Discussion questions#
You run a study on classroom distractions and learning. If your dataset contains exactly two classrooms and you only care about those two rooms, how should classroom be treated in the model? If you want to say something about classrooms in general, how should it be treated?
Yarkoni (2019) argued that mixed effects models should be standard practice for increasing generalizability of findings in psychology and neuroscience.
What problem are mixed-effects models trying to fix in traditional statistical analyses?
How do random effects actually change what a study’s conclusion means?
Why might a statistically “worse” model produce scientifically better evidence?
Under what circumstances could mixed-effects models reduce generalizability instead of improve it?