OCIS · Fall 2020¶
- 共 14 场 · 14 篇精读
本季导览¶
自动生成：归纳本季主线与值得先看的几场，不打分、不排名。
这一季 OCIS Fall 2020 的 14 场报告可归纳为四条主线：（1）随机化实验中的设计与分析（Luke Miratrix、Fan Li、Peng Ding、Esther Duflo 访谈），（2）选择偏误与缺失数据的因果视角（Qingyuan Zhao、Karthika Mohan & David Hirshberg），（3）因果效应识别与估计的图模型与半参数方法（Vanessa Didelez、Emilija Perkovic、Joris Mooij、Nathan Kallus & Xiaojie Mao），（4）异质性处理效应与可解释性（Falco Bargagli Stoffi & Eli Ben-Michael）。此外，两场思想性访谈（Judea Pearl、Donald Rubin）和一场方法论辩论（David Blei 的 Deconfounder）贯穿多条主线，提供历史与批判性视角。
主线一：随机化实验的设计与分析 是这一季最密集的主题。Luke Miratrix 利用英格兰国家学生数据库进行元分析，量化教育评估中未观测混杂与损耗偏差的实际规模，属于实证偏差基准化。Fan Li 和 Peng Ding 分别从加权和回归调整两条路径提升 RCT 的精度：Li 提出重叠权重（overlap weights）作为比 IPW 更稳定的协变量调整工具，Ding 则系统比较 rerandomization（设计阶段）与含交互项的回归调整（分析阶段）的联合效率。Esther Duflo 的访谈从实践者角度补充了 RCT 在发展中经济学的操作挑战与政策转化缺口。
主线二：选择偏误与缺失数据 被 Qingyuan Zhao 和 Karthika Mohan & David Hirshberg 从不同角度切入。Zhao 以 COVID-19 增长率和 incubation period 的早期误估为案例，展示选择偏误在热点问题中的灾难性影响，呼吁从第一性原理校正。Mohan & Hirshberg 则回到平衡权重方法论，讨论如何通过直接优化协变量矩平衡（而非先估计倾向性评分）来建立处理组与对照组的可比性，属于稳定平衡权重（stable balancing weights）的推广。
主线三：图模型与半参数效率 覆盖识别与估计两个层面。Emilija Perkovic 处理 MPDAG（部分有向图）中因果效应的识别与有效估计，推广了广义调整准则。Joris Mooij 的 Joint Causal Inference（JCI）将因果发现统一为多来源数据（观测+实验）的联合推断框架，提升可识别性。Nathan Kallus & Xiaojie Mao 的 Localized DML 则解决分位数处理效应（QTE）中干扰项依赖参数的问题，通过局部化 Neyman 正交性实现高效推断。Vanessa Didelez 专门讨论生存分析中的因果 estimand 定义，指出 hazard ratio 因 survivor selection bias 而缺乏因果解释，并比较中介分析与竞争事件处理的替代方案。
主线四：异质性处理效应与可解释性 由 Falco Bargagli Stoffi 的 Causal Rule Ensemble（CRE）和 Eli Ben-Michael 的 Partially Pooled Synthetic Control 组成。CRE 在因果森林的高精度与单棵决策树的可解释性之间取得平衡，输出“如果-那么”规则；Ben-Michael 则针对交错采纳（staggered adoption）的面板数据，用部分池化合成控制估计 ATT，兼顾精度与稳定性。
推荐入口：若想快速把握这一季的技术核心，建议按以下路径：（1）RCT 设计分析：先看 Peng Ding（rerandomization + 回归调整）和 Fan Li（重叠权重），再读 Luke Miratrix（实证偏差基准）作为背景；（2）选择偏误与缺失数据：从 Qingyuan Zhao（案例驱动）入手，再读 Karthika Mohan & David Hirshberg（平衡权重方法论）；（3）图模型与半参数效率：先看 Emilija Perkovic（MPDAG 识别）和 Joris Mooij（JCI 统一框架），再读 Nathan Kallus & Xiaojie Mao（Localized DML）作为进阶；（4）异质性处理效应：从 Falco Bargagli Stoffi & Eli Ben-Michael（CRE + 合成控制）开始。两场访谈（Pearl、Rubin）和 Deconfounder 辩论（Blei）适合作为思想性补充，不依赖技术细节。
报告列表¶
Using national data and meta-analysis techniques to get a handle on how bad some biases might be in practice¶
讲者: Luke Miratrix · 讨论人: Elizabeth Tipton · 2020-12-15
链接：视频 · 幻灯片
摘要
Different designs come with different risks for bias and researchers and funding agencies have limited data on the magnitude of these different biases. How much do unobserved factors bias quasi-experimental designs in education evaluations? How problematic is attrition bias in randomized experiments? Across two projects we make use of a unique administrative dataset in England, linked to a large archive of RCTs. This allows us to complete two types of “within study comparison” analyses, one to s…Selection bias in 2020¶
讲者: Qingyuan Zhao · 讨论人: Louisa Smith · 2020-12-08
链接：视频 · 幻灯片
摘要
This talk will examine the selection bias that occurred in studying some most contentious problems in 2020. In the first case study, we will look at the estimation of the growth rate and incubation period of COVID-19 and demonstrate how early studies drastically misestimated them. In the second case study, we will review and hopefully clarify a recent debate on post-treatment selection in studying racial discrimination in policing. In the era of data, causal inference researchers are uniquely po…Causal reasoning in survival and time-to-event analyses¶
讲者: Vanessa Didelez · 讨论人: Els Goetghebeur · 2020-12-01
链接：视频 · 幻灯片
摘要
In this talk I will discuss why causal inference should pay special attention to survival and time-to-event settings. Even in an apparently simple case of a randomized point-treatment it is common that events other than the event of interest occur, sometimes called intercurrent events such as (semi-)competing events or time-varying mediators, and of course censoring. The choice of causal estimand in such situations should anticipate these issues and suitably represent the research question. Rece…Propensity score weighting for covariate adjustment in randomized clinical trials¶
讲者: Fan Li · 讨论人: Kari Lock Morgan · 2020-11-24
链接：视频 · 幻灯片 · arXiv
摘要
