This was something we used a few years ago in one of our research projects and in the paper, Difficulty of selecting among multilevel models using predictive accuracy, with Wei Wang, but didn’t follow up on. I think it’s such a great idea I want to share it with all of you. We were applying …