This paper argues that the sometimes-conflicting results of a modern revisionist literature on data mining in econometrics reflect different approaches to solving the central problem of model uncertainty in a science of non-experimental data. The literature has entered an exciting phase with...