[论文解读] Image-based Treatment Effect Heterogeneity

Connor Jerzak, Fredrik Johansson+1 more|arXiv (Cornell University)|Jun 13, 2022

Advanced Causal Inference Techniques被引用 4

一句话总结

本文提出一种基于变分贝叶斯推断的图像类型聚类模型，用于在随机对照试验（RCT）中识别图像数据中的处理效应异质性。通过建模能够预测异质性因果效应的图像衍生聚类，该方法实现了可解释的、带有不确定性感知的显著性分析，并支持样本外的政策目标定位，在乌干达一项反贫困RCT中，相较于仅使用表格数据的方法，显著提升了对异质性的检测能力。

ABSTRACT

Randomized controlled trials (RCTs) are considered the gold standard for estimating the average treatment effect (ATE) of interventions. One use of RCTs is to study the causes of global poverty -- a subject explicitly cited in the 2019 Nobel Memorial Prize awarded to Duflo, Banerjee, and Kremer "for their experimental approach to alleviating global poverty." Because the ATE is a population summary, anti-poverty experiments often seek to unpack the effect variation around the ATE by conditioning (CATE) on tabular variables such as age and ethnicity that were measured during the RCT data collection. Although such variables are key to unpacking CATE, using only such variables may fail to capture historical, geographical, or neighborhood-specific contributors to effect variation, as tabular RCT data are often only observed near the time of the experiment. In global poverty research, when the location of the experiment units is approximately known, satellite imagery can provide a window into such factors important for understanding heterogeneity. However, there is no method that specifically enables applied researchers to analyze CATE from images. In this paper, using a deep probabilistic modeling framework, we develop such a method that estimates latent clusters of images by identifying images with similar treatment effects distributions. Our interpretable image CATE model also includes a sensitivity factor that quantifies the importance of image segments contributing to the effect cluster prediction. We compare the proposed methods against alternatives in simulation; also, we show how the model works in an actual RCT, estimating the effects of an anti-poverty intervention in northern Uganda and obtaining a posterior predictive distribution over effects for the rest of the country where no experimental data was collected. We make all models available in open-source software.

研究动机与目标

• 解决在随机对照试验（RCT）中缺乏利用非结构化图像数据分析因果效应异质性的方法问题。
• 建模能够预测异质性处理效应的图像衍生聚类，以提升可解释性和科学洞察力。
• 利用贝叶斯推断实现聚类分配和效应预测中的不确定性量化。
• 开发显著性度量，识别在决定预测处理效应聚类中最具影响力的图像区域。
• 通过引入卫星图像作为效应修饰信息的来源，将因果推断从表格协变量扩展至更广泛的数据类型。

提出的方法

• 提出一种概率图像类型效应聚类模型，其中图像类型 $Z_i$ 决定处理效应 $\tau(z)$ 的分布，且 $\tau(z) = \mu_{\tau,z}$，方差为 $\sigma_0^2 + \sigma_1^2 + \sigma_{\tau}^2$。
• 使用变分贝叶斯推断近似后验分布 $p(\mathbf{Z}, \boldsymbol{\Theta} \mid \mathbf{D})$，通过随机梯度下降最大化证据下界（ELBO）。
• 应用重参数化梯度，实现对离散潜在变量（图像聚类）的反向传播，支持可微分推断。
• 采用蒙特卡洛近似计算期望显著性：$s_{whk}^{\text{Direction}} = \sum_c \mathbb{E}\left[ \frac{\partial \text{Pr}(Z_i = k \mid M_i = m)}{\partial m_{whc}} \right]$。
• 引入基于幅值的显著性度量 $s_{whk}^{\text{Magnitude}} = \sum_c \left( \frac{\partial \mathbb{E}[\text{Pr}(Z_i = k \mid M_i = m)]}{\partial m_{whc}} \right)^2$，以识别高影响力的图像区域。
• 通过 $p(\tau_i^{\text{Out}} \mid M_i^{\text{Out}}, \mathbf{D}) = \sum_z \int p(\tau_i \mid Z_i^\text{Out}=z; \boldsymbol{\theta}) p(Z_i^\text{Out}=z \mid M_i^\text{Out}; \boldsymbol{\theta}) p(\boldsymbol{\theta} \mid \mathbf{D}) d\boldsymbol{\theta}$ 生成样本外的处理效应预测分布。

实验结果

研究问题

• RQ1 基于图像的数据能否在随机对照试验（RCT）中揭示出表格协变量之外的、此前未被发现的处理效应异质性？
• RQ2 如何对图像衍生聚类进行建模，以概率化且可解释的方式总结和解释异质性处理效应？
• RQ3 哪些图像区域最影响对处理效应聚类的预测？我们如何量化其影响？
• RQ4 聚类分配中的不确定性如何影响显著性和政策建议的可靠性？
• RQ5 该模型能否推广至样本外图像，并支持在新地理或情境设置下的政策目标定位？

主要发现

• 该模型成功识别出能够预测乌干达一项反贫困RCT中异质性处理效应的图像基聚类，揭示了表格协变量未能捕捉到的异质性。
• 显著性分析识别出特定图像区域（尤其是土地利用和基础设施特征）显著影响预测的处理效应聚类。
• 在正交化后，聚类概率与表格CATE之间的相关性仍高达0.85，表明卫星图像为异质性提供了独立且非冗余的信息。
• 对全国范围的后验预测聚类概率分析揭示了处理效应异质性的空间模式，支持大规模政策相关的目标定位。
• 由于缺乏可微分的聚类分配，该模型的不确定性感知显著性度量在事后聚类方法中不可行。
• 该方法通过为新图像形成处理效应的预测分布，实现了样本外的政策目标定位，甚至无需依赖表格协变量。