Hey #rstats and #stats people, does anyone know of any references that explain the hazards of trying to run or interpret models with all combinations of many predictors? (i.e., including multiple 2-way and 3-way interactions)