Econ Papers Daily / 期刊论文
可视化合理的处理效应路径
(Visualizing) Plausible Treatment Effect Paths
摘要
We study inference on dynamic treatment effects. We propose simple additions to standard visualizations aimed at enhancing their informativeness. First, we recommend reporting joint hypothesis tests and summary measures that are often of direct interest alongside the standard event-study plot. Second, we introduce a data-driven procedure that estimates a smooth approximation to the treatment-effect path within a prespecified, economically motivated class and produces corresponding confidence regions. The resulting intervals can be substantially tighter than conventional uniform bands while maintaining formal coverage guarantees through post-selection inference (Berk et al. 2013). The accompanying R package plausibounds implements our suggestions.
中文摘要
我们研究动态处理效应的推断问题。我们提出对标准可视化方法的简单补充，旨在增强其信息量。首先，我们建议在标准事件研究图旁报告联合假设检验和通常直接相关的汇总度量。其次，我们引入一种数据驱动程序，在预设的、具有经济动机的类别内估计处理效应路径的光滑近似，并生成相应的置信区域。所得区间可能比常规均匀带更紧凑，同时通过选择后推断（Berk等人，2013）保持正式覆盖保证。随附的R包plausibounds实现了我们的建议。