Primary Submission Category: Measurement error and missing data
Measurement Error in Causal Inference: A Review
Authors: Keith Barnatchez, Kevin Josey, Rachel Nethery,
Presenting Author: Kevin Josey*
In both the scientific application and development of causal inference methods, it is often implicitly assumed that all relevant variables are measured without error. Despite the extensive literature studying the impact of measurement error in association studies, the development of methods at the intersection of measurement error and causal inference is in a relatively early, yet rapidly growing, stage. In this paper, we provide an overview of the burgeoning field of measurement error in causal inference. We detail the key role of study design in addressing measurement error, before examining a variety of methods for addressing confounder and exposure measurement error in causal inference studies, synthesizing the existing methods in measurement error correction with common causal estimators of the means of the potential outcomes. To facilitate the comparison of existing methods and development of future methods, we frame all methods in terms of causal assumptions and their associated study design requirements. After introducing the different choices for measurement error correction, we conduct a simulation study to evaluate their relative merits. We conclude with a set of recommendations for causal inference researchers suspecting measurement error in their analysis.