Estimating the total treatment effect in randomized experiments with unknown network structure.
Estimating the total treatment effect in randomized experiments with unknown network structure.
复制标题
DOI：
10.1073/pnas.2208975119
复制
发表时间：
2022-11
影响因子：
11.1
通讯作者：
中科院分区：
文献类型：
--
作者：
在许多领域，我们希望在我们怀疑存在网络干扰的情况下估计总处理效果（TTE）。然而，我们经常无法测量网络或隐含的依赖结构。令人惊讶的是，我们能够在不了解网络的情况下开发设计随机实验的原则，这表明在合理的条件下，人们仍然可以估计TTE，考虑到未知网络上的干扰。所建议的设计原则和相关的估计器与广泛的结果模型一起工作。我们的估计器在简单的随机设计下具有低方差，从而在复杂网络效应存在的情况下有效实用地估计总处理效果。我们详细介绍了所提出的方法工作的假设，并讨论了它们可能失败的情况。在许多科学领域，从医学和医疗保健到物理和生物科学，从社会科学到工程学，从公共政策到技术产业，随机实验被广泛用于估计拟议治疗的因果效应。在这里，我们考虑的情况是，由于混淆网络效应，用于估计目标人群总治疗效果的经典方法存在相当大的偏差，即个体的治疗可能影响其邻居的结果，这一问题被称为网络干扰或非个体化治疗反应。在这些情况下的一个关键挑战是，网络通常是未知的，很难或昂贵的测量。我们假设了一个具有异质性可加性网络效应的潜在结果模型，该模型涵盖了广泛的网络干扰源，包括溢出效应、对等效应和传染。首先，我们描述了在不了解驱动干扰的网络的情况下估计总体治疗效果的局限性。相比之下，我们随后开发了一个简单的估计器和有效的随机设计，在实验之前获得平均历史基线测量的情况下，输出具有低方差的无偏估计。我们的解决方案不需要底层网络结构的知识，并且它为广泛的模型类别提供了统计保证。由于它们易于解释和实施，以及它们的理论保证，我们相信我们的结果将对随机实验的设计产生重大影响。
In many domains, we want to estimate the total treatment effect (TTE) in situations where we suspect network interference is present. However, we often cannot measure the network or the implied dependency structure. Surprisingly, we are able to develop principles for designing randomized experiments without knowledge of the network, showing that under reasonable conditions one can nonetheless estimate the TTE, accounting for interference on the unknown network. The proposed design principles, and related estimator, work with a broad class of outcome models. Our estimator has low variance under simple randomized designs, resulting in an efficient and practical solution for estimating total treatment effect in the presence of complex network effects. We detail the assumptions under which the proposed methods work and discuss situations when they may fail. Randomized experiments are widely used to estimate the causal effects of a proposed treatment in many areas of science, from medicine and healthcare to the physical and biological sciences, from the social sciences to engineering, and from public policy to the technology industry. Here we consider situations where classical methods for estimating the total treatment effect on a target population are considerably biased due to confounding network effects, i.e., the fact that the treatment of an individual may impact its neighbors’ outcomes, an issue referred to as network interference or as nonindividualized treatment response. A key challenge in these situations is that the network is often unknown and difficult or costly to measure. We assume a potential outcomes model with heterogeneous additive network effects, encompassing a broad class of network interference sources, including spillover, peer effects, and contagion. First, we characterize the limitations in estimating the total treatment effect without knowledge of the network that drives interference. By contrast, we subsequently develop a simple estimator and efficient randomized design that outputs an unbiased estimate with low variance in situations where one is given access to average historical baseline measurements prior to the experiment. Our solution does not require knowledge of the underlying network structure, and it comes with statistical guarantees for a broad class of models. Due to their ease of interpretation and implementation, and their theoretical guarantees, we believe our results will have significant impact on the design of randomized experiments.
登录
查看更多内容
影响因子：
5.4
作者：
Bowers, Jake;Fredrickson, Mark M.;Panagopoulos, Costas
通讯作者：
Panagopoulos, Costas
影响因子：
6.2
作者：
Cai, Jing;De Janvry, Alain;Sadoulet, Elisabeth
通讯作者：
Sadoulet, Elisabeth
影响因子：
1.4
作者：
Chin, Alex
通讯作者：
Chin, Alex
影响因子：
1.4
作者：
Eckles, Dean;Karrer, Brian;Ugander, Johan
通讯作者：
Ugander, Johan
影响因子：
1.9
作者：
Manski, Charles F.
通讯作者：
Manski, Charles F.