Department of Education and Human Services, 1687Lehigh University, Bethlehem, PA, USA.
Department of Education Reform, 3341The University of Arkansas, Fayetteville, AR, USA.
Eval Rev. 2021 Feb-Apr;45(1-2):70-104. doi: 10.1177/0193841X211034363.
In randomized controlled trials, attrition rates often differ by treatment status, jeopardizing causal inference. Inverse probability weighting methods and estimation of treatment effect bounds have been used to adjust for this bias. We compare the performance of various methods within two samples, both generated through lottery-based randomization: one with considerable differential attrition and an augmented dataset with less problematic attrition. We assess the performance of various correction methods within the dataset with problematic attrition. In addition, we conduct simulation analyses. Within the more problematic dataset, we find the correction methods often performed poorly. Simulation analyses indicate that deviations from the underlying assumptions for bounding approaches damage the performance of estimated bounds. We recommend the verification of the underlying assumptions in attrition correction methods whenever possible and, when verification is not possible, using these methods with caution.
在随机对照试验中，失访率往往因治疗状况而异，从而危及因果推断。反概率加权法和治疗效果边界估计已被用于调整这种偏差。我们在两个样本中比较了各种方法的性能，这两个样本都是通过基于彩票的随机化产生的：一个样本的失访率差异较大，另一个样本的失访率较小。我们在存在问题的失访率数据集内评估了各种校正方法的性能。此外，我们还进行了模拟分析。在更具问题性的数据集内，我们发现校正方法的性能往往不佳。模拟分析表明，边界方法中偏离基本假设会损害估计边界的性能。我们建议在可能的情况下对失访校正方法的基本假设进行验证，并且在无法验证的情况下，谨慎使用这些方法。
