Modeling is a central part not only of statistical inquiry, but also of everyday human sense-making. We use models as metaphors for the world, in a broader sense. Of course, a model that explains the world better (than some other model) is to be preferred, all other things being equal. In this post, we demonstrate that a more “clever” statistical model reduces the residual variance. It should be noted that this “noise reduction” comes at a cost, however: The model gets more complex; there a more parameters in the …