Don't Drop Noncompliers: Instrumental Variables and Encouragement Designs Explained Clearly
What do you do with participants who don't cooperate with instructions in a given condition? We examined an intention-to-treat (ITT) approach alongside what happens when noncomplying participants are dropped from the analysis to demonstrate how results change under these circumstances; specifically, how findings are pushed into statistical significance where there otherwise would not be any. But the problem with using an ITT approach is that it can weaken statistical power and make effect sizes appear smaller. As an alternate approach, we describe a way of keeping everyone while maintaining the ability to claim causality using a method borrowed from econometrics, an instrumental variable (IV). Using examples from previously published research, we explain this concept in nontechnical language, so that researchers in the psychological sciences can adapt and implement IV analyses in their own experiments, avoiding altogether the error of breaking randomization as a result of dropping participants.