Learning When-to-Treat Policies
Learning When-to-Treat Policies
复制标题
DOI：
10.1080/01621459.2020.1831925
复制
发表时间：
2020-11-28
影响因子：
3.7
通讯作者：
Wager, Stefan
中科院分区：
文献类型：
--
作者：
Nie, Xinkun;Brunskill, Emma;Wager, Stefan
许多应用决策问题都有一个动态的组成部分：决策者不仅需要选择谁来治疗，而且还需要选择何时开始治疗。例如，医生可以在推迟治疗（观察等待）和在患者多次就诊期间开出几种可用治疗之一之间进行选择。我们开发了一个“优势双鲁棒”的估计学习这样的动态治疗规则，使用观测数据的假设下，顺序可扩展性。我们证明了福利遗憾界，推广双鲁棒学习的结果，在单步设置，并在几个不同的情况下表现出有前途的经验表现。我们的方法对于策略优化是实用的，并且不需要任何结构性的（例如，Markovian）假设。可以在网上找到。
Many applied decision-making problems have a dynamic component: The policymaker needs not only to choose whom to treat, but also when to start which treatment. For example, a medical doctor may choose between postponing treatment (watchful waiting) and prescribing one of several available treatments during the many visits from a patient. We develop an "advantage doubly robust" estimator for learning such dynamic treatment rules using observational data under the assumption of sequential ignorability. We prove welfare regret bounds that generalize results for doubly robust learning in the single-step setting, and show promising empirical performance in several different contexts. Our approach is practical for policy optimization, and does not need any structural (e.g., Markovian) assumptions. for this article are available online.