Bayesian Matrix Completion Approach to Causal Inference with Panel Data
Bayesian Matrix Completion Approach to Causal Inference with Panel Data
复制标题
DOI：
10.1007/s42519-021-00188-x
复制
发表时间：
2019-11
影响因子：
0.6
通讯作者：
Masahiro Tanaka
中科院分区：
文献类型：
--
作者：
Masahiro Tanaka
本研究提出了一种新的贝叶斯方法来推断二元治疗效果。该方法将反事实的未经处理的结果视为缺失的观察结果，并通过使用数据增强技术完成由已实现的和潜在的未经处理的结果组成的矩阵来推断它们。我们还开发了一个量身定制的先验知识，有助于识别参数，并诱导未经处理的结果矩阵近似为低秩。使用马尔可夫链蒙特卡罗采样器模拟后验绘制。虽然所提出的方法是类似的综合控制方法和其他相关的方法，它有几个显着的优点。首先，与合成控制方法不同，所提出的方法不需要严格的假设。第二，与非贝叶斯方法相比，所提出的方法可以以直接和一致的方式量化推理的不确定性。通过一系列的仿真研究，我们表明，我们的建议有一个更好的有限样本性能比现有的方法。
This study proposes a new Bayesian approach to infer binary treatment effects. The approach treats counterfactual untreated outcomes as missing observations and infers them by completing a matrix composed of realized and potential untreated outcomes using a data augmentation technique. We also develop a tailored prior that helps in the identification of parameters and induces the matrix of untreated outcomes to be approximately low rank. Posterior draws are simulated using a Markov Chain Monte Carlo sampler. While the proposed approach is similar to synthetic control methods and other related methods, it has several notable advantages. First, unlike synthetic control methods, the proposed approach does not require stringent assumptions. Second, in contrast to non-Bayesian approaches, the proposed method can quantify uncertainty about inferences in a straightforward and consistent manner. By means of a series of simulation studies, we show that our proposal has a better finite sample performance than that of the existing approaches.