响是非线性的,但却是通过
目录 The average treatment effect Difference-in-means estimation IID Sampling and Population Asymptotics Example: The linear model Regression adjustments with a linear model 随机对照试验(RCT)是统计因果推论