"Using independent covariates in experimental designs: Quantifying the tradeoff between power boost and Type I error inflation"
The practice of using covariates in experimental designs is controversial. Traditionally touted by statisticians as a useful method to soak up noise in a dependent variable and boost power, the practice has recently been recast in a negative light because of its inflation of Type I error rate. In order to make informed decisions about practices like this one, researchers need to know more about the actual size of the benefits and costs of these practices. In this talk, I present simulations on the impact of flexibly analyzing data with an unanticipated, independent covariate on power and Type I error rate, and I offer recommendations on when and how to use independent covariates in experimental designs. I close with a consideration of the tradeoff between power boost and Type I error inflation when using covariates in nonexperimental designs, particularly in the context of incremental validity testing.
Meeting ID: 961 8485 4291