Noncompliance and Instrumental Variables for 2^<i>K</i> Factorial Experiments

Factorial experiments are widely used to assess the marginal, joint, and interactive effects of multiple concurrent factors. While a robust literature covers design analysis these experiments, there is less work on how handle treatment noncompliance in this setting. To fill gap, we introduce new methodology that uses potential outcomes framework for analyzing 2K factorial with any number This builds extends both instrumental variables several ways. First, define novel, complier-specific quantities interest setting show generalize key assumptions. Second, partial compliance across factors gives researchers choice over different types compliers target estimation. Third, conduct inference estimands from finite-population superpopulation asymptotic perspectives. Finally, illustrate techniques by applying them field experiment effectiveness forms get-out-the-vote canvassing. New easy-to-use, open-source software implements methodology. Supplementary materials article available online.