Chance imbalance in baseline characteristics is common in randomized clinical trials. Regression adjustment such as the analysis of covariance (ANCOVA) is often used to account for imbalance and increase precision of the treatment effect estimate. An objective alternative is through inverse probability weighting (IPW) of the propensity scores. Although IPW and ANCOVA are asymptotically equivalent, the former may demonstrate inferior performance in finite samples. In this article, we point out th…- Data versus Science: Contesting the Soul of Data-Science - Radical Empiricism and Machine Learning Research - What Statisticians Want to Know about Causal Inference and The Book of Why (Interview by David Hand) - The Seven Tools of Causal Inference with Reflections on Machine Learning¶
讲者: Interview with Judea Pearl · 2020-11-17
链接：视频 · 幻灯片 · arXiv
ocis-2020-11-03-interview-with-donald-rubin¶
讲者: Interview with Donald Rubin · 2020-11-03
链接：视频
The Deconfounder: What is it? What is its theory? Is it useful?¶
讲者: David Blei · 讨论人: Guido Im bens · 2020-10-27
链接：视频 · 幻灯片 · arXiv
摘要
I will discuss the deconfounder algorithm and the assumptions it requires. Several refinements have been suggested around the theory of the deconfounder. Among these, Imai and Jiang clarified the assumption of "no unobserved single-cause confounders." Using their assumption, I will clarify the theory. Finally, I will discuss whether the deconfounder is useful in practice. This talk will largely follow Wang and Blei (2020).ocis-2020-10-12-interview-with-esther-duflo¶
讲者: Interview with Esther Duflo · 2020-10-12
链接：视频
Randomization and Regression Adjustment¶
讲者: Peng Ding · 讨论人: Tirthankar DasGupta · 2020-10-07
链接：视频 · 幻灯片 · arXiv
Causal effects in maximally oriented partially directed acyclic graphs (MPDAGs): Identification and efficient estimation¶
讲者: Emilija Perkovic · 讨论人: Thomas Richardson · 2020-09-29
链接：视频 · 幻灯片 · arXiv
Causal Rule Ensemble: Interpretable Inference of Heterogeneous Treatment Effects¶
讲者: Falco Bargagli Stoffi ; Eli Ben-Michael · 2020-09-23
链接：视频 · 幻灯片
摘要
In environmental epidemiology, it is critically important to identify subpopulations that are most vulnerable to the adverse effects of air pollution so we can develop targeted interventions. In recent years, there have been many methodological developments for addressing heterogeneity of treatment effects in causal inference. A common approach is to estimate the conditional average treatment effect (CATE) for a pre-specified covariate set. However, this approach does not provide an easy-to-inte…Localized Debiased Machine Learning: Efficient Inference on Quantile Treatment Effects and Beyond¶
讲者: Nathan Kallus and Xiaojie Mao · 讨论人: Alex andre Belloni · 2020-09-15
链接：视频 · 幻灯片 · arXiv
摘要
We consider the efficient estimation of a low-dimensional parameter in an estimating equation involving high-dimensional nuisances that depend on the parameter of interest. An important example is the (local) quantile treatment effect ((L)QTE) in causal inference, where the efficient estimating equation involves as a nuisance the covariate-conditional cumulative distribution function evaluated at the quantile to be estimated. Debiased machine learning (DML) is a data-splitting approach to addres…Joint Causal Inference: A Unifying Perspective on Causal Discovery¶
讲者: Joris Mooij · 讨论人: Philip Dawid · 2020-09-08
链接：视频 · 幻灯片
摘要
Many questions in science, policy making and everyday life are of a causal nature: how would a change of A affect B? An important research topic is therefore how cause-effect relationships can be discovered from data and how these can be used for making predictions in situations where a system has been perturbed by an external intervention. In this talk, I will introduce the basics of two, apparently quite different, approaches to causal discovery. I will discuss how both approaches can be elega…Causal Graphical Models for Handling Missing Data¶
讲者: Karthika Mohan ; David Hirshberg · 2020-09-02
链接：视频 · 幻灯片
摘要
In causal inference, various notions of comparability between samples are used to justify the interpretation of observed differences as causal. Collectively, these are called balance . By reweighting units, we can establish balance between otherwise incomparable samples. We will discuss this tradition, and recent generalizations that allow the estimation of a variety of causal summaries.Maintained by 陈星宇 · Homepage · Source