Overfitting is a common problem in data analysis. Some go as far as saying that “most of” published research is false (John Ionnadis); overfitting being one, maybe central, problem of it. In this post, we explore some aspects on the notion of overfitting. Assume we have 10 metric variables v (personality/health/behavior/gene indicator variables), and, say, 10 variables for splitting up subgroups (aged vs. young, female vs. male, etc.), so 10 dichotomic …