Abstract

Nonignorable missingness and noncompliance can occur even in well-designed randomized experiments, making the intervention effect that the experiment was designed to estimate nonidentifiable. Nonparametric causal bounds provide a way to narrow the range of possible values for a nonidentifiable causal effect with minimal assumptions. We derive novel bounds for the causal risk difference for a binary outcome and intervention in randomized experiments with nonignorable missingness that is caused by a variety of mechanisms, with both perfect and imperfect compliance. We show that the so-called worst-case imputation, whereby all missing subjects on the intervention arm are assumed to have events and all missing subjects on the control or placebo arm are assumed to be event-free, can be too pessimistic in blinded studies with perfect compliance, and is not bounding the correct estimand with imperfect compliance. We illustrate the use of the proposed bounds in our motivating data example of peanut consumption on the development of peanut allergies in infants. We find that, even accounting for potentially nonignorable missingness and noncompliance, our derived bounds confirm that regular exposure to peanuts reduces the risk of development of peanut allergies, making the results of this study much more compelling.

摘要

即使在设计良好的随机实验中，也可能会出现不可忽略的缺失和不依从性，从而使实验旨在估计的干预效果无法识别。非参数因果界限提供了一种方法，可以使用最少的假设来缩小不可识别因果效应的可能值范围。我们推导出二元结果和随机实验干预的因果风险差异的新界限，这些实验具有由各种机制引起的不可忽略的缺失，具有完美和不完美的依从性。我们表明所谓的最坏情况归因，即假定干预组中所有缺失的受试者都有事件，而对照组或安慰剂组中所有缺失的受试者都假定为无事件，在盲法研究中可能过于悲观完美合规，并且不会以不完美的合规性来约束正确的估计值。我们在花生消费对婴儿花生过敏发展的激励数据示例中说明了建议界限的使用。我们发现，即使考虑到潜在的不可忽视的缺失和不依从性，我们得出的界限也证实经常接触花生会降低花生过敏的风险，使这项研究的结果更具说服力。