Chapter 9 Understanding Type I error inflation using simulation

The simplest situation we encountered in the earliest chapters was to carry out a single hypothesis test, having set Type I error to 0.05. In the linear mixed model setting, even this simple scenario is fraught with danger; Type I error can become startlingly large depending on the kind of linear mixed model you fit. We begin this chapter by illustrating this problem through an example of a simple two-condition experiment. However, there are at least two other sources of Type I error inflation: multiple comparison, and model misspecification. We illustrate these issues below with an example from a \(2\times 2\times 2\) repeated measures design